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 | |
c6ae04d3 KW |
10 | If you haven't used regular expressions before, a tutorial introduction |
11 | is available in L<perlretut>. If you know just a little about them, | |
12 | a quick-start introduction is available in L<perlrequick>. | |
13 | ||
c57a3350 KW |
14 | Except for L</The Basics> section, this page assumes you are familiar |
15 | with regular expression basics, like what is a "pattern", what does it | |
16 | look like, and how it is basically used. For a reference on how they | |
17 | are used, plus various examples of the same, see discussions of C<m//>, | |
18 | C<s///>, C<qr//> and C<"??"> in L<perlop/"Regexp Quote-Like Operators">. | |
cb1a09d0 | 19 | |
67cdf558 KW |
20 | New in v5.22, L<C<use re 'strict'>|re/'strict' mode> applies stricter |
21 | rules than otherwise when compiling regular expression patterns. It can | |
22 | find things that, while legal, may not be what you intended. | |
0d017f4d | 23 | |
c57a3350 KW |
24 | =head2 The Basics |
25 | X<regular expression, version 8> X<regex, version 8> X<regexp, version 8> | |
26 | ||
27 | Regular expressions are strings with the very particular syntax and | |
28 | meaning described in this document and auxiliary documents referred to | |
29 | by this one. The strings are called "patterns". Patterns are used to | |
30 | determine if some other string, called the "target", has (or doesn't | |
31 | have) the characteristics specified by the pattern. We call this | |
32 | "matching" the target string against the pattern. Usually the match is | |
33 | done by having the target be the first operand, