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 "$" from matching | |
1e4aa4a5 AP |
34 | the start or end of line only at the left and right ends of the string to |
35 | matching them anywhere within the string. | |
55497cff | 36 | |
54310121 | 37 | =item s |
d74e8afc ITB |
38 | X</s> X<regex, single-line> X<regexp, single-line> |
39 | X<regular expression, single-line> | |
55497cff | 40 | |
41 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 42 | whatsoever, even a newline, which normally it would not match. |
55497cff | 43 | |
34d67d80 | 44 | Used together, as C</ms>, they let the "." match any character whatsoever, |
fb55449c | 45 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 46 | and just before newlines within the string. |
7b8d334a | 47 | |
87e95b7f YO |
48 | =item i |
49 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> | |
50 | X<regular expression, case-insensitive> | |
51 | ||
52 | Do case-insensitive pattern matching. | |
53 | ||
5027a30b KW |
54 | If locale matching rules are in effect, the case map is taken from the |
55 | current | |
17580e7a | 56 | locale for code points less than 255, and from Unicode rules for larger |
ed7efc79 KW |
57 | code points. However, matches that would cross the Unicode |
58 | rules/non-Unicode rules boundary (ords 255/256) will not succeed. See | |
59 | L<perllocale>. | |
60 | ||
61 | There are a number of Unicode characters that match multiple characters | |
62 | under C</i>. For example, C<LATIN SMALL LIGATURE FI> | |
63 | should match the sequence C<fi>. Perl is not | |
64 | currently able to do this when the multiple characters are in the pattern and | |
65 | are split between groupings, or when one or more are quantified. Thus | |
66 | ||
67 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches | |
68 | "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match! | |
69 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match! | |
70 | ||
71 | # The below doesn't match, and it isn't clear what $1 and $2 would | |
72 | # be even if it did!! | |
73 | "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match! | |
74 | ||
1f59b283 KW |
75 | Perl doesn't match multiple characters in an inverted bracketed |
76 | character class, which otherwise could be highly confusing. See | |
77 | L<perlrecharclass/Negation>. | |
78 | ||
25275596 KW |
79 | Another bug involves character classes that match both a sequence of |
80 | multiple characters, and an initial sub-string of that sequence. For | |
81 | example, | |
82 | ||
83 | /[s\xDF]/i | |
84 | ||
85 | should match both a single and a double "s", since C<\xDF> (on ASCII | |
86 | platforms) matches "ss". However, this bug | |
87 | (L<[perl #89774]|https://rt.perl.org/rt3/Ticket/Display.html?id=89774>) | |
88 | causes it to only match a single "s", even if the final larger match | |
89 | fails, and matching the double "ss" would have succeeded. | |
90 | ||
1f59b283 | 91 | Also, Perl matching doesn't fully conform to the current Unicode C</i> |
ed7efc79 KW |
92 | recommendations, which ask that the matching be made upon the NFD |
93 | (Normalization Form Decomposed) of the text. However, Unicode is | |
94 | in the process of reconsidering and revising their recommendations. | |
87e95b7f | 95 | |
54310121 | 96 | =item x |
d74e8afc | 97 | X</x> |
55497cff | 98 | |
99 | Extend your pattern's legibility by permitting whitespace and comments. | |
ed7efc79 | 100 | Details in L</"/x"> |
55497cff | 101 | |
87e95b7f YO |
102 | =item p |
103 | X</p> X<regex, preserve> X<regexp, preserve> | |
104 | ||
632a1772 | 105 | Preserve the string matched such that ${^PREMATCH}, ${^MATCH}, and |
87e95b7f YO |
106 | ${^POSTMATCH} are available for use after matching. |
107 | ||
e2e6bec7 DN |
108 | =item g and c |
109 | X</g> X</c> | |
110 | ||
111 | Global matching, and keep the Current position after failed matching. | |
112 | Unlike i, m, s and x, these two flags affect the way the regex is used | |
113 | rather than the regex itself. See | |
114 | L<perlretut/"Using regular expressions in Perl"> for further explanation | |
115 | of the g and c modifiers. | |
116 | ||
b6fa137b FC |
117 | =item a, d, l and u |
118 | X</a> X</d> X</l> X</u> | |
119 | ||
516074bb KW |
120 | These modifiers, all new in 5.14, affect which character-set semantics |
121 | (Unicode, etc.) are used, as described below in | |
ed7efc79 | 122 | L</Character set modifiers>. |
b6fa137b | 123 | |
55497cff | 124 | =back |
a0d0e21e | 125 | |
516074bb KW |
126 | Regular expression modifiers are usually written in documentation |
127 | as e.g., "the C</x> modifier", even though the delimiter | |
b6fa137b | 128 | in question might not really be a slash. The modifiers C</imsxadlup> |
ab7bb42d | 129 | may also be embedded within the regular expression itself using |
ed7efc79 | 130 | the C<(?...)> construct, see L</Extended Patterns> below. |
b6fa137b | 131 | |
ed7efc79 KW |
132 | =head3 /x |
133 | ||
b6fa137b | 134 | C</x> tells |
7b059540 | 135 | the regular expression parser to ignore most whitespace that is neither |
55497cff | 136 | backslashed nor within a character class. You can use this to break up |
4633a7c4 | 137 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 138 | character is also treated as a metacharacter introducing a comment, |
55497cff | 139 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 140 | whitespace or C<#> characters in the pattern (outside a character |
f9a3ff1a | 141 | class, where they are unaffected by C</x>), then you'll either have to |
7b059540 KW |
142 | escape them (using backslashes or C<\Q...\E>) or encode them using octal, |
143 | hex, or C<\N{}> escapes. Taken together, these features go a long way towards | |
8933a740 RGS |
144 | making Perl's regular expressions more readable. Note that you have to |
145 | be careful not to include the pattern delimiter in the comment--perl has | |
146 | no way of knowing you did not intend to close the pattern early. See | |
147 | the C-comment deletion code in L<perlop>. Also note that anything inside | |
7651b971 | 148 | a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect |
0b928c2f | 149 | space interpretation within a single multi-character construct. For |
7651b971 | 150 | example in C<\x{...}>, regardless of the C</x> modifier, there can be no |
9bb1f947 | 151 | spaces. Same for a L<quantifier|/Quantifiers> such as C<{3}> or |
f9e949fd KW |
152 | C<{5,}>. Similarly, C<(?:...)> can't have a space between the C<?> and C<:>, |
153 | but can between the C<(> and C<?>. Within any delimiters for such a | |
154 | construct, allowed spaces are not affected by C</x>, and depend on the | |
155 | construct. For example, C<\x{...}> can't have spaces because hexadecimal | |
156 | numbers don't have spaces in them. But, Unicode properties can have spaces, so | |
0b928c2f | 157 | in C<\p{...}> there can be spaces that follow the Unicode rules, for which see |
9bb1f947 | 158 | L<perluniprops/Properties accessible through \p{} and \P{}>. |
d74e8afc | 159 | X</x> |
a0d0e21e | 160 | |
ed7efc79 KW |
161 | =head3 Character set modifiers |
162 | ||
163 | C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called | |
164 | the character set modifiers; they affect the character set semantics | |
165 | used for the regular expression. | |
166 | ||
808432af KW |
167 | The C</d>, C</u>, and C</l> modifiers are not likely to be of much use |
168 | to you, and so you need not worry about them very much. They exist for | |
169 | Perl's internal use, so that complex regular expression data structures | |
170 | can be automatically serialized and later exactly reconstituted, | |
171 | including all their nuances. But, since Perl can't keep a secret, and | |
172 | there may be rare instances where they are useful, they are documented | |
173 | here. | |
ed7efc79 | 174 | |
808432af KW |
175 | The C</a> modifier, on the other hand, may be useful. Its purpose is to |
176 | allow code that is to work mostly on ASCII data to not have to concern | |
177 | itself with Unicode. | |
ca9560b2 | 178 | |
808432af KW |
179 | Briefly, C</l> sets the character set to that of whatever B<L>ocale is in |
180 | effect at the time of the execution of the pattern match. | |
ca9560b2 | 181 | |
808432af KW |
182 | C</u> sets the character set to B<U>nicode. |
183 | ||
184 | C</a> also sets the character set to Unicode, BUT adds several | |
185 | restrictions for B<A>SCII-safe matching. | |
186 | ||
187 | C</d> is the old, problematic, pre-5.14 B<D>efault character set | |
188 | behavior. Its only use is to force that old behavior. | |
189 | ||
190 | At any given time, exactly one of these modifiers is in effect. Their | |
191 | existence allows Perl to keep the originally compiled behavior of a | |
192 | regular expression, regardless of what rules are in effect when it is | |
193 | actually executed. And if it is interpolated into a larger regex, the | |
194 | original's rules continue to apply to it, and only it. | |
195 | ||
196 | The C</l> and C</u> modifiers are automatically selected for | |
197 | regular expressions compiled within the scope of various pragmas, | |
198 | and we recommend that in general, you use those pragmas instead of | |
199 | specifying these modifiers explicitly. For one thing, the modifiers | |
200 | affect only pattern matching, and do not extend to even any replacement | |
201 | done, whereas using the pragmas give consistent results for all | |
202 | appropriate operations within their scopes. For example, | |
203 | ||
204 | s/foo/\Ubar/il | |
205 | ||
206 | will match "foo" using the locale's rules for case-insensitive matching, | |
207 | but the C</l> does not affect how the C<\U> operates. Most likely you | |
208 | want both of them to use locale rules. To do this, instead compile the | |
209 | regular expression within the scope of C<use locale>. This both | |
210 | implicitly adds the C</l> and applies locale rules to the C<\U>. The | |
211 | lesson is to C<use locale> and not C</l> explicitly. | |
212 | ||
213 | Similarly, it would be better to use C<use feature 'unicode_strings'> | |
214 | instead of, | |
215 | ||
216 | s/foo/\Lbar/iu | |
217 | ||
218 | to get Unicode rules, as the C<\L> in the former (but not necessarily | |
219 | the latter) would also use Unicode rules. | |
220 | ||
221 | More detail on each of the modifiers follows. Most likely you don't | |
222 | need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead | |
223 | to L<E<sol>a|/E<sol>a (and E<sol>aa)>. | |
ca9560b2 | 224 | |
ed7efc79 KW |
225 | =head4 /l |
226 | ||
227 | means to use the current locale's rules (see L<perllocale>) when pattern | |
228 | matching. For example, C<\w> will match the "word" characters of that | |
229 | locale, and C<"/i"> case-insensitive matching will match according to | |
230 | the locale's case folding rules. The locale used will be the one in | |
231 | effect at the time of execution of the pattern match. This may not be | |
232 | the same as the compilation-time locale, and can differ from one match | |
233 | to another if there is an intervening call of the | |
b6fa137b | 234 | L<setlocale() function|perllocale/The setlocale function>. |
ed7efc79 KW |
235 | |
236 | Perl only supports single-byte locales. This means that code points | |
237 | above 255 are treated as Unicode no matter what locale is in effect. | |
238 | Under Unicode rules, there are a few case-insensitive matches that cross | |
239 | the 255/256 boundary. These are disallowed under C</l>. For example, | |
516074bb KW |
240 | 0xFF (on ASCII platforms) does not caselessly match the character at |
241 | 0x178, C<LATIN CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be | |
242 | C<LATIN SMALL LETTER Y WITH DIAERESIS> in the current locale, and Perl | |
243 | has no way of knowing if that character even exists in the locale, much | |
244 | less what code point it is. | |
ed7efc79 KW |
245 | |
246 | This modifier may be specified to be the default by C<use locale>, but | |
247 | see L</Which character set modifier is in effect?>. | |
b6fa137b FC |
248 | X</l> |
249 | ||
ed7efc79 KW |
250 | =head4 /u |
251 | ||
252 | means to use Unicode rules when pattern matching. On ASCII platforms, | |
253 | this means that the code points between 128 and 255 take on their | |
808432af KW |
254 | Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's). |
255 | (Otherwise Perl considers their meanings to be undefined.) Thus, | |
256 | under this modifier, the ASCII platform effectively becomes a Unicode | |
257 | platform; and hence, for example, C<\w> will match any of the more than | |
258 | 100_000 word characters in Unicode. | |
ed7efc79 KW |
259 | |
260 | Unlike most locales, which are specific to a language and country pair, | |
516074bb KW |
261 | Unicode classifies all the characters that are letters I<somewhere> in |
262 | the world as | |
ed7efc79 KW |
263 | C<\w>. For example, your locale might not think that C<LATIN SMALL |
264 | LETTER ETH> is a letter (unless you happen to speak Icelandic), but | |
265 | Unicode does. Similarly, all the characters that are decimal digits | |
266 | somewhere in the world will match C<\d>; this is hundreds, not 10, | |
267 | possible matches. And some of those digits look like some of the 10 | |
268 | ASCII digits, but mean a different number, so a human could easily think | |
269 | a number is a different quantity than it really is. For example, | |
270 | C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an | |
271 | C<ASCII DIGIT EIGHT> (U+0038). And, C<\d+>, may match strings of digits | |
272 | that are a mixture from different writing systems, creating a security | |
67592e11 | 273 | issue. L<Unicode::UCD/num()> can be used to sort |
516074bb KW |
274 | this out. Or the C</a> modifier can be used to force C<\d> to match |
275 | just the ASCII 0 through 9. | |
ed7efc79 | 276 | |
516074bb KW |
277 | Also, under this modifier, case-insensitive matching works on the full |
278 | set of Unicode | |
ed7efc79 KW |
279 | characters. The C<KELVIN SIGN>, for example matches the letters "k" and |
280 | "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which, | |
281 | if you're not prepared, might make it look like a hexadecimal constant, | |
282 | presenting another potential security issue. See | |
283 | L<http://unicode.org/reports/tr36> for a detailed discussion of Unicode | |
284 | security issues. | |
285 | ||
6368643f KW |
286 | On the EBCDIC platforms that Perl handles, the native character set is |
287 | equivalent to Latin-1. Thus this modifier changes behavior only when | |
ed7efc79 KW |
288 | the C<"/i"> modifier is also specified, and it turns out it affects only |
289 | two characters, giving them full Unicode semantics: the C<MICRO SIGN> | |
6368643f | 290 | will match the Greek capital and small letters C<MU>, otherwise not; and |
ed7efc79 KW |
291 | the C<LATIN CAPITAL LETTER SHARP S> will match any of C<SS>, C<Ss>, |
292 | C<sS>, and C<ss>, otherwise not. | |
293 | ||
294 | This modifier may be specified to be the default by C<use feature | |
66cbab2c KW |
295 | 'unicode_strings>, C<use locale ':not_characters'>, or |
296 | C<L<use 5.012|perlfunc/use VERSION>> (or higher), | |
808432af | 297 | but see L</Which character set modifier is in effect?>. |
b6fa137b FC |
298 | X</u> |
299 | ||
ed7efc79 KW |
300 | =head4 /d |
301 | ||
302 | This modifier means to use the "Default" native rules of the platform | |
303 | except when there is cause to use Unicode rules instead, as follows: | |
304 | ||
305 | =over 4 | |
306 | ||
307 | =item 1 | |
308 | ||
309 | the target string is encoded in UTF-8; or | |
310 | ||
311 | =item 2 | |
312 | ||
313 | the pattern is encoded in UTF-8; or | |
314 | ||
315 | =item 3 | |
316 | ||
317 | the pattern explicitly mentions a code point that is above 255 (say by | |
318 | C<\x{100}>); or | |
319 | ||
320 | =item 4 | |
b6fa137b | 321 | |
ed7efc79 KW |
322 | the pattern uses a Unicode name (C<\N{...}>); or |
323 | ||
324 | =item 5 | |
325 | ||
326 | the pattern uses a Unicode property (C<\p{...}>) | |
327 | ||
328 | =back | |
329 | ||
330 | Another mnemonic for this modifier is "Depends", as the rules actually | |
331 | used depend on various things, and as a result you can get unexpected | |
808432af KW |
332 | results. See L<perlunicode/The "Unicode Bug">. The Unicode Bug has |
333 | become rather infamous, leading to yet another (printable) name for this | |
334 | modifier, "Dodgy". | |
ed7efc79 KW |
335 | |
336 | On ASCII platforms, the native rules are ASCII, and on EBCDIC platforms | |
337 | (at least the ones that Perl handles), they are Latin-1. | |
338 | ||
339 | Here are some examples of how that works on an ASCII platform: | |
340 | ||
341 | $str = "\xDF"; # $str is not in UTF-8 format. | |
342 | $str =~ /^\w/; # No match, as $str isn't in UTF-8 format. | |
343 | $str .= "\x{0e0b}"; # Now $str is in UTF-8 format. | |
344 | $str =~ /^\w/; # Match! $str is now in UTF-8 format. | |
345 | chop $str; | |
346 | $str =~ /^\w/; # Still a match! $str remains in UTF-8 format. | |
347 | ||
808432af KW |
348 | This modifier is automatically selected by default when none of the |
349 | others are, so yet another name for it is "Default". | |
350 | ||
351 | Because of the unexpected behaviors associated with this modifier, you | |
352 | probably should only use it to maintain weird backward compatibilities. | |
353 | ||
354 | =head4 /a (and /aa) | |
355 | ||
356 | This modifier stands for ASCII-restrict (or ASCII-safe). This modifier, | |
357 | unlike the others, may be doubled-up to increase its effect. | |
358 | ||
359 | When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and | |
360 | the Posix character classes to match only in the ASCII range. They thus | |
361 | revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d> | |
362 | always means precisely the digits C<"0"> to C<"9">; C<\s> means the five | |
363 | characters C<[ \f\n\r\t]>; C<\w> means the 63 characters | |
364 | C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as | |
365 | C<[[:print:]]> match only the appropriate ASCII-range characters. | |
366 | ||
367 | This modifier is useful for people who only incidentally use Unicode, | |
368 | and who do not wish to be burdened with its complexities and security | |
369 | concerns. | |
370 | ||
371 | With C</a>, one can write C<\d> with confidence that it will only match | |
372 | ASCII characters, and should the need arise to match beyond ASCII, you | |
373 | can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are | |
374 | similar C<\p{...}> constructs that can match beyond ASCII both white | |
375 | space (see L<perlrecharclass/Whitespace>), and Posix classes (see | |
376 | L<perlrecharclass/POSIX Character Classes>). Thus, this modifier | |
377 | doesn't mean you can't use Unicode, it means that to get Unicode | |
378 | matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that | |
379 | signals Unicode. | |
380 | ||
381 | As you would expect, this modifier causes, for example, C<\D> to mean | |
382 | the same thing as C<[^0-9]>; in fact, all non-ASCII characters match | |
383 | C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary | |
384 | between C<\w> and C<\W>, using the C</a> definitions of them (similarly | |
385 | for C<\B>). | |
386 | ||
387 | Otherwise, C</a> behaves like the C</u> modifier, in that | |
388 | case-insensitive matching uses Unicode semantics; for example, "k" will | |
389 | match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code | |
390 | points in the Latin1 range, above ASCII will have Unicode rules when it | |
391 | comes to case-insensitive matching. | |
392 | ||
393 | To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>), | |
394 | specify the "a" twice, for example C</aai> or C</aia>. (The first | |
395 | occurrence of "a" restricts the C<\d>, etc., and the second occurrence | |
396 | adds the C</i> restrictions.) But, note that code points outside the | |
397 | ASCII range will use Unicode rules for C</i> matching, so the modifier | |
398 | doesn't really restrict things to just ASCII; it just forbids the | |
399 | intermixing of ASCII and non-ASCII. | |
400 | ||
401 | To summarize, this modifier provides protection for applications that | |
402 | don't wish to be exposed to all of Unicode. Specifying it twice | |
403 | gives added protection. | |
404 | ||
405 | This modifier may be specified to be the default by C<use re '/a'> | |
406 | or C<use re '/aa'>. If you do so, you may actually have occasion to use | |
407 | the C</u> modifier explictly if there are a few regular expressions | |
408 | where you do want full Unicode rules (but even here, it's best if | |
409 | everything were under feature C<"unicode_strings">, along with the | |
410 | C<use re '/aa'>). Also see L</Which character set modifier is in | |
411 | effect?>. | |
412 | X</a> | |
413 | X</aa> | |
414 | ||
ed7efc79 KW |
415 | =head4 Which character set modifier is in effect? |
416 | ||
417 | Which of these modifiers is in effect at any given point in a regular | |
808432af KW |
418 | expression depends on a fairly complex set of interactions. These have |
419 | been designed so that in general you don't have to worry about it, but | |
420 | this section gives the gory details. As | |
ed7efc79 KW |
421 | explained below in L</Extended Patterns> it is possible to explicitly |
422 | specify modifiers that apply only to portions of a regular expression. | |
423 | The innermost always has priority over any outer ones, and one applying | |
6368643f KW |
424 | to the whole expression has priority over any of the default settings that are |
425 | described in the remainder of this section. | |
ed7efc79 | 426 | |
916cec3f | 427 | The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set |
ed7efc79 KW |
428 | default modifiers (including these) for regular expressions compiled |
429 | within its scope. This pragma has precedence over the other pragmas | |
516074bb | 430 | listed below that also change the defaults. |
ed7efc79 KW |
431 | |
432 | Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>; | |
66cbab2c | 433 | and C<L<use feature 'unicode_strings|feature>>, or |
ed7efc79 KW |
434 | C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to |
435 | C</u> when not in the same scope as either C<L<use locale|perllocale>> | |
66cbab2c KW |
436 | or C<L<use bytes|bytes>>. |
437 | (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also | |
438 | sets the default to C</u>, overriding any plain C<use locale>.) | |
439 | Unlike the mechanisms mentioned above, these | |
6368643f KW |
440 | affect operations besides regular expressions pattern matching, and so |
441 | give more consistent results with other operators, including using | |
442 | C<\U>, C<\l>, etc. in substitution replacements. | |
ed7efc79 KW |
443 | |
444 | If none of the above apply, for backwards compatibility reasons, the | |
445 | C</d> modifier is the one in effect by default. As this can lead to | |
446 | unexpected results, it is best to specify which other rule set should be | |
447 | used. | |
448 | ||
449 | =head4 Character set modifier behavior prior to Perl 5.14 | |
450 | ||
451 | Prior to 5.14, there were no explicit modifiers, but C</l> was implied | |
452 | for regexes compiled within the scope of C<use locale>, and C</d> was | |
453 | implied otherwise. However, interpolating a regex into a larger regex | |
454 | would ignore the original compilation in favor of whatever was in effect | |
455 | at the time of the second compilation. There were a number of | |
456 | inconsistencies (bugs) with the C</d> modifier, where Unicode rules | |
457 | would be used when inappropriate, and vice versa. C<\p{}> did not imply | |
458 | Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12. | |
b6fa137b | 459 | |
a0d0e21e LW |
460 | =head2 Regular Expressions |
461 | ||
04838cea RGS |
462 | =head3 Metacharacters |
463 | ||
384f06ae | 464 | The patterns used in Perl pattern matching evolved from those supplied in |
14218588 | 465 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
466 | (distantly) from Henry Spencer's freely redistributable reimplementation |
467 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
468 | details. | |
a0d0e21e LW |
469 | |
470 | In particular the following metacharacters have their standard I<egrep>-ish | |
471 | meanings: | |
d74e8afc ITB |
472 | X<metacharacter> |
473 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
474 | ||
a0d0e21e | 475 | |
f793d64a KW |
476 | \ Quote the next metacharacter |
477 | ^ Match the beginning of the line | |
478 | . Match any character (except newline) | |
479 | $ Match the end of the line (or before newline at the end) | |
480 | | Alternation | |
481 | () Grouping | |
482 | [] Bracketed Character class | |
a0d0e21e | 483 | |
14218588 GS |
484 | By default, the "^" character is guaranteed to match only the |
485 | beginning of the string, the "$" character only the end (or before the | |
486 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
487 | assumption that the string contains only one line. Embedded newlines |
488 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 489 | string as a multi-line buffer, such that the "^" will match after any |
0d520e8e YO |
490 | newline within the string (except if the newline is the last character in |
491 | the string), and "$" will match before any newline. At the | |
a0d0e21e LW |
492 | cost of a little more overhead, you can do this by using the /m modifier |
493 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
db7cd43a | 494 | but this option was removed in perl 5.10.) |
d74e8afc | 495 | X<^> X<$> X</m> |
a0d0e21e | 496 | |
14218588 | 497 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 498 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
f02c194e | 499 | the string is a single line--even if it isn't. |
d74e8afc | 500 | X<.> X</s> |
a0d0e21e | 501 | |
04838cea RGS |
502 | =head3 Quantifiers |
503 | ||
a0d0e21e | 504 | The following standard quantifiers are recognized: |
d74e8afc | 505 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e | 506 | |
f793d64a KW |
507 | * Match 0 or more times |
508 | + Match 1 or more times | |
509 | ? Match 1 or 0 times | |
510 | {n} Match exactly n times | |
511 | {n,} Match at least n times | |
512 | {n,m} Match at least n but not more than m times | |
a0d0e21e | 513 | |
0b928c2f FC |
514 | (If a curly bracket occurs in any other context and does not form part of |
515 | a backslashed sequence like C<\x{...}>, it is treated | |
9af81bfe KW |
516 | as a regular character. In particular, the lower quantifier bound |
517 | is not optional. However, in Perl v5.18, it is planned to issue a | |
518 | deprecation warning for all such occurrences, and in Perl v5.20 to | |
519 | require literal uses of a curly bracket to be escaped, say by preceding | |
520 | them with a backslash or enclosing them within square brackets, (C<"\{"> | |
521 | or C<"[{]">). This change will allow for future syntax extensions (like | |
522 | making the lower bound of a quantifier optional), and better error | |
523 | checking of quantifiers. Now, a typo in a quantifier silently causes | |
524 | it to be treated as the literal characters. For example, | |
525 | ||
526 | /o{4,3}/ | |
527 | ||
528 | looks like a quantifier that matches 0 times, since 4 is greater than 3, | |
529 | but it really means to match the sequence of six characters | |
530 | S<C<"o { 4 , 3 }">>.) | |
531 | ||
532 | The "*" quantifier is equivalent to C<{0,}>, the "+" | |
527e91da | 533 | quantifier to C<{1,}>, and the "?" quantifier to C<{0,1}>. n and m are limited |
d0b16107 | 534 | to non-negative integral values less than a preset limit defined when perl is built. |
9c79236d GS |
535 | This is usually 32766 on the most common platforms. The actual limit can |
536 | be seen in the error message generated by code such as this: | |
537 | ||
820475bd | 538 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 539 | |
54310121 | 540 | By default, a quantified subpattern is "greedy", that is, it will match as |
541 | many times as possible (given a particular starting location) while still | |
542 | allowing the rest of the pattern to match. If you want it to match the | |
543 | minimum number of times possible, follow the quantifier with a "?". Note | |
544 | that the meanings don't change, just the "greediness": | |
0d017f4d | 545 | X<metacharacter> X<greedy> X<greediness> |
d74e8afc | 546 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{n,m}?> |
a0d0e21e | 547 | |
f793d64a KW |
548 | *? Match 0 or more times, not greedily |
549 | +? Match 1 or more times, not greedily | |
550 | ?? Match 0 or 1 time, not greedily | |
0b928c2f | 551 | {n}? Match exactly n times, not greedily (redundant) |
f793d64a KW |
552 | {n,}? Match at least n times, not greedily |
553 | {n,m}? Match at least n but not more than m times, not greedily | |
a0d0e21e | 554 | |
b9b4dddf YO |
555 | By default, when a quantified subpattern does not allow the rest of the |
556 | overall pattern to match, Perl will backtrack. However, this behaviour is | |
0d017f4d | 557 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
558 | as well. |
559 | ||
f793d64a KW |
560 | *+ Match 0 or more times and give nothing back |
561 | ++ Match 1 or more times and give nothing back | |
562 | ?+ Match 0 or 1 time and give nothing back | |
563 | {n}+ Match exactly n times and give nothing back (redundant) | |
564 | {n,}+ Match at least n times and give nothing back | |
565 | {n,m}+ Match at least n but not more than m times and give nothing back | |
b9b4dddf YO |
566 | |
567 | For instance, | |
568 | ||
569 | 'aaaa' =~ /a++a/ | |
570 | ||
571 | will never match, as the C<a++> will gobble up all the C<a>'s in the | |
572 | string and won't leave any for the remaining part of the pattern. This | |
573 | feature can be extremely useful to give perl hints about where it | |
574 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
575 | string" problem can be most efficiently performed when written as: | |
576 | ||
577 | /"(?:[^"\\]++|\\.)*+"/ | |
578 | ||
0d017f4d | 579 | as we know that if the final quote does not match, backtracking will not |
0b928c2f FC |
580 | help. See the independent subexpression |
581 | L</C<< (?>pattern) >>> for more details; | |
b9b4dddf YO |
582 | possessive quantifiers are just syntactic sugar for that construct. For |
583 | instance the above example could also be written as follows: | |
584 | ||
585 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
586 | ||
04838cea RGS |
587 | =head3 Escape sequences |
588 | ||
0b928c2f | 589 | Because patterns are processed as double-quoted strings, the following |
a0d0e21e LW |
590 | also work: |
591 | ||
f793d64a KW |
592 | \t tab (HT, TAB) |
593 | \n newline (LF, NL) | |
594 | \r return (CR) | |
595 | \f form feed (FF) | |
596 | \a alarm (bell) (BEL) | |
597 | \e escape (think troff) (ESC) | |
f793d64a | 598 | \cK control char (example: VT) |
dc0d9c48 | 599 | \x{}, \x00 character whose ordinal is the given hexadecimal number |
fb121860 | 600 | \N{name} named Unicode character or character sequence |
f793d64a | 601 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) |
f0a2b745 | 602 | \o{}, \000 character whose ordinal is the given octal number |
f793d64a KW |
603 | \l lowercase next char (think vi) |
604 | \u uppercase next char (think vi) | |
605 | \L lowercase till \E (think vi) | |
606 | \U uppercase till \E (think vi) | |
607 | \Q quote (disable) pattern metacharacters till \E | |
608 | \E end either case modification or quoted section, think vi | |
a0d0e21e | 609 | |
9bb1f947 | 610 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 611 | |
e1d1eefb | 612 | =head3 Character Classes and other Special Escapes |
04838cea | 613 | |
a0d0e21e | 614 | In addition, Perl defines the following: |
d0b16107 | 615 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 616 | |
f793d64a KW |
617 | Sequence Note Description |
618 | [...] [1] Match a character according to the rules of the | |
619 | bracketed character class defined by the "...". | |
620 | Example: [a-z] matches "a" or "b" or "c" ... or "z" | |
621 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
622 | character class "..." within the outer bracketed | |
623 | character class. Example: [[:upper:]] matches any | |
624 | uppercase character. | |
d35dd6c6 KW |
625 | \w [3] Match a "word" character (alphanumeric plus "_", plus |
626 | other connector punctuation chars plus Unicode | |
0b928c2f | 627 | marks) |
f793d64a KW |
628 | \W [3] Match a non-"word" character |
629 | \s [3] Match a whitespace character | |
630 | \S [3] Match a non-whitespace character | |
631 | \d [3] Match a decimal digit character | |
632 | \D [3] Match a non-digit character | |
633 | \pP [3] Match P, named property. Use \p{Prop} for longer names | |
634 | \PP [3] Match non-P | |
635 | \X [4] Match Unicode "eXtended grapheme cluster" | |
636 | \C Match a single C-language char (octet) even if that is | |
637 | part of a larger UTF-8 character. Thus it breaks up | |
638 | characters into their UTF-8 bytes, so you may end up | |
639 | with malformed pieces of UTF-8. Unsupported in | |
640 | lookbehind. | |
c27a5cfe | 641 | \1 [5] Backreference to a specific capture group or buffer. |
f793d64a KW |
642 | '1' may actually be any positive integer. |
643 | \g1 [5] Backreference to a specific or previous group, | |
644 | \g{-1} [5] The number may be negative indicating a relative | |
c27a5cfe | 645 | previous group and may optionally be wrapped in |
f793d64a KW |
646 | curly brackets for safer parsing. |
647 | \g{name} [5] Named backreference | |
648 | \k<name> [5] Named backreference | |
649 | \K [6] Keep the stuff left of the \K, don't include it in $& | |
650 | \N [7] Any character but \n (experimental). Not affected by | |
651 | /s modifier | |
652 | \v [3] Vertical whitespace | |
653 | \V [3] Not vertical whitespace | |
654 | \h [3] Horizontal whitespace | |
655 | \H [3] Not horizontal whitespace | |
656 | \R [4] Linebreak | |
e1d1eefb | 657 | |
9bb1f947 KW |
658 | =over 4 |
659 | ||
660 | =item [1] | |
661 | ||
662 | See L<perlrecharclass/Bracketed Character Classes> for details. | |
df225385 | 663 | |
9bb1f947 | 664 | =item [2] |
b8c5462f | 665 | |
9bb1f947 | 666 | See L<perlrecharclass/POSIX Character Classes> for details. |
b8c5462f | 667 | |
9bb1f947 | 668 | =item [3] |
5496314a | 669 | |
9bb1f947 | 670 | See L<perlrecharclass/Backslash sequences> for details. |
5496314a | 671 | |
9bb1f947 | 672 | =item [4] |
5496314a | 673 | |
9bb1f947 | 674 | See L<perlrebackslash/Misc> for details. |
d0b16107 | 675 | |
9bb1f947 | 676 | =item [5] |
b8c5462f | 677 | |
c27a5cfe | 678 | See L</Capture groups> below for details. |
93733859 | 679 | |
9bb1f947 | 680 | =item [6] |
b8c5462f | 681 | |
9bb1f947 KW |
682 | See L</Extended Patterns> below for details. |
683 | ||
684 | =item [7] | |
685 | ||
686 | Note that C<\N> has two meanings. When of the form C<\N{NAME}>, it matches the | |
fb121860 KW |
687 | character or character sequence whose name is C<NAME>; and similarly |
688 | when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode | |
689 | code point is I<hex>. Otherwise it matches any character but C<\n>. | |
9bb1f947 KW |
690 | |
691 | =back | |
d0b16107 | 692 | |
04838cea RGS |
693 | =head3 Assertions |
694 | ||
a0d0e21e | 695 | Perl defines the following zero-width assertions: |
d74e8afc ITB |
696 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
697 | X<regexp, zero-width assertion> | |
698 | X<regular expression, zero-width assertion> | |
699 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e | 700 | |
9bb1f947 KW |
701 | \b Match a word boundary |
702 | \B Match except at a word boundary | |
703 | \A Match only at beginning of string | |
704 | \Z Match only at end of string, or before newline at the end | |
705 | \z Match only at end of string | |
706 | \G Match only at pos() (e.g. at the end-of-match position | |
9da458fc | 707 | of prior m//g) |
a0d0e21e | 708 | |
14218588 | 709 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
710 | that has a C<\w> on one side of it and a C<\W> on the other side |
711 | of it (in either order), counting the imaginary characters off the | |
712 | beginning and end of the string as matching a C<\W>. (Within | |
713 | character classes C<\b> represents backspace rather than a word | |
714 | boundary, just as it normally does in any double-quoted string.) | |
715 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
716 | won't match multiple times when the C</m> modifier is used, while | |
717 | "^" and "$" will match at every internal line boundary. To match | |
718 | the actual end of the string and not ignore an optional trailing | |
719 | newline, use C<\z>. | |
d74e8afc | 720 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
721 | |
722 | The C<\G> assertion can be used to chain global matches (using | |
723 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
724 | It is also useful when writing C<lex>-like scanners, when you have | |
725 | several patterns that you want to match against consequent substrings | |
0b928c2f | 726 | of your string; see the previous reference. The actual location |
19799a22 | 727 | where C<\G> will match can also be influenced by using C<pos()> as |
58e23c8d | 728 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
0b928c2f FC |
729 | matches (see L</"Repeated Patterns Matching a Zero-length Substring">) |
730 | is modified somewhat, in that contents to the left of C<\G> are | |
58e23c8d YO |
731 | not counted when determining the length of the match. Thus the following |
732 | will not match forever: | |
d74e8afc | 733 | X<\G> |
c47ff5f1 | 734 | |
e761bb84 CO |
735 | my $string = 'ABC'; |
736 | pos($string) = 1; | |
737 | while ($string =~ /(.\G)/g) { | |
738 | print $1; | |
739 | } | |
58e23c8d YO |
740 | |
741 | It will print 'A' and then terminate, as it considers the match to | |
742 | be zero-width, and thus will not match at the same position twice in a | |
743 | row. | |
744 | ||
745 | It is worth noting that C<\G> improperly used can result in an infinite | |
746 | loop. Take care when using patterns that include C<\G> in an alternation. | |
747 | ||
c27a5cfe | 748 | =head3 Capture groups |
04838cea | 749 | |
c27a5cfe KW |
750 | The bracketing construct C<( ... )> creates capture groups (also referred to as |
751 | capture buffers). To refer to the current contents of a group later on, within | |
d8b950dc KW |
752 | the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>) |
753 | for the second, and so on. | |
754 | This is called a I<backreference>. | |
d74e8afc | 755 | X<regex, capture buffer> X<regexp, capture buffer> |
c27a5cfe | 756 | X<regex, capture group> X<regexp, capture group> |
d74e8afc | 757 | X<regular expression, capture buffer> X<backreference> |
c27a5cfe | 758 | X<regular expression, capture group> X<backreference> |
1f1031fe | 759 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
d8b950dc KW |
760 | X<named capture buffer> X<regular expression, named capture buffer> |
761 | X<named capture group> X<regular expression, named capture group> | |
762 | X<%+> X<$+{name}> X<< \k<name> >> | |
763 | There is no limit to the number of captured substrings that you may use. | |
764 | Groups are numbered with the leftmost open parenthesis being number 1, etc. If | |
765 | a group did not match, the associated backreference won't match either. (This | |
766 | can happen if the group is optional, or in a different branch of an | |
767 | alternation.) | |
768 | You can omit the C<"g">, and write C<"\1">, etc, but there are some issues with | |
769 | this form, described below. | |
770 | ||
771 | You can also refer to capture groups relatively, by using a negative number, so | |
772 | that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture | |
773 | group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For | |
774 | example: | |
5624f11d YO |
775 | |
776 | / | |
c27a5cfe KW |
777 | (Y) # group 1 |
778 | ( # group 2 | |
779 | (X) # group 3 | |
780 | \g{-1} # backref to group 3 | |
781 | \g{-3} # backref to group 1 | |
5624f11d YO |
782 | ) |
783 | /x | |
784 | ||
d8b950dc KW |
785 | would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to |
786 | interpolate regexes into larger regexes and not have to worry about the | |
787 | capture groups being renumbered. | |
788 | ||
789 | You can dispense with numbers altogether and create named capture groups. | |
790 | The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to | |
791 | reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may | |
792 | also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.) | |
793 | I<name> must not begin with a number, nor contain hyphens. | |
794 | When different groups within the same pattern have the same name, any reference | |
795 | to that name assumes the leftmost defined group. Named groups count in | |
796 | absolute and relative numbering, and so can also be referred to by those | |
797 | numbers. | |
798 | (It's possible to do things with named capture groups that would otherwise | |
799 | require C<(??{})>.) | |
800 | ||
801 | Capture group contents are dynamically scoped and available to you outside the | |
802 | pattern until the end of the enclosing block or until the next successful | |
803 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
804 | You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">, | |
805 | etc); or by name via the C<%+> hash, using C<"$+{I<name>}">. | |
806 | ||
807 | Braces are required in referring to named capture groups, but are optional for | |
808 | absolute or relative numbered ones. Braces are safer when creating a regex by | |
809 | concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a> | |
810 | contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which | |
811 | is probably not what you intended. | |
812 | ||
813 | The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that | |
814 | there were no named nor relative numbered capture groups. Absolute numbered | |
0b928c2f FC |
815 | groups were referred to using C<\1>, |
816 | C<\2>, etc., and this notation is still | |
d8b950dc KW |
817 | accepted (and likely always will be). But it leads to some ambiguities if |
818 | there are more than 9 capture groups, as C<\10> could mean either the tenth | |
819 | capture group, or the character whose ordinal in octal is 010 (a backspace in | |
820 | ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference | |
821 | only if at least 10 left parentheses have opened before it. Likewise C<\11> is | |
822 | a backreference only if at least 11 left parentheses have opened before it. | |
e1f120a9 KW |
823 | And so on. C<\1> through C<\9> are always interpreted as backreferences. |
824 | There are several examples below that illustrate these perils. You can avoid | |
825 | the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups; | |
826 | and for octal constants always using C<\o{}>, or for C<\077> and below, using 3 | |
827 | digits padded with leading zeros, since a leading zero implies an octal | |
828 | constant. | |
d8b950dc KW |
829 | |
830 | The C<\I<digit>> notation also works in certain circumstances outside | |
ed7efc79 | 831 | the pattern. See L</Warning on \1 Instead of $1> below for details. |
81714fb9 | 832 | |
14218588 | 833 | Examples: |
a0d0e21e LW |
834 | |
835 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
836 | ||
d8b950dc | 837 | /(.)\g1/ # find first doubled char |
81714fb9 YO |
838 | and print "'$1' is the first doubled character\n"; |
839 | ||
840 | /(?<char>.)\k<char>/ # ... a different way | |
841 | and print "'$+{char}' is the first doubled character\n"; | |
842 | ||
d8b950dc | 843 | /(?'char'.)\g1/ # ... mix and match |
81714fb9 | 844 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 845 | |
14218588 | 846 | if (/Time: (..):(..):(..)/) { # parse out values |
f793d64a KW |
847 | $hours = $1; |
848 | $minutes = $2; | |
849 | $seconds = $3; | |
a0d0e21e | 850 | } |
c47ff5f1 | 851 | |
9d860678 KW |
852 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference |
853 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal | |
854 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference | |
855 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal | |
856 | ||
857 | $a = '(.)\1'; # Creates problems when concatenated. | |
858 | $b = '(.)\g{1}'; # Avoids the problems. | |
859 | "aa" =~ /${a}/; # True | |
860 | "aa" =~ /${b}/; # True | |
861 | "aa0" =~ /${a}0/; # False! | |
862 | "aa0" =~ /${b}0/; # True | |
dc0d9c48 KW |
863 | "aa\x08" =~ /${a}0/; # True! |
864 | "aa\x08" =~ /${b}0/; # False | |
9d860678 | 865 | |
14218588 GS |
866 | Several special variables also refer back to portions of the previous |
867 | match. C<$+> returns whatever the last bracket match matched. | |
868 | C<$&> returns the entire matched string. (At one point C<$0> did | |
869 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
870 | everything before the matched string. C<$'> returns everything |
871 | after the matched string. And C<$^N> contains whatever was matched by | |
872 | the most-recently closed group (submatch). C<$^N> can be used in | |
873 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 874 | variable. |
d74e8afc | 875 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 876 | |
d8b950dc KW |
877 | These special variables, like the C<%+> hash and the numbered match variables |
878 | (C<$1>, C<$2>, C<$3>, etc.) are dynamically scoped | |
14218588 GS |
879 | until the end of the enclosing block or until the next successful |
880 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
881 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
882 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
883 | ||
0d017f4d | 884 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 885 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
886 | specific cases and remembers the best match. |
887 | ||
14218588 GS |
888 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or |
889 | C<$'> anywhere in the program, it has to provide them for every | |
890 | pattern match. This may substantially slow your program. Perl | |
d8b950dc | 891 | uses the same mechanism to produce C<$1>, C<$2>, etc, so you also pay a |
14218588 GS |
892 | price for each pattern that contains capturing parentheses. (To |
893 | avoid this cost while retaining the grouping behaviour, use the | |
894 | extended regular expression C<(?: ... )> instead.) But if you never | |
895 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
896 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
897 | if you can, but if you can't (and some algorithms really appreciate | |
898 | them), once you've used them once, use them at will, because you've | |
c149d39e DM |
899 | already paid the price. As of 5.17.4, the presence of each of the three |
900 | variables in a program is recorded separately, and depending on | |
901 | circumstances, perl may be able be more efficient knowing that only C<$&> | |
902 | rather than all three have been seen, for example. | |
d74e8afc | 903 | X<$&> X<$`> X<$'> |
68dc0745 | 904 | |
99d59c4d | 905 | As a workaround for this problem, Perl 5.10.0 introduces C<${^PREMATCH}>, |
cde0cee5 YO |
906 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
907 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 908 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 YO |
909 | The use of these variables incurs no global performance penalty, unlike |
910 | their punctuation char equivalents, however at the trade-off that you | |
911 | have to tell perl when you want to use them. | |
87e95b7f | 912 | X</p> X<p modifier> |
cde0cee5 | 913 | |
9d727203 KW |
914 | =head2 Quoting metacharacters |
915 | ||
19799a22 GS |
916 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
917 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
918 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 919 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
920 | interpreted as a literal character, not a metacharacter. This was |
921 | once used in a common idiom to disable or quote the special meanings | |
922 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 923 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
924 | |
925 | $pattern =~ s/(\W)/\\$1/g; | |
926 | ||
f1cbbd6e | 927 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
928 | Today it is more common to use the quotemeta() function or the C<\Q> |
929 | metaquoting escape sequence to disable all metacharacters' special | |
930 | meanings like this: | |
a0d0e21e LW |
931 | |
932 | /$unquoted\Q$quoted\E$unquoted/ | |
933 | ||
9da458fc IZ |
934 | Beware that if you put literal backslashes (those not inside |
935 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
936 | backslash interpolation may lead to confusing results. If you | |
937 | I<need> to use literal backslashes within C<\Q...\E>, | |
938 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
939 | ||
736fe711 KW |
940 | C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>. |
941 | ||
19799a22 GS |
942 | =head2 Extended Patterns |
943 | ||
14218588 | 944 | Perl also defines a consistent extension syntax for features not |
0b928c2f FC |
945 | found in standard tools like B<awk> and |
946 | B<lex>. The syntax for most of these is a | |
14218588 GS |
947 | pair of parentheses with a question mark as the first thing within |
948 | the parentheses. The character after the question mark indicates | |
949 | the extension. | |
19799a22 | 950 | |
14218588 GS |
951 | The stability of these extensions varies widely. Some have been |
952 | part of the core language for many years. Others are experimental | |
953 | and may change without warning or be completely removed. Check | |
954 | the documentation on an individual feature to verify its current | |
955 | status. | |
19799a22 | 956 | |
14218588 GS |
957 | A question mark was chosen for this and for the minimal-matching |
958 | construct because 1) question marks are rare in older regular | |
959 | expressions, and 2) whenever you see one, you should stop and | |
0b928c2f | 960 | "question" exactly what is going on. That's psychology.... |
a0d0e21e | 961 | |
70ca8714 | 962 | =over 4 |
a0d0e21e | 963 | |
cc6b7395 | 964 | =item C<(?#text)> |
d74e8afc | 965 | X<(?#)> |
a0d0e21e | 966 | |
14218588 | 967 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 968 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
969 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
970 | C<)> in the comment. | |
a0d0e21e | 971 | |
cfaf538b | 972 | =item C<(?adlupimsx-imsx)> |
fb85c044 | 973 | |
cfaf538b | 974 | =item C<(?^alupimsx)> |
fb85c044 | 975 | X<(?)> X<(?^)> |
19799a22 | 976 | |
0b6d1084 JH |
977 | One or more embedded pattern-match modifiers, to be turned on (or |
978 | turned off, if preceded by C<->) for the remainder of the pattern or | |
fb85c044 KW |
979 | the remainder of the enclosing pattern group (if any). |
980 | ||
fb85c044 | 981 | This is particularly useful for dynamic patterns, such as those read in from a |
0d017f4d | 982 | configuration file, taken from an argument, or specified in a table |
0b928c2f FC |
983 | somewhere. Consider the case where some patterns want to be |
984 | case-sensitive and some do not: The case-insensitive ones merely need to | |
0d017f4d | 985 | include C<(?i)> at the front of the pattern. For example: |
19799a22 GS |
986 | |
987 | $pattern = "foobar"; | |
5d458dd8 | 988 | if ( /$pattern/i ) { } |
19799a22 GS |
989 | |
990 | # more flexible: | |
991 | ||
992 | $pattern = "(?i)foobar"; | |
5d458dd8 | 993 | if ( /$pattern/ ) { } |
19799a22 | 994 | |
0b6d1084 | 995 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 | 996 | |
d8b950dc | 997 | ( (?i) blah ) \s+ \g1 |
19799a22 | 998 | |
0d017f4d WL |
999 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
1000 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
1001 | modifier outside this group. | |
19799a22 | 1002 | |
8eb5594e | 1003 | These modifiers do not carry over into named subpatterns called in the |
dd72e27b | 1004 | enclosing group. In other words, a pattern such as C<((?i)(?&NAME))> does not |
8eb5594e DR |
1005 | change the case-sensitivity of the "NAME" pattern. |
1006 | ||
dc925305 KW |
1007 | Any of these modifiers can be set to apply globally to all regular |
1008 | expressions compiled within the scope of a C<use re>. See | |
a0bbd6ff | 1009 | L<re/"'/flags' mode">. |
dc925305 | 1010 | |
9de15fec KW |
1011 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
1012 | after the C<"?"> is a shorthand equivalent to C<d-imsx>. Flags (except | |
1013 | C<"d">) may follow the caret to override it. | |
1014 | But a minus sign is not legal with it. | |
1015 | ||
dc925305 | 1016 | Note that the C<a>, C<d>, C<l>, C<p>, and C<u> modifiers are special in |
e1d8d8ac | 1017 | that they can only be enabled, not disabled, and the C<a>, C<d>, C<l>, and |
dc925305 | 1018 | C<u> modifiers are mutually exclusive: specifying one de-specifies the |
ed7efc79 KW |
1019 | others, and a maximum of one (or two C<a>'s) may appear in the |
1020 | construct. Thus, for | |
0b928c2f | 1021 | example, C<(?-p)> will warn when compiled under C<use warnings>; |
b6fa137b | 1022 | C<(?-d:...)> and C<(?dl:...)> are fatal errors. |
9de15fec KW |
1023 | |
1024 | Note also that the C<p> modifier is special in that its presence | |
1025 | anywhere in a pattern has a global effect. | |
cde0cee5 | 1026 | |
5a964f20 | 1027 | =item C<(?:pattern)> |
d74e8afc | 1028 | X<(?:)> |
a0d0e21e | 1029 | |
cfaf538b | 1030 | =item C<(?adluimsx-imsx:pattern)> |
ca9dfc88 | 1031 | |
cfaf538b | 1032 | =item C<(?^aluimsx:pattern)> |
fb85c044 KW |
1033 | X<(?^:)> |
1034 | ||
5a964f20 TC |
1035 | This is for clustering, not capturing; it groups subexpressions like |
1036 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 1037 | |
5a964f20 | 1038 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
1039 | |
1040 | is like | |
1041 | ||
5a964f20 | 1042 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 1043 | |
19799a22 GS |
1044 | but doesn't spit out extra fields. It's also cheaper not to capture |
1045 | characters if you don't need to. | |
a0d0e21e | 1046 | |
19799a22 | 1047 | Any letters between C<?> and C<:> act as flags modifiers as with |
cfaf538b | 1048 | C<(?adluimsx-imsx)>. For example, |
ca9dfc88 IZ |
1049 | |
1050 | /(?s-i:more.*than).*million/i | |
1051 | ||
14218588 | 1052 | is equivalent to the more verbose |
ca9dfc88 IZ |
1053 | |
1054 | /(?:(?s-i)more.*than).*million/i | |
1055 | ||
fb85c044 | 1056 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
9de15fec KW |
1057 | after the C<"?"> is a shorthand equivalent to C<d-imsx>. Any positive |
1058 | flags (except C<"d">) may follow the caret, so | |
fb85c044 KW |
1059 | |
1060 | (?^x:foo) | |
1061 | ||
1062 | is equivalent to | |
1063 | ||
1064 | (?x-ims:foo) | |
1065 | ||
1066 | The caret tells Perl that this cluster doesn't inherit the flags of any | |
0b928c2f | 1067 | surrounding pattern, but uses the system defaults (C<d-imsx>), |
fb85c044 KW |
1068 | modified by any flags specified. |
1069 | ||
1070 | The caret allows for simpler stringification of compiled regular | |
1071 | expressions. These look like | |
1072 | ||
1073 | (?^:pattern) | |
1074 | ||
1075 | with any non-default flags appearing between the caret and the colon. | |
1076 | A test that looks at such stringification thus doesn't need to have the | |
1077 | system default flags hard-coded in it, just the caret. If new flags are | |
1078 | added to Perl, the meaning of the caret's expansion will change to include | |
1079 | the default for those flags, so the test will still work, unchanged. | |
1080 | ||
1081 | Specifying a negative flag after the caret is an error, as the flag is | |
1082 | redundant. | |
1083 | ||
1084 | Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is | |
1085 | to match at the beginning. | |
1086 | ||
594d7033 YO |
1087 | =item C<(?|pattern)> |
1088 | X<(?|)> X<Branch reset> | |
1089 | ||
1090 | This is the "branch reset" pattern, which has the special property | |
c27a5cfe | 1091 | that the capture groups are numbered from the same starting point |
99d59c4d | 1092 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 1093 | |
c27a5cfe | 1094 | Capture groups are numbered from left to right, but inside this |
693596a8 | 1095 | construct the numbering is restarted for each branch. |
4deaaa80 | 1096 | |
c27a5cfe | 1097 | The numbering within each branch will be as normal, and any groups |
4deaaa80 PJ |
1098 | following this construct will be numbered as though the construct |
1099 | contained only one branch, that being the one with the most capture | |
c27a5cfe | 1100 | groups in it. |
4deaaa80 | 1101 | |
0b928c2f | 1102 | This construct is useful when you want to capture one of a |
4deaaa80 PJ |
1103 | number of alternative matches. |
1104 | ||
1105 | Consider the following pattern. The numbers underneath show in | |
c27a5cfe | 1106 | which group the captured content will be stored. |
594d7033 YO |
1107 | |
1108 | ||
1109 | # before ---------------branch-reset----------- after | |
1110 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x | |
1111 | # 1 2 2 3 2 3 4 | |
1112 | ||
ab106183 A |
1113 | Be careful when using the branch reset pattern in combination with |
1114 | named captures. Named captures are implemented as being aliases to | |
c27a5cfe | 1115 | numbered groups holding the captures, and that interferes with the |
ab106183 A |
1116 | implementation of the branch reset pattern. If you are using named |
1117 | captures in a branch reset pattern, it's best to use the same names, | |
1118 | in the same order, in each of the alternations: | |
1119 | ||
1120 | /(?| (?<a> x ) (?<b> y ) | |
1121 | | (?<a> z ) (?<b> w )) /x | |
1122 | ||
1123 | Not doing so may lead to surprises: | |
1124 | ||
1125 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
1126 | say $+ {a}; # Prints '12' | |
1127 | say $+ {b}; # *Also* prints '12'. | |
1128 | ||
c27a5cfe KW |
1129 | The problem here is that both the group named C<< a >> and the group |
1130 | named C<< b >> are aliases for the group belonging to C<< $1 >>. | |
90a18110 | 1131 | |
ee9b8eae YO |
1132 | =item Look-Around Assertions |
1133 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> | |
1134 | ||
0b928c2f | 1135 | Look-around assertions are zero-width patterns which match a specific |
ee9b8eae YO |
1136 | pattern without including it in C<$&>. Positive assertions match when |
1137 | their subpattern matches, negative assertions match when their subpattern | |
1138 | fails. Look-behind matches text up to the current match position, | |
1139 | look-ahead matches text following the current match position. | |
1140 | ||
1141 | =over 4 | |
1142 | ||
5a964f20 | 1143 | =item C<(?=pattern)> |
d74e8afc | 1144 | X<(?=)> X<look-ahead, positive> X<lookahead, positive> |
a0d0e21e | 1145 | |
19799a22 | 1146 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
1147 | matches a word followed by a tab, without including the tab in C<$&>. |
1148 | ||
5a964f20 | 1149 | =item C<(?!pattern)> |
d74e8afc | 1150 | X<(?!)> X<look-ahead, negative> X<lookahead, negative> |
a0d0e21e | 1151 | |
19799a22 | 1152 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 1153 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
1154 | however that look-ahead and look-behind are NOT the same thing. You cannot |
1155 | use this for look-behind. | |
7b8d334a | 1156 | |
5a964f20 | 1157 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
1158 | will not do what you want. That's because the C<(?!foo)> is just saying that |
1159 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
0b928c2f | 1160 | match. Use look-behind instead (see below). |
c277df42 | 1161 | |
ee9b8eae YO |
1162 | =item C<(?<=pattern)> C<\K> |
1163 | X<(?<=)> X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 1164 | |
c47ff5f1 | 1165 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
1166 | matches a word that follows a tab, without including the tab in C<$&>. |
1167 | Works only for fixed-width look-behind. | |
c277df42 | 1168 | |
ee9b8eae YO |
1169 | There is a special form of this construct, called C<\K>, which causes the |
1170 | regex engine to "keep" everything it had matched prior to the C<\K> and | |
0b928c2f | 1171 | not include it in C<$&>. This effectively provides variable-length |
ee9b8eae YO |
1172 | look-behind. The use of C<\K> inside of another look-around assertion |
1173 | is allowed, but the behaviour is currently not well defined. | |
1174 | ||
c62285ac | 1175 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
1176 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
1177 | situations where you want to efficiently remove something following | |
1178 | something else in a string. For instance | |
1179 | ||
1180 | s/(foo)bar/$1/g; | |
1181 | ||
1182 | can be rewritten as the much more efficient | |
1183 | ||
1184 | s/foo\Kbar//g; | |
1185 | ||
5a964f20 | 1186 | =item C<(?<!pattern)> |
d74e8afc | 1187 | X<(?<!)> X<look-behind, negative> X<lookbehind, negative> |
c277df42 | 1188 | |
19799a22 GS |
1189 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
1190 | matches any occurrence of "foo" that does not follow "bar". Works | |
1191 | only for fixed-width look-behind. | |
c277df42 | 1192 | |
ee9b8eae YO |
1193 | =back |
1194 | ||
81714fb9 YO |
1195 | =item C<(?'NAME'pattern)> |
1196 | ||
1197 | =item C<< (?<NAME>pattern) >> | |
1198 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> | |
1199 | ||
c27a5cfe | 1200 | A named capture group. Identical in every respect to normal capturing |
0b928c2f FC |
1201 | parentheses C<()> but for the additional fact that the group |
1202 | can be referred to by name in various regular expression | |
1203 | constructs (like C<\g{NAME}>) and can be accessed by name | |
1204 | after a successful match via C<%+> or C<%->. See L<perlvar> | |
90a18110 | 1205 | for more details on the C<%+> and C<%-> hashes. |
81714fb9 | 1206 | |
c27a5cfe KW |
1207 | If multiple distinct capture groups have the same name then the |
1208 | $+{NAME} will refer to the leftmost defined group in the match. | |
81714fb9 | 1209 | |
0d017f4d | 1210 | The forms C<(?'NAME'pattern)> and C<< (?<NAME>pattern) >> are equivalent. |
81714fb9 YO |
1211 | |
1212 | B<NOTE:> While the notation of this construct is the same as the similar | |
c27a5cfe | 1213 | function in .NET regexes, the behavior is not. In Perl the groups are |
81714fb9 YO |
1214 | numbered sequentially regardless of being named or not. Thus in the |
1215 | pattern | |
1216 | ||
1217 | /(x)(?<foo>y)(z)/ | |
1218 | ||
1219 | $+{foo} will be the same as $2, and $3 will contain 'z' instead of | |
1220 | the opposite which is what a .NET regex hacker might expect. | |
1221 | ||
1f1031fe YO |
1222 | Currently NAME is restricted to simple identifiers only. |
1223 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or | |
1224 | its Unicode extension (see L<utf8>), | |
1225 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 1226 | |
1f1031fe | 1227 | B<NOTE:> In order to make things easier for programmers with experience |
ae5648b3 | 1228 | with the Python or PCRE regex engines, the pattern C<< (?PE<lt>NAMEE<gt>pattern) >> |
0d017f4d | 1229 | may be used instead of C<< (?<NAME>pattern) >>; however this form does not |
64c5a566 | 1230 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 1231 | |
1f1031fe YO |
1232 | =item C<< \k<NAME> >> |
1233 | ||
1234 | =item C<< \k'NAME' >> | |
81714fb9 YO |
1235 | |
1236 | Named backreference. Similar to numeric backreferences, except that | |
1237 | the group is designated by name and not number. If multiple groups | |
1238 | have the same name then it refers to the leftmost defined group in | |
1239 | the current match. | |
1240 | ||
0d017f4d | 1241 | It is an error to refer to a name not defined by a C<< (?<NAME>) >> |
81714fb9 YO |
1242 | earlier in the pattern. |
1243 | ||
1244 | Both forms are equivalent. | |
1245 | ||
1f1031fe | 1246 | B<NOTE:> In order to make things easier for programmers with experience |
0d017f4d | 1247 | with the Python or PCRE regex engines, the pattern C<< (?P=NAME) >> |
64c5a566 | 1248 | may be used instead of C<< \k<NAME> >>. |
1f1031fe | 1249 | |
cc6b7395 | 1250 | =item C<(?{ code })> |
d74e8afc | 1251 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 1252 | |
19799a22 | 1253 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
1254 | experimental, and may be changed without notice. Code executed that |
1255 | has side effects may not perform identically from version to version | |
e128ab2c DM |
1256 | due to the effect of future optimisations in the regex engine. The |
1257 | implementation of this feature was radically overhauled for the 5.18.0 | |
1258 | release, and its behaviour in earlier versions of perl was much buggier, | |
1259 | especially in relation to parsing, lexical vars, scoping, recursion and | |
1260 | reentrancy. | |
c277df42 | 1261 | |
e128ab2c DM |
1262 | This zero-width assertion executes any embedded Perl code. It always |
1263 | succeeds, and its return value is set as C<$^R>. | |
19799a22 | 1264 | |
e128ab2c DM |
1265 | In literal patterns, the code is parsed at the same time as the |
1266 | surrounding code. While within the pattern, control is passed temporarily | |
1267 | back to the perl parser, until the logically-balancing closing brace is | |
1268 | encountered. This is similar to the way that an array index expression in | |
1269 | a literal string is handled, for example | |
77ea4f6d | 1270 | |
e128ab2c DM |
1271 | "abc$array[ 1 + f('[') + g()]def" |
1272 | ||
1273 | In particular, braces do not need to be balanced: | |
1274 | ||
1275 | /abc(?{ f('{'); })/def/ | |
1276 | ||
1277 | Even in a pattern that is interpolated and compiled at run-time, literal | |
1278 | code blocks will be compiled once, at perl compile time; the following | |
1279 | prints "ABCD": | |
1280 | ||
1281 | print "D"; | |
1282 | my $qr = qr/(?{ BEGIN { print "A" } })/; | |
1283 | my $foo = "foo"; | |
1284 | /$foo$qr(?{ BEGIN { print "B" } })/; | |
1285 | BEGIN { print "C" } | |
1286 | ||
1287 | In patterns where the text of the code is derived from run-time | |
1288 | information rather than appearing literally in a source code /pattern/, | |
1289 | the code is compiled at the same time that the pattern is compiled, and | |
1290 | fro reasons of security, C<use re 'eval'> must be in scope. This is to | |
1291 | stop user-supplied patterns containing code snippets from being | |
1292 | executable. | |
1293 | ||
1294 | In situations where you need enable this with C<use re 'eval'>, you should | |
1295 | also have taint checking enabled. Better yet, use the carefully | |
1296 | constrained evaluation within a Safe compartment. See L<perlsec> for | |
1297 | details about both these mechanisms. | |
1298 | ||
1299 | From the viewpoint of parsing, lexical variable scope and closures, | |
1300 | ||
1301 | /AAA(?{ BBB })CCC/ | |
1302 | ||
1303 | behaves approximately like | |
1304 | ||
1305 | /AAA/ && do { BBB } && /CCC/ | |
1306 | ||
1307 | Similarly, | |
1308 | ||
1309 | qr/AAA(?{ BBB })CCC/ | |
1310 | ||
1311 | behaves approximately like | |
77ea4f6d | 1312 | |
e128ab2c DM |
1313 | sub { /AAA/ && do { BBB } && /CCC/ } |
1314 | ||
1315 | In particular: | |
1316 | ||
1317 | { my $i = 1; $r = qr/(?{ print $i })/ } | |
1318 | my $i = 2; | |
1319 | /$r/; # prints "1" | |
1320 | ||
1321 | Inside a C<(?{...})> block, C<$_> refers to the string the regular | |
754091cb | 1322 | expression is matching against. You can also use C<pos()> to know what is |
fa11829f | 1323 | the current position of matching within this string. |
754091cb | 1324 | |
e128ab2c DM |
1325 | The code block introduces a new scope from the perspective of lexical |
1326 | variable declarations, but B<not> from the perspective of C<local> and | |
1327 | similar localizing behaviours. So later code blocks within the same | |
1328 | pattern will still see the values which were localized in earlier blocks. | |
1329 | These accumulated localizations are undone either at the end of a | |
1330 | successful match, or if the assertion is backtracked (compare | |
1331 | L<"Backtracking">). For example, | |
b9ac3b5b GS |
1332 | |
1333 | $_ = 'a' x 8; | |
5d458dd8 | 1334 | m< |
d1fbf752 | 1335 | (?{ $cnt = 0 }) # Initialize $cnt. |
b9ac3b5b | 1336 | ( |
5d458dd8 | 1337 | a |
b9ac3b5b | 1338 | (?{ |
d1fbf752 KW |
1339 | local $cnt = $cnt + 1; # Update $cnt, |
1340 | # backtracking-safe. | |
b9ac3b5b | 1341 | }) |
5d458dd8 | 1342 | )* |
b9ac3b5b | 1343 | aaaa |
d1fbf752 KW |
1344 | (?{ $res = $cnt }) # On success copy to |
1345 | # non-localized location. | |
b9ac3b5b GS |
1346 | >x; |
1347 | ||
e128ab2c DM |
1348 | will initially increment C<$cnt> up to 8; then during backtracking, its |
1349 | value will be unwound back to 4, which is the value assigned to C<$res>. | |
1350 | At the end of the regex execution, $cnt will be wound back to its initial | |
1351 | value of 0. | |
1352 | ||
1353 | This assertion may be used as the condition in a | |
b9ac3b5b | 1354 | |
e128ab2c DM |
1355 | (?(condition)yes-pattern|no-pattern) |
1356 | ||
1357 | switch. If I<not> used in this way, the result of evaluation of C<code> | |
1358 | is put into the special variable C<$^R>. This happens immediately, so | |
1359 | C<$^R> can be used from other C<(?{ code })> assertions inside the same | |
1360 | regular expression. | |
b9ac3b5b | 1361 | |
19799a22 GS |
1362 | The assignment to C<$^R> above is properly localized, so the old |
1363 | value of C<$^R> is restored if the assertion is backtracked; compare | |
1364 | L<"Backtracking">. | |
b9ac3b5b | 1365 | |
e128ab2c DM |
1366 | Note that the special variable C<$^N> is particularly useful with code |
1367 | blocks to capture the results of submatches in variables without having to | |
1368 | keep track of the number of nested parentheses. For example: | |
1369 | ||
1370 | $_ = "The brown fox jumps over the lazy dog"; | |
1371 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
1372 | print "color = $color, animal = $animal\n"; | |
1373 | ||
8988a1bb | 1374 | |
14455d6c | 1375 | =item C<(??{ code })> |
d74e8afc ITB |
1376 | X<(??{})> |
1377 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 1378 | |
19799a22 | 1379 | B<WARNING>: This extended regular expression feature is considered |
b9b4dddf YO |
1380 | experimental, and may be changed without notice. Code executed that |
1381 | has side effects may not perform identically from version to version | |
1382 | due to the effect of future optimisations in the regex engine. | |
0f5d15d6 | 1383 | |
e128ab2c DM |
1384 | This is a "postponed" regular subexpression. It behaves in I<exactly> the |
1385 | same way as a C<(?{ code })> code block as described above, except that | |
1386 | its return value, rather than being assigned to C<$^R>, is treated as a | |
1387 | pattern, compiled if it's a string (or used as-is if its a qr// object), | |
1388 | then matched as if it were inserted instead of this construct. | |
6bda09f9 | 1389 | |
e128ab2c DM |
1390 | During the matching of this sub-pattern, it has its own set of |
1391 | captures which are valid during the sub-match, but are discarded once | |
1392 | control returns to the main pattern. For example, the following matches, | |
1393 | with the inner pattern capturing "B" and matching "BB", while the outer | |
1394 | pattern captures "A"; | |
1395 | ||
1396 | my $inner = '(.)\1'; | |
1397 | "ABBA" =~ /^(.)(??{ $inner })\1/; | |
1398 | print $1; # prints "A"; | |
6bda09f9 | 1399 | |
e128ab2c DM |
1400 | Note that this means that there is no way for the inner pattern to refer |
1401 | to a capture group defined outside. (The code block itself can use C<$1>, | |
1402 | etc., to refer to the enclosing pattern's capture groups.) Thus, although | |
0f5d15d6 | 1403 | |
e128ab2c DM |
1404 | ('a' x 100)=~/(??{'(.)' x 100})/ |
1405 | ||
1406 | I<will> match, it will I<not> set $1 on exit. | |
19799a22 GS |
1407 | |
1408 | The following pattern matches a parenthesized group: | |
0f5d15d6 | 1409 | |
d1fbf752 KW |
1410 | $re = qr{ |
1411 | \( | |
1412 | (?: | |
1413 | (?> [^()]+ ) # Non-parens without backtracking | |
1414 | | | |
1415 | (??{ $re }) # Group with matching parens | |
1416 | )* | |
1417 | \) | |
1418 | }x; | |
0f5d15d6 | 1419 | |
6bda09f9 YO |
1420 | See also C<(?PARNO)> for a different, more efficient way to accomplish |
1421 | the same task. | |
1422 | ||
e128ab2c DM |
1423 | Executing a postponed regular expression 50 times without consuming any |
1424 | input string will result in a fatal error. The maximum depth is compiled | |
1425 | into perl, so changing it requires a custom build. | |
6bda09f9 | 1426 | |
542fa716 YO |
1427 | =item C<(?PARNO)> C<(?-PARNO)> C<(?+PARNO)> C<(?R)> C<(?0)> |
1428 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> | |
6bda09f9 | 1429 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
542fa716 | 1430 | X<regex, relative recursion> |
6bda09f9 | 1431 | |
e128ab2c DM |
1432 | Similar to C<(??{ code })> except that it does not involve executing any |
1433 | code or potentially compiling a returned pattern string; instead it treats | |
1434 | the part of the current pattern contained within a specified capture group | |
1435 | as an independent pattern that must match at the current position. | |
1436 | Capture groups contained by the pattern will have the value as determined | |
1437 | by the outermost recursion. | |
6bda09f9 | 1438 | |
894be9b7 | 1439 | PARNO is a sequence of digits (not starting with 0) whose value reflects |
c27a5cfe | 1440 | the paren-number of the capture group to recurse to. C<(?R)> recurses to |
894be9b7 | 1441 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
542fa716 | 1442 | C<(?R)>. If PARNO is preceded by a plus or minus sign then it is assumed |
c27a5cfe | 1443 | to be relative, with negative numbers indicating preceding capture groups |
542fa716 | 1444 | and positive ones following. Thus C<(?-1)> refers to the most recently |
c27a5cfe | 1445 | declared group, and C<(?+1)> indicates the next group to be declared. |
c74340f9 | 1446 | Note that the counting for relative recursion differs from that of |
c27a5cfe | 1447 | relative backreferences, in that with recursion unclosed groups B<are> |
c74340f9 | 1448 | included. |
6bda09f9 | 1449 | |
81714fb9 | 1450 | The following pattern matches a function foo() which may contain |
f145b7e9 | 1451 | balanced parentheses as the argument. |
6bda09f9 | 1452 | |
d1fbf752 | 1453 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 | 1454 | foo |
d1fbf752 | 1455 | ( # paren group 2 (parens) |
6bda09f9 | 1456 | \( |
d1fbf752 | 1457 | ( # paren group 3 (contents of parens) |
6bda09f9 | 1458 | (?: |
d1fbf752 | 1459 | (?> [^()]+ ) # Non-parens without backtracking |
6bda09f9 | 1460 | | |
d1fbf752 | 1461 | (?2) # Recurse to start of paren group 2 |
6bda09f9 YO |
1462 | )* |
1463 | ) | |
1464 | \) | |
1465 | ) | |
1466 | ) | |
1467 | }x; | |
1468 | ||
1469 | If the pattern was used as follows | |
1470 | ||
1471 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
1472 | and print "\$1 = $1\n", | |
1473 | "\$2 = $2\n", | |
1474 | "\$3 = $3\n"; | |
1475 | ||
1476 | the output produced should be the following: | |
1477 | ||
1478 | $1 = foo(bar(baz)+baz(bop)) | |
1479 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 1480 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 1481 | |
c27a5cfe | 1482 | If there is no corresponding capture group defined, then it is a |
61528107 | 1483 | fatal error. Recursing deeper than 50 times without consuming any input |
81714fb9 | 1484 | string will also result in a fatal error. The maximum depth is compiled |
6bda09f9 YO |
1485 | into perl, so changing it requires a custom build. |
1486 | ||
542fa716 YO |
1487 | The following shows how using negative indexing can make it |
1488 | easier to embed recursive patterns inside of a C<qr//> construct | |
1489 | for later use: | |
1490 | ||
1491 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
c77257ed | 1492 | if (/foo $parens \s+ \+ \s+ bar $parens/x) { |
542fa716 YO |
1493 | # do something here... |
1494 | } | |
1495 | ||
81714fb9 | 1496 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 1497 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 1498 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 YO |
1499 | as atomic. Also, modifiers are resolved at compile time, so constructs |
1500 | like (?i:(?1)) or (?:(?i)(?1)) do not affect how the sub-pattern will | |
1501 | be processed. | |
6bda09f9 | 1502 | |
894be9b7 YO |
1503 | =item C<(?&NAME)> |
1504 | X<(?&NAME)> | |
1505 | ||
0d017f4d WL |
1506 | Recurse to a named subpattern. Identical to C<(?PARNO)> except that the |
1507 | parenthesis to recurse to is determined by name. If multiple parentheses have | |
894be9b7 YO |
1508 | the same name, then it recurses to the leftmost. |
1509 | ||
1510 | It is an error to refer to a name that is not declared somewhere in the | |
1511 | pattern. | |
1512 | ||
1f1031fe YO |
1513 | B<NOTE:> In order to make things easier for programmers with experience |
1514 | with the Python or PCRE regex engines the pattern C<< (?P>NAME) >> | |
64c5a566 | 1515 | may be used instead of C<< (?&NAME) >>. |
1f1031fe | 1516 | |
e2e6a0f1 YO |
1517 | =item C<(?(condition)yes-pattern|no-pattern)> |
1518 | X<(?()> | |
286f584a | 1519 | |
e2e6a0f1 | 1520 | =item C<(?(condition)yes-pattern)> |
286f584a | 1521 | |
41ef34de ML |
1522 | Conditional expression. Matches C<yes-pattern> if C<condition> yields |
1523 | a true value, matches C<no-pattern> otherwise. A missing pattern always | |
1524 | matches. | |
1525 | ||
25e26d77 | 1526 | C<(condition)> should be one of: 1) an integer in |
e2e6a0f1 | 1527 | parentheses (which is valid if the corresponding pair of parentheses |
25e26d77 | 1528 | matched); 2) a look-ahead/look-behind/evaluate zero-width assertion; 3) a |
c27a5cfe | 1529 | name in angle brackets or single quotes (which is valid if a group |
25e26d77 | 1530 | with the given name matched); or 4) the special symbol (R) (true when |
e2e6a0f1 YO |
1531 | evaluated inside of recursion or eval). Additionally the R may be |
1532 | followed by a number, (which will be true when evaluated when recursing | |
1533 | inside of the appropriate group), or by C<&NAME>, in which case it will | |
1534 | be true only when evaluated during recursion in the named group. | |
1535 | ||
1536 | Here's a summary of the possible predicates: | |
1537 | ||
1538 | =over 4 | |
1539 | ||
1540 | =item (1) (2) ... | |
1541 | ||
c27a5cfe | 1542 | Checks if the numbered capturing group has matched something. |
e2e6a0f1 YO |
1543 | |
1544 | =item (<NAME>) ('NAME') | |
1545 | ||
c27a5cfe | 1546 | Checks if a group with the given name has matched something. |
e2e6a0f1 | 1547 | |
f01cd190 FC |
1548 | =item (?=...) (?!...) (?<=...) (?<!...) |
1549 | ||
1550 | Checks whether the pattern matches (or does not match, for the '!' | |
1551 | variants). | |
1552 | ||
e2e6a0f1 YO |
1553 | =item (?{ CODE }) |
1554 | ||
f01cd190 | 1555 | Treats the return value of the code block as the condition. |
e2e6a0f1 YO |
1556 | |
1557 | =item (R) | |
1558 | ||
1559 | Checks if the expression has been evaluated inside of recursion. | |
1560 | ||
1561 | =item (R1) (R2) ... | |
1562 | ||
1563 | Checks if the expression has been evaluated while executing directly | |
1564 | inside of the n-th capture group. This check is the regex equivalent of | |
1565 | ||
1566 | if ((caller(0))[3] eq 'subname') { ... } | |
1567 | ||
1568 | In other words, it does not check the full recursion stack. | |
1569 | ||
1570 | =item (R&NAME) | |
1571 | ||
1572 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
1573 | directly inside of the leftmost group with a given name (this is the same | |
1574 | logic used by C<(?&NAME)> to disambiguate). It does not check the full | |
1575 | stack, but only the name of the innermost active recursion. | |
1576 | ||
1577 | =item (DEFINE) | |
1578 | ||
1579 | In this case, the yes-pattern is never directly executed, and no | |
1580 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
1581 | See below for details. | |
1582 | ||
1583 | =back | |
1584 | ||
1585 | For example: | |
1586 | ||
1587 | m{ ( \( )? | |
1588 | [^()]+ | |
1589 | (?(1) \) ) | |
1590 | }x | |
1591 | ||
1592 | matches a chunk of non-parentheses, possibly included in parentheses | |
1593 | themselves. | |
1594 | ||
0b928c2f FC |
1595 | A special form is the C<(DEFINE)> predicate, which never executes its |
1596 | yes-pattern directly, and does not allow a no-pattern. This allows one to | |
1597 | define subpatterns which will be executed only by the recursion mechanism. | |
e2e6a0f1 YO |
1598 | This way, you can define a set of regular expression rules that can be |
1599 | bundled into any pattern you choose. | |
1600 | ||
1601 | It is recommended that for this usage you put the DEFINE block at the | |
1602 | end of the pattern, and that you name any subpatterns defined within it. | |
1603 | ||
1604 | Also, it's worth noting that patterns defined this way probably will | |
1605 | not be as efficient, as the optimiser is not very clever about | |
1606 | handling them. | |
1607 | ||
1608 | An example of how this might be used is as follows: | |
1609 | ||
2bf803e2 | 1610 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 1611 | (?(DEFINE) |
2bf803e2 YO |
1612 | (?<NAME_PAT>....) |
1613 | (?<ADRESS_PAT>....) | |
e2e6a0f1 YO |
1614 | )/x |
1615 | ||
c27a5cfe KW |
1616 | Note that capture groups matched inside of recursion are not accessible |
1617 | after the recursion returns, so the extra layer of capturing groups is | |
e2e6a0f1 YO |
1618 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
1619 | C<$+{NAME}> would be. | |
286f584a | 1620 | |
51a1303c BF |
1621 | Finally, keep in mind that subpatterns created inside a DEFINE block |
1622 | count towards the absolute and relative number of captures, so this: | |
1623 | ||
1624 | my @captures = "a" =~ /(.) # First capture | |
1625 | (?(DEFINE) | |
1626 | (?<EXAMPLE> 1 ) # Second capture | |
1627 | )/x; | |
1628 | say scalar @captures; | |
1629 | ||
1630 | Will output 2, not 1. This is particularly important if you intend to | |
1631 | compile the definitions with the C<qr//> operator, and later | |
1632 | interpolate them in another pattern. | |
1633 | ||
c47ff5f1 | 1634 | =item C<< (?>pattern) >> |
6bda09f9 | 1635 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 1636 | |
19799a22 GS |
1637 | An "independent" subexpression, one which matches the substring |
1638 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 1639 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
1640 | construct is useful for optimizations of what would otherwise be |
1641 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
1642 | It may also be useful in places where the "grab all you can, and do not |
1643 | give anything back" semantic is desirable. | |
19799a22 | 1644 | |
c47ff5f1 | 1645 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
1646 | (anchored at the beginning of string, as above) will match I<all> |
1647 | characters C<a> at the beginning of string, leaving no C<a> for | |
1648 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
1649 | since the match of the subgroup C<a*> is influenced by the following | |
1650 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
1651 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
1652 | this makes the tail match. | |
1653 | ||
0b928c2f FC |
1654 | C<< (?>pattern) >> does not disable backtracking altogether once it has |
1655 | matched. It is still possible to backtrack past the construct, but not | |
1656 | into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar". | |
1657 | ||
c47ff5f1 | 1658 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
0b928c2f FC |
1659 | C<(?=(pattern))\g{-1}>. This matches the same substring as a standalone |
1660 | C<a+>, and the following C<\g{-1}> eats the matched string; it therefore | |
c47ff5f1 | 1661 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
1662 | (The difference between these two constructs is that the second one |
1663 | uses a capturing group, thus shifting ordinals of backreferences | |
1664 | in the rest of a regular expression.) | |
1665 | ||
1666 | Consider this pattern: | |
c277df42 | 1667 | |
871b0233 | 1668 | m{ \( |
e2e6a0f1 | 1669 | ( |
f793d64a | 1670 | [^()]+ # x+ |
e2e6a0f1 | 1671 | | |
871b0233 IZ |
1672 | \( [^()]* \) |
1673 | )+ | |
e2e6a0f1 | 1674 | \) |
871b0233 | 1675 | }x |
5a964f20 | 1676 | |
19799a22 GS |
1677 | That will efficiently match a nonempty group with matching parentheses |
1678 | two levels deep or less. However, if there is no such group, it | |
1679 | will take virtually forever on a long string. That's because there | |
1680 | are so many different ways to split a long string into several | |
1681 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
1682 | to a subpattern of the above pattern. Consider how the pattern | |
1683 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
1684 | seconds, but that each extra letter doubles this time. This | |
1685 | exponential performance will make it appear that your program has | |
14218588 | 1686 | hung. However, a tiny change to this pattern |
5a964f20 | 1687 | |
e2e6a0f1 YO |
1688 | m{ \( |
1689 | ( | |
f793d64a | 1690 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
e2e6a0f1 | 1691 | | |
871b0233 IZ |
1692 | \( [^()]* \) |
1693 | )+ | |
e2e6a0f1 | 1694 | \) |
871b0233 | 1695 | }x |
c277df42 | 1696 | |
c47ff5f1 | 1697 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
1698 | this yourself would be a productive exercise), but finishes in a fourth |
1699 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
0b928c2f FC |
1700 | however, that, when this construct is followed by a |
1701 | quantifier, it currently triggers a warning message under | |
9f1b1f2d | 1702 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 1703 | C<"matches null string many times in regex">. |
c277df42 | 1704 | |
c47ff5f1 | 1705 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 1706 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
1707 | This was only 4 times slower on a string with 1000000 C<a>s. |
1708 | ||
9da458fc IZ |
1709 | The "grab all you can, and do not give anything back" semantic is desirable |
1710 | in many situations where on the first sight a simple C<()*> looks like | |
1711 | the correct solution. Suppose we parse text with comments being delimited | |
1712 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 1713 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
1714 | the comment delimiter, because it may "give up" some whitespace if |
1715 | the remainder of the pattern can be made to match that way. The correct | |
1716 | answer is either one of these: | |
1717 | ||
1718 | (?>#[ \t]*) | |
1719 | #[ \t]*(?![ \t]) | |
1720 | ||
1721 | For example, to grab non-empty comments into $1, one should use either | |
1722 | one of these: | |
1723 | ||
1724 | / (?> \# [ \t]* ) ( .+ ) /x; | |
1725 | / \# [ \t]* ( [^ \t] .* ) /x; | |
1726 | ||
1727 | Which one you pick depends on which of these expressions better reflects | |
1728 | the above specification of comments. | |
1729 | ||
6bda09f9 YO |
1730 | In some literature this construct is called "atomic matching" or |
1731 | "possessive matching". | |
1732 | ||
b9b4dddf YO |
1733 | Possessive quantifiers are equivalent to putting the item they are applied |
1734 | to inside of one of these constructs. The following equivalences apply: | |
1735 | ||
1736 | Quantifier Form Bracketing Form | |
1737 | --------------- --------------- | |
1738 | PAT*+ (?>PAT*) | |
1739 | PAT++ (?>PAT+) | |
1740 | PAT?+ (?>PAT?) | |
1741 | PAT{min,max}+ (?>PAT{min,max}) | |
1742 | ||
e2e6a0f1 YO |
1743 | =back |
1744 | ||
1745 | =head2 Special Backtracking Control Verbs | |
1746 | ||
1747 | B<WARNING:> These patterns are experimental and subject to change or | |
0d017f4d | 1748 | removal in a future version of Perl. Their usage in production code should |
e2e6a0f1 YO |
1749 | be noted to avoid problems during upgrades. |
1750 | ||
1751 | These special patterns are generally of the form C<(*VERB:ARG)>. Unless | |
1752 | otherwise stated the ARG argument is optional; in some cases, it is | |
1753 | forbidden. | |
1754 | ||
1755 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 1756 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
1757 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
1758 | rules apply: | |
e2e6a0f1 | 1759 | |
5d458dd8 YO |
1760 | On failure, the C<$REGERROR> variable will be set to the ARG value of the |
1761 | verb pattern, if the verb was involved in the failure of the match. If the | |
1762 | ARG part of the pattern was omitted, then C<$REGERROR> will be set to the | |
1763 | name of the last C<(*MARK:NAME)> pattern executed, or to TRUE if there was | |
1764 | none. Also, the C<$REGMARK> variable will be set to FALSE. | |
e2e6a0f1 | 1765 | |
5d458dd8 YO |
1766 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
1767 | the C<$REGMARK> variable will be set to the name of the last | |
1768 | C<(*MARK:NAME)> pattern executed. See the explanation for the | |
1769 | C<(*MARK:NAME)> verb below for more details. | |
e2e6a0f1 | 1770 | |
5d458dd8 | 1771 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
0b928c2f | 1772 | and most other regex-related variables. They are not local to a scope, nor |
5d458dd8 YO |
1773 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. |
1774 | Use C<local> to localize changes to them to a specific scope if necessary. | |
e2e6a0f1 YO |
1775 | |
1776 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 1777 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 | 1778 | |
70ca8714 | 1779 | =over 3 |
e2e6a0f1 YO |
1780 | |
1781 | =item Verbs that take an argument | |
1782 | ||
1783 | =over 4 | |
1784 | ||
5d458dd8 | 1785 | =item C<(*PRUNE)> C<(*PRUNE:NAME)> |
f7819f85 | 1786 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 1787 | |
5d458dd8 YO |
1788 | This zero-width pattern prunes the backtracking tree at the current point |
1789 | when backtracked into on failure. Consider the pattern C<A (*PRUNE) B>, | |
1790 | where A and B are complex patterns. Until the C<(*PRUNE)> verb is reached, | |
1791 | A may backtrack as necessary to match. Once it is reached, matching | |
1792 | continues in B, which may also backtrack as necessary; however, should B | |
1793 | not match, then no further backtracking will take place, and the pattern | |
1794 | will fail outright at the current starting position. | |
54612592 YO |
1795 | |
1796 | The following example counts all the possible matching strings in a | |
1797 | pattern (without actually matching any of them). | |
1798 | ||
e2e6a0f1 | 1799 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1800 | print "Count=$count\n"; |
1801 | ||
1802 | which produces: | |
1803 | ||
1804 | aaab | |
1805 | aaa | |
1806 | aa | |
1807 | a | |
1808 | aab | |
1809 | aa | |
1810 | a | |
1811 | ab | |
1812 | a | |
1813 | Count=9 | |
1814 | ||
5d458dd8 | 1815 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 1816 | |
5d458dd8 | 1817 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
1818 | print "Count=$count\n"; |
1819 | ||
0b928c2f | 1820 | we prevent backtracking and find the count of the longest matching string |
353c6505 | 1821 | at each matching starting point like so: |
54612592 YO |
1822 | |
1823 | aaab | |
1824 | aab | |
1825 | ab | |
1826 | Count=3 | |
1827 | ||
5d458dd8 | 1828 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 1829 | |
5d458dd8 YO |
1830 | See also C<< (?>pattern) >> and possessive quantifiers for other ways to |
1831 | control backtracking. In some cases, the use of C<(*PRUNE)> can be | |
1832 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
1833 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
1834 | C<< (?>pattern) >> alone. | |
54612592 | 1835 | |
5d458dd8 YO |
1836 | =item C<(*SKIP)> C<(*SKIP:NAME)> |
1837 | X<(*SKIP)> | |
e2e6a0f1 | 1838 | |
5d458dd8 | 1839 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 1840 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
1841 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
1842 | of this pattern. This effectively means that the regex engine "skips" forward | |
1843 | to this position on failure and tries to match again, (assuming that | |
1844 | there is sufficient room to match). | |
1845 | ||
1846 | The name of the C<(*SKIP:NAME)> pattern has special significance. If a | |
1847 | C<(*MARK:NAME)> was encountered while matching, then it is that position | |
1848 | which is used as the "skip point". If no C<(*MARK)> of that name was | |
1849 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
1850 | without a name the "skip point" is where the match point was when | |
1851 | executing the (*SKIP) pattern. | |
1852 | ||
0b928c2f | 1853 | Compare the following to the examples in C<(*PRUNE)>; note the string |
24b23f37 YO |
1854 | is twice as long: |
1855 | ||
d1fbf752 KW |
1856 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
1857 | print "Count=$count\n"; | |
24b23f37 YO |
1858 | |
1859 | outputs | |
1860 | ||
1861 | aaab | |
1862 | aaab | |
1863 | Count=2 | |
1864 | ||
5d458dd8 | 1865 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 1866 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
1867 | C<(*SKIP)> was executed. |
1868 | ||
5d458dd8 | 1869 | =item C<(*MARK:NAME)> C<(*:NAME)> |
b16db30f | 1870 | X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)> |
5d458dd8 YO |
1871 | |
1872 | This zero-width pattern can be used to mark the point reached in a string | |
1873 | when a certain part of the pattern has been successfully matched. This | |
1874 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
1875 | forward to that point if backtracked into on failure. Any number of | |
b4222fa9 | 1876 | C<(*MARK)> patterns are allowed, and the NAME portion may be duplicated. |
5d458dd8 YO |
1877 | |
1878 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:NAME)> | |
1879 | can be used to "label" a pattern branch, so that after matching, the | |
1880 | program can determine which branches of the pattern were involved in the | |
1881 | match. | |
1882 | ||
1883 | When a match is successful, the C<$REGMARK> variable will be set to the | |
1884 | name of the most recently executed C<(*MARK:NAME)> that was involved | |
1885 | in the match. | |
1886 | ||
1887 | This can be used to determine which branch of a pattern was matched | |
c27a5cfe | 1888 | without using a separate capture group for each branch, which in turn |
5d458dd8 YO |
1889 | can result in a performance improvement, as perl cannot optimize |
1890 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
1891 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
1892 | ||
1893 | When a match has failed, and unless another verb has been involved in | |
1894 | failing the match and has provided its own name to use, the C<$REGERROR> | |
1895 | variable will be set to the name of the most recently executed | |
1896 | C<(*MARK:NAME)>. | |
1897 | ||
42ac7c82 | 1898 | See L</(*SKIP)> for more details. |
5d458dd8 | 1899 | |
b62d2d15 YO |
1900 | As a shortcut C<(*MARK:NAME)> can be written C<(*:NAME)>. |
1901 | ||
5d458dd8 YO |
1902 | =item C<(*THEN)> C<(*THEN:NAME)> |
1903 | ||
ac9d8485 | 1904 | This is similar to the "cut group" operator C<::> from Perl 6. Like |
5d458dd8 YO |
1905 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
1906 | failure, it causes the regex engine to try the next alternation in the | |
ac9d8485 FC |
1907 | innermost enclosing group (capturing or otherwise) that has alternations. |
1908 | The two branches of a C<(?(condition)yes-pattern|no-pattern)> do not | |
1909 | count as an alternation, as far as C<(*THEN)> is concerned. | |
5d458dd8 YO |
1910 | |
1911 | Its name comes from the observation that this operation combined with the | |
1912 | alternation operator (C<|>) can be used to create what is essentially a | |
1913 | pattern-based if/then/else block: | |
1914 | ||
1915 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
1916 | ||
1917 | Note that if this operator is used and NOT inside of an alternation then | |
1918 | it acts exactly like the C<(*PRUNE)> operator. | |
1919 | ||
1920 | / A (*PRUNE) B / | |
1921 | ||
1922 | is the same as | |
1923 | ||
1924 | / A (*THEN) B / | |
1925 | ||
1926 | but | |
1927 | ||
25e26d77 | 1928 | / ( A (*THEN) B | C ) / |
5d458dd8 YO |
1929 | |
1930 | is not the same as | |
1931 | ||
25e26d77 | 1932 | / ( A (*PRUNE) B | C ) / |
5d458dd8 YO |
1933 | |
1934 | as after matching the A but failing on the B the C<(*THEN)> verb will | |
1935 | backtrack and try C; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 1936 | |
cbeadc21 JV |
1937 | =back |
1938 | ||
1939 | =item Verbs without an argument | |
1940 | ||
1941 | =over 4 | |
1942 | ||
e2e6a0f1 YO |
1943 | =item C<(*COMMIT)> |
1944 | X<(*COMMIT)> | |
24b23f37 | 1945 | |
241e7389 | 1946 | This is the Perl 6 "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
1947 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
1948 | into on failure it causes the match to fail outright. No further attempts | |
1949 | to find a valid match by advancing the start pointer will occur again. | |
1950 | For example, | |
24b23f37 | 1951 | |
d1fbf752 KW |
1952 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
1953 | print "Count=$count\n"; | |
24b23f37 YO |
1954 | |
1955 | outputs | |
1956 | ||
1957 | aaab | |
1958 | Count=1 | |
1959 | ||
e2e6a0f1 YO |
1960 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
1961 | does not match, the regex engine will not try any further matching on the | |
1962 | rest of the string. | |
c277df42 | 1963 | |
e2e6a0f1 YO |
1964 | =item C<(*FAIL)> C<(*F)> |
1965 | X<(*FAIL)> X<(*F)> | |
9af228c6 | 1966 | |
e2e6a0f1 YO |
1967 | This pattern matches nothing and always fails. It can be used to force the |
1968 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
1969 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. | |
9af228c6 | 1970 | |
e2e6a0f1 | 1971 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 1972 | |
e2e6a0f1 YO |
1973 | =item C<(*ACCEPT)> |
1974 | X<(*ACCEPT)> | |
9af228c6 | 1975 | |
e2e6a0f1 YO |
1976 | B<WARNING:> This feature is highly experimental. It is not recommended |
1977 | for production code. | |
9af228c6 | 1978 | |
e2e6a0f1 YO |
1979 | This pattern matches nothing and causes the end of successful matching at |
1980 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
1981 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 1982 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 1983 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 1984 | |
c27a5cfe | 1985 | If the C<(*ACCEPT)> is inside of capturing groups then the groups are |
e2e6a0f1 YO |
1986 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
1987 | For instance: | |
9af228c6 | 1988 | |
e2e6a0f1 | 1989 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 1990 | |
e2e6a0f1 | 1991 | will match, and C<$1> will be C<AB> and C<$2> will be C<B>, C<$3> will not |
0b928c2f | 1992 | be set. If another branch in the inner parentheses was matched, such as in the |
e2e6a0f1 | 1993 | string 'ACDE', then the C<D> and C<E> would have to be matched as well. |
9af228c6 YO |
1994 | |
1995 | =back | |
c277df42 | 1996 | |
a0d0e21e LW |
1997 | =back |
1998 | ||
c07a80fd | 1999 | =head2 Backtracking |
d74e8afc | 2000 | X<backtrack> X<backtracking> |
c07a80fd | 2001 | |
35a734be IZ |
2002 | NOTE: This section presents an abstract approximation of regular |
2003 | expression behavior. For a more rigorous (and complicated) view of | |
2004 | the rules involved in selecting a match among possible alternatives, | |
0d017f4d | 2005 | see L<Combining RE Pieces>. |
35a734be | 2006 | |
c277df42 | 2007 | A fundamental feature of regular expression matching involves the |
5a964f20 | 2008 | notion called I<backtracking>, which is currently used (when needed) |
0d017f4d | 2009 | by all regular non-possessive expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
2010 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
2011 | internally, but the general principle outlined here is valid. | |
c07a80fd | 2012 | |
2013 | For a regular expression to match, the I<entire> regular expression must | |
2014 | match, not just part of it. So if the beginning of a pattern containing a | |
2015 | quantifier succeeds in a way that causes later parts in the pattern to | |
2016 | fail, the matching engine backs up and recalculates the beginning | |
2017 | part--that's why it's called backtracking. | |
2018 | ||
2019 | Here is an example of backtracking: Let's say you want to find the | |
2020 | word following "foo" in the string "Food is on the foo table.": | |
2021 | ||
2022 | $_ = "Food is on the foo table."; | |
2023 | if ( /\b(foo)\s+(\w+)/i ) { | |
f793d64a | 2024 | print "$2 follows $1.\n"; |
c07a80fd | 2025 | } |
2026 | ||
2027 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
2028 | finds a possible match right at the beginning of the string, and loads up | |
2029 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
2030 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 2031 | mistake and starts over again one character after where it had the |
c07a80fd | 2032 | tentative match. This time it goes all the way until the next occurrence |
2033 | of "foo". The complete regular expression matches this time, and you get | |
2034 | the expected output of "table follows foo." | |
2035 | ||
2036 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
2037 | everything between "foo" and "bar". Initially, you write something | |
2038 | like this: | |
2039 | ||
2040 | $_ = "The food is under the bar in the barn."; | |
2041 | if ( /foo(.*)bar/ ) { | |
f793d64a | 2042 | print "got <$1>\n"; |
c07a80fd | 2043 | } |
2044 | ||
2045 | Which perhaps unexpectedly yields: | |
2046 | ||
2047 | got <d is under the bar in the > | |
2048 | ||
2049 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 2050 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 2051 | to use minimal matching to make sure you get the text between a "foo" |
2052 | and the first "bar" thereafter. | |
2053 | ||
2054 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
2055 | got <d is under the > | |
2056 | ||
0d017f4d | 2057 | Here's another example. Let's say you'd like to match a number at the end |
b6e13d97 | 2058 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 2059 | So you write this: |
2060 | ||
2061 | $_ = "I have 2 numbers: 53147"; | |
f793d64a KW |
2062 | if ( /(.*)(\d*)/ ) { # Wrong! |
2063 | print "Beginning is <$1>, number is <$2>.\n"; | |
c07a80fd | 2064 | } |
2065 | ||
2066 | That won't work at all, because C<.*> was greedy and gobbled up the | |
2067 | whole string. As C<\d*> can match on an empty string the complete | |
2068 | regular expression matched successfully. | |
2069 | ||
8e1088bc | 2070 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 2071 | |
2072 | Here are some variants, most of which don't work: | |
2073 | ||
2074 | $_ = "I have 2 numbers: 53147"; | |
2075 | @pats = qw{ | |
f793d64a KW |
2076 | (.*)(\d*) |
2077 | (.*)(\d+) | |
2078 | (.*?)(\d*) | |
2079 | (.*?)(\d+) | |
2080 | (.*)(\d+)$ | |
2081 | (.*?)(\d+)$ | |
2082 | (.*)\b(\d+)$ | |
2083 | (.*\D)(\d+)$ | |
c07a80fd | 2084 | }; |
2085 | ||
2086 | for $pat (@pats) { | |
f793d64a KW |
2087 | printf "%-12s ", $pat; |
2088 | if ( /$pat/ ) { | |
2089 | print "<$1> <$2>\n"; | |
2090 | } else { | |
2091 | print "FAIL\n"; | |
2092 | } | |
c07a80fd | 2093 | } |
2094 | ||
2095 | That will print out: | |
2096 | ||
2097 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
2098 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
2099 | (.*?)(\d*) <> <> | |
2100 | (.*?)(\d+) <I have > <2> | |
2101 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
2102 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
2103 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
2104 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
2105 | ||
2106 | As you see, this can be a bit tricky. It's important to realize that a | |
2107 | regular expression is merely a set of assertions that gives a definition | |
2108 | of success. There may be 0, 1, or several different ways that the | |
2109 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
2110 | multiple ways it might succeed, you need to understand backtracking to |
2111 | know which variety of success you will achieve. | |
c07a80fd | 2112 | |
19799a22 | 2113 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 2114 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 2115 | followed by "123". You might try to write that as |
2116 | ||
871b0233 | 2117 | $_ = "ABC123"; |
f793d64a KW |
2118 | if ( /^\D*(?!123)/ ) { # Wrong! |
2119 | print "Yup, no 123 in $_\n"; | |
871b0233 | 2120 | } |
c07a80fd | 2121 | |
2122 | But that isn't going to match; at least, not the way you're hoping. It | |
2123 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 2124 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 2125 | |
4358a253 SS |
2126 | $x = 'ABC123'; |
2127 | $y = 'ABC445'; | |
c07a80fd | 2128 | |
4358a253 SS |
2129 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; |
2130 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
c07a80fd | 2131 | |
4358a253 SS |
2132 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; |
2133 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
c07a80fd | 2134 | |
2135 | This prints | |
2136 | ||
2137 | 2: got ABC | |
2138 | 3: got AB | |
2139 | 4: got ABC | |
2140 | ||
5f05dabc | 2141 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 2142 | general purpose version of test 1. The important difference between |
2143 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
2144 | backtracking, whereas test 1 will not. What's happening is | |
2145 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 2146 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 2147 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 2148 | fail. |
14218588 | 2149 | |
c07a80fd | 2150 | The search engine will initially match C<\D*> with "ABC". Then it will |
0b928c2f | 2151 | try to match C<(?!123)> with "123", which fails. But because |
c07a80fd | 2152 | a quantifier (C<\D*>) has been used in the regular expression, the |
2153 | search engine can backtrack and retry the match differently | |
54310121 | 2154 | in the hope of matching the complete regular expression. |
c07a80fd | 2155 | |
5a964f20 TC |
2156 | The pattern really, I<really> wants to succeed, so it uses the |
2157 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 2158 | time. Now there's indeed something following "AB" that is not |
14218588 | 2159 | "123". It's "C123", which suffices. |
c07a80fd | 2160 | |
14218588 GS |
2161 | We can deal with this by using both an assertion and a negation. |
2162 | We'll say that the first part in $1 must be followed both by a digit | |
2163 | and by something that's not "123". Remember that the look-aheads | |
2164 | are zero-width expressions--they only look, but don't consume any | |
2165 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 2166 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
2167 | ||
4358a253 SS |
2168 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; |
2169 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
c07a80fd | 2170 | |
2171 | 6: got ABC | |
2172 | ||
5a964f20 | 2173 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 2174 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 2175 | matches only if you're at the beginning of the line AND the end of the |
2176 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
2177 | regular expressions always means AND, except when you write an explicit OR | |
2178 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
2179 | although the attempted matches are made at different positions because "a" | |
2180 | is not a zero-width assertion, but a one-width assertion. | |
2181 | ||
0d017f4d | 2182 | B<WARNING>: Particularly complicated regular expressions can take |
14218588 | 2183 | exponential time to solve because of the immense number of possible |
0d017f4d | 2184 | ways they can use backtracking to try for a match. For example, without |
9da458fc IZ |
2185 | internal optimizations done by the regular expression engine, this will |
2186 | take a painfully long time to run: | |
c07a80fd | 2187 | |
e1901655 IZ |
2188 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
2189 | ||
2190 | And if you used C<*>'s in the internal groups instead of limiting them | |
2191 | to 0 through 5 matches, then it would take forever--or until you ran | |
2192 | out of stack space. Moreover, these internal optimizations are not | |
2193 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
2194 | on the external group, no current optimization is applicable, and the | |
2195 | match takes a long time to finish. | |
c07a80fd | 2196 | |
9da458fc IZ |
2197 | A powerful tool for optimizing such beasts is what is known as an |
2198 | "independent group", | |
96090e4f | 2199 | which does not backtrack (see L</C<< (?>pattern) >>>). Note also that |
9da458fc | 2200 | zero-length look-ahead/look-behind assertions will not backtrack to make |
5d458dd8 | 2201 | the tail match, since they are in "logical" context: only |
14218588 | 2202 | whether they match is considered relevant. For an example |
9da458fc | 2203 | where side-effects of look-ahead I<might> have influenced the |
96090e4f | 2204 | following match, see L</C<< (?>pattern) >>>. |
c277df42 | 2205 | |
a0d0e21e | 2206 | =head2 Version 8 Regular Expressions |
d74e8afc | 2207 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> |
a0d0e21e | 2208 | |
5a964f20 | 2209 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
2210 | routines, here are the pattern-matching rules not described above. |
2211 | ||
54310121 | 2212 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 2213 | with a special meaning described here or above. You can cause |
5a964f20 | 2214 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 2215 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
0d017f4d WL |
2216 | character; "\\" matches a "\"). This escape mechanism is also required |
2217 | for the character used as the pattern delimiter. | |
2218 | ||
2219 | A series of characters matches that series of characters in the target | |
0b928c2f | 2220 | string, so the pattern C<blurfl> would match "blurfl" in the target |
0d017f4d | 2221 | string. |
a0d0e21e LW |
2222 | |
2223 | You can specify a character class, by enclosing a list of characters | |
5d458dd8 | 2224 | in C<[]>, which will match any character from the list. If the |
a0d0e21e | 2225 | first character after the "[" is "^", the class matches any character not |
14218588 | 2226 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 2227 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
2228 | inclusive. If you want either "-" or "]" itself to be a member of a |
2229 | class, put it at the start of the list (possibly after a "^"), or | |
2230 | escape it with a backslash. "-" is also taken literally when it is | |
2231 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
2232 | following all specify the same class of three characters: C<[-az]>, |
2233 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
5d458dd8 YO |
2234 | specifies a class containing twenty-six characters, even on EBCDIC-based |
2235 | character sets.) Also, if you try to use the character | |
2236 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
2237 | a range, the "-" is understood literally. | |
a0d0e21e | 2238 | |
8ada0baa JH |
2239 | Note also that the whole range idea is rather unportable between |
2240 | character sets--and even within character sets they may cause results | |
2241 | you probably didn't expect. A sound principle is to use only ranges | |
0d017f4d | 2242 | that begin from and end at either alphabetics of equal case ([a-e], |
8ada0baa JH |
2243 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, |
2244 | spell out the character sets in full. | |
2245 | ||
54310121 | 2246 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
2247 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
2248 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
dc0d9c48 | 2249 | of three octal digits, matches the character whose coded character set value |
5d458dd8 | 2250 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, |
dc0d9c48 | 2251 | matches the character whose ordinal is I<nn>. The expression \cI<x> |
5d458dd8 | 2252 | matches the character control-I<x>. Finally, the "." metacharacter |
fb55449c | 2253 | matches any character except "\n" (unless you use C</s>). |
a0d0e21e LW |
2254 | |
2255 | You can specify a series of alternatives for a pattern using "|" to | |
2256 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 2257 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e | 2258 | first alternative includes everything from the last pattern delimiter |
0b928c2f | 2259 | ("(", "(?:", etc. or the beginning of the pattern) up to the first "|", and |
a0d0e21e | 2260 | the last alternative contains everything from the last "|" to the next |
0b928c2f | 2261 | closing pattern delimiter. That's why it's common practice to include |
14218588 | 2262 | alternatives in parentheses: to minimize confusion about where they |
a3cb178b GS |
2263 | start and end. |
2264 | ||
5a964f20 | 2265 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
2266 | alternative found for which the entire expression matches, is the one that |
2267 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 2268 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
2269 | part will match, as that is the first alternative tried, and it successfully |
2270 | matches the target string. (This might not seem important, but it is | |
2271 | important when you are capturing matched text using parentheses.) | |
2272 | ||
5a964f20 | 2273 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 2274 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 2275 | |
14218588 GS |
2276 | Within a pattern, you may designate subpatterns for later reference |
2277 | by enclosing them in parentheses, and you may refer back to the | |
2278 | I<n>th subpattern later in the pattern using the metacharacter | |
0b928c2f | 2279 | \I<n> or \gI<n>. Subpatterns are numbered based on the left to right order |
14218588 GS |
2280 | of their opening parenthesis. A backreference matches whatever |
2281 | actually matched the subpattern in the string being examined, not | |
d8b950dc | 2282 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\g1\d*> will |
14218588 GS |
2283 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern |
2284 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
2285 | the leading 0 in the second number. | |
cb1a09d0 | 2286 | |
0d017f4d | 2287 | =head2 Warning on \1 Instead of $1 |
cb1a09d0 | 2288 | |
5a964f20 | 2289 | Some people get too used to writing things like: |
cb1a09d0 AD |
2290 | |
2291 | $pattern =~ s/(\W)/\\\1/g; | |
2292 | ||
3ff1c45a KW |
2293 | This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid |
2294 | shocking the | |
cb1a09d0 | 2295 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 | 2296 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
2297 | the usual double-quoted string means a control-A. The customary Unix |
2298 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
2299 | of doing that, you get yourself into trouble if you then add an C</e> | |
2300 | modifier. | |
2301 | ||
f793d64a | 2302 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
2303 | |
2304 | Or if you try to do | |
2305 | ||
2306 | s/(\d+)/\1000/; | |
2307 | ||
2308 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 2309 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
2310 | with the operation of matching a backreference. Certainly they mean two |
2311 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 2312 | |
0d017f4d | 2313 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 2314 | |
19799a22 | 2315 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
2316 | |
2317 | Regular expressions provide a terse and powerful programming language. As | |
2318 | with most other power tools, power comes together with the ability | |
2319 | to wreak havoc. | |
2320 | ||
2321 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 2322 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
2323 | |
2324 | 'foo' =~ m{ ( o? )* }x; | |
2325 | ||
0d017f4d | 2326 | The C<o?> matches at the beginning of C<'foo'>, and since the position |
c84d73f1 | 2327 | in the string is not moved by the match, C<o?> would match again and again |
527e91da | 2328 | because of the C<*> quantifier. Another common way to create a similar cycle |
c84d73f1 IZ |
2329 | is with the looping modifier C<//g>: |
2330 | ||
2331 | @matches = ( 'foo' =~ m{ o? }xg ); | |
2332 | ||
2333 | or | |
2334 | ||
2335 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
2336 | ||
2337 | or the loop implied by split(). | |
2338 | ||
2339 | However, long experience has shown that many programming tasks may | |
14218588 GS |
2340 | be significantly simplified by using repeated subexpressions that |
2341 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 | 2342 | |
d1fbf752 | 2343 | @chars = split //, $string; # // is not magic in split |
c84d73f1 IZ |
2344 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
2345 | ||
9da458fc | 2346 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 2347 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 2348 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
c84d73f1 IZ |
2349 | ones like the C</g> modifier or split() operator. |
2350 | ||
19799a22 GS |
2351 | The lower-level loops are I<interrupted> (that is, the loop is |
2352 | broken) when Perl detects that a repeated expression matched a | |
2353 | zero-length substring. Thus | |
c84d73f1 IZ |
2354 | |
2355 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
2356 | ||
5d458dd8 | 2357 | is made equivalent to |
c84d73f1 | 2358 | |
0b928c2f FC |
2359 | m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x; |
2360 | ||
2361 | For example, this program | |
2362 | ||
2363 | #!perl -l | |
2364 | "aaaaab" =~ / | |
2365 | (?: | |
2366 | a # non-zero | |
2367 | | # or | |
2368 | (?{print "hello"}) # print hello whenever this | |
2369 | # branch is tried | |
2370 | (?=(b)) # zero-width assertion | |
2371 | )* # any number of times | |
2372 | /x; | |
2373 | print $&; | |
2374 | print $1; | |
c84d73f1 | 2375 | |
0b928c2f FC |
2376 | prints |
2377 | ||
2378 | hello | |
2379 | aaaaa | |
2380 | b | |
2381 | ||
2382 | Notice that "hello" is only printed once, as when Perl sees that the sixth | |
2383 | iteration of the outermost C<(?:)*> matches a zero-length string, it stops | |
2384 | the C<*>. | |
2385 | ||
2386 | The higher-level loops preserve an additional state between iterations: | |
5d458dd8 | 2387 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 2388 | match after a zero-length match is prohibited to have a length of zero. |
5d458dd8 | 2389 | This prohibition interacts with backtracking (see L<"Backtracking">), |
c84d73f1 IZ |
2390 | and so the I<second best> match is chosen if the I<best> match is of |
2391 | zero length. | |
2392 | ||
19799a22 | 2393 | For example: |
c84d73f1 IZ |
2394 | |
2395 | $_ = 'bar'; | |
2396 | s/\w??/<$&>/g; | |
2397 | ||
20fb949f | 2398 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 2399 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
2400 | best> match is what is matched by C<\w>. Thus zero-length matches |
2401 | alternate with one-character-long matches. | |
2402 | ||
5d458dd8 | 2403 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
2404 | position one notch further in the string. |
2405 | ||
19799a22 | 2406 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 2407 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
2408 | Zero-length matches at the end of the previous match are ignored |
2409 | during C<split>. | |
c84d73f1 | 2410 | |
0d017f4d | 2411 | =head2 Combining RE Pieces |
35a734be IZ |
2412 | |
2413 | Each of the elementary pieces of regular expressions which were described | |
2414 | before (such as C<ab> or C<\Z>) could match at most one substring | |
2415 | at the given position of the input string. However, in a typical regular | |
2416 | expression these elementary pieces are combined into more complicated | |
0b928c2f | 2417 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc. |
35a734be IZ |
2418 | (in these examples C<S> and C<T> are regular subexpressions). |
2419 | ||
2420 | Such combinations can include alternatives, leading to a problem of choice: | |
2421 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
2422 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
2423 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
2424 | However, this description is too low-level and makes you think | |
2425 | in terms of a particular implementation. | |
2426 | ||
2427 | Another description starts with notions of "better"/"worse". All the | |
2428 | substrings which may be matched by the given regular expression can be | |
2429 | sorted from the "best" match to the "worst" match, and it is the "best" | |
2430 | match which is chosen. This substitutes the question of "what is chosen?" | |
2431 | by the question of "which matches are better, and which are worse?". | |
2432 | ||
2433 | Again, for elementary pieces there is no such question, since at most | |
2434 | one match at a given position is possible. This section describes the | |
2435 | notion of better/worse for combining operators. In the description | |
2436 | below C<S> and C<T> are regular subexpressions. | |
2437 | ||
13a2d996 | 2438 | =over 4 |
35a734be IZ |
2439 | |
2440 | =item C<ST> | |
2441 | ||
2442 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
2443 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
5d458dd8 | 2444 | which can be matched by C<T>. |
35a734be | 2445 | |
0b928c2f | 2446 | If C<A> is a better match for C<S> than C<A'>, C<AB> is a better |
35a734be IZ |
2447 | match than C<A'B'>. |
2448 | ||
2449 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
0b928c2f | 2450 | C<B> is a better match for C<T> than C<B'>. |
35a734be IZ |
2451 | |
2452 | =item C<S|T> | |
2453 | ||
2454 | When C<S> can match, it is a better match than when only C<T> can match. | |
2455 | ||
2456 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
2457 | two matches for C<T>. | |
2458 | ||
2459 | =item C<S{REPEAT_COUNT}> | |
2460 | ||
2461 | Matches as C<SSS...S> (repeated as many times as necessary). | |
2462 | ||
2463 | =item C<S{min,max}> | |
2464 | ||
2465 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
2466 | ||
2467 | =item C<S{min,max}?> | |
2468 | ||
2469 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
2470 | ||
2471 | =item C<S?>, C<S*>, C<S+> | |
2472 | ||
2473 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
2474 | ||
2475 | =item C<S??>, C<S*?>, C<S+?> | |
2476 | ||
2477 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
2478 | ||
c47ff5f1 | 2479 | =item C<< (?>S) >> |
35a734be IZ |
2480 | |
2481 | Matches the best match for C<S> and only that. | |
2482 | ||
2483 | =item C<(?=S)>, C<(?<=S)> | |
2484 | ||
2485 | Only the best match for C<S> is considered. (This is important only if | |
2486 | C<S> has capturing parentheses, and backreferences are used somewhere | |
2487 | else in the whole regular expression.) | |
2488 | ||
2489 | =item C<(?!S)>, C<(?<!S)> | |
2490 | ||
2491 | For this grouping operator there is no need to describe the ordering, since | |
2492 | only whether or not C<S> can match is important. | |
2493 | ||
6bda09f9 | 2494 | =item C<(??{ EXPR })>, C<(?PARNO)> |
35a734be IZ |
2495 | |
2496 | The ordering is the same as for the regular expression which is | |
c27a5cfe | 2497 | the result of EXPR, or the pattern contained by capture group PARNO. |
35a734be IZ |
2498 | |
2499 | =item C<(?(condition)yes-pattern|no-pattern)> | |
2500 | ||
2501 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
2502 | already determined. The ordering of the matches is the same as for the | |
2503 | chosen subexpression. | |
2504 | ||
2505 | =back | |
2506 | ||
2507 | The above recipes describe the ordering of matches I<at a given position>. | |
2508 | One more rule is needed to understand how a match is determined for the | |
2509 | whole regular expression: a match at an earlier position is always better | |
2510 | than a match at a later position. | |
2511 | ||
0d017f4d | 2512 | =head2 Creating Custom RE Engines |
c84d73f1 | 2513 | |
0b928c2f FC |
2514 | As of Perl 5.10.0, one can create custom regular expression engines. This |
2515 | is not for the faint of heart, as they have to plug in at the C level. See | |
2516 | L<perlreapi> for more details. | |
2517 | ||
2518 | As an alternative, overloaded constants (see L<overload>) provide a simple | |
2519 | way to extend the functionality of the RE engine, by substituting one | |
2520 | pattern for another. | |
c84d73f1 IZ |
2521 | |
2522 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 2523 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
2524 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
2525 | at these positions, so we want to have each C<\Y|> in the place of the | |
2526 | more complicated version. We can create a module C<customre> to do | |
2527 | this: | |
2528 | ||
2529 | package customre; | |
2530 | use overload; | |
2531 | ||
2532 | sub import { | |
2533 | shift; | |
2534 | die "No argument to customre::import allowed" if @_; | |
2535 | overload::constant 'qr' => \&convert; | |
2536 | } | |
2537 | ||
2538 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
2539 | ||
580a9fe1 RGS |
2540 | # We must also take care of not escaping the legitimate \\Y| |
2541 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 2542 | my %rules = ( '\\' => '\\\\', |
f793d64a | 2543 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
c84d73f1 IZ |
2544 | sub convert { |
2545 | my $re = shift; | |
5d458dd8 | 2546 | $re =~ s{ |
c84d73f1 IZ |
2547 | \\ ( \\ | Y . ) |
2548 | } | |
5d458dd8 | 2549 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
2550 | return $re; |
2551 | } | |
2552 | ||
2553 | Now C<use customre> enables the new escape in constant regular | |
2554 | expressions, i.e., those without any runtime variable interpolations. | |
2555 | As documented in L<overload>, this conversion will work only over | |
2556 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
2557 | part of this regular expression needs to be converted explicitly | |
2558 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
2559 | ||
2560 | use customre; | |
2561 | $re = <>; | |
2562 | chomp $re; | |
2563 | $re = customre::convert $re; | |
2564 | /\Y|$re\Y|/; | |
2565 | ||
0b928c2f | 2566 | =head2 PCRE/Python Support |
1f1031fe | 2567 | |
0b928c2f | 2568 | As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions |
1f1031fe | 2569 | to the regex syntax. While Perl programmers are encouraged to use the |
0b928c2f | 2570 | Perl-specific syntax, the following are also accepted: |
1f1031fe YO |
2571 | |
2572 | =over 4 | |
2573 | ||
ae5648b3 | 2574 | =item C<< (?PE<lt>NAMEE<gt>pattern) >> |
1f1031fe | 2575 | |
c27a5cfe | 2576 | Define a named capture group. Equivalent to C<< (?<NAME>pattern) >>. |
1f1031fe YO |
2577 | |
2578 | =item C<< (?P=NAME) >> | |
2579 | ||
c27a5cfe | 2580 | Backreference to a named capture group. Equivalent to C<< \g{NAME} >>. |
1f1031fe YO |
2581 | |
2582 | =item C<< (?P>NAME) >> | |
2583 | ||
c27a5cfe | 2584 | Subroutine call to a named capture group. Equivalent to C<< (?&NAME) >>. |
1f1031fe | 2585 | |
ee9b8eae | 2586 | =back |
1f1031fe | 2587 | |
19799a22 GS |
2588 | =head1 BUGS |
2589 | ||
88c9975e KW |
2590 | Many regular expression constructs don't work on EBCDIC platforms. |
2591 | ||
ed7efc79 KW |
2592 | There are a number of issues with regard to case-insensitive matching |
2593 | in Unicode rules. See C<i> under L</Modifiers> above. | |
2594 | ||
9da458fc IZ |
2595 | This document varies from difficult to understand to completely |
2596 | and utterly opaque. The wandering prose riddled with jargon is | |
2597 | hard to fathom in several places. | |
2598 | ||
2599 | This document needs a rewrite that separates the tutorial content | |
2600 | from the reference content. | |
19799a22 GS |
2601 | |
2602 | =head1 SEE ALSO | |
9fa51da4 | 2603 | |
91e0c79e MJD |
2604 | L<perlrequick>. |
2605 | ||
2606 | L<perlretut>. | |
2607 | ||
9b599b2a GS |
2608 | L<perlop/"Regexp Quote-Like Operators">. |
2609 | ||
1e66bd83 PP |
2610 | L<perlop/"Gory details of parsing quoted constructs">. |
2611 | ||
14218588 GS |
2612 | L<perlfaq6>. |
2613 | ||
9b599b2a GS |
2614 | L<perlfunc/pos>. |
2615 | ||
2616 | L<perllocale>. | |
2617 | ||
fb55449c JH |
2618 | L<perlebcdic>. |
2619 | ||
14218588 GS |
2620 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
2621 | by O'Reilly and Associates. |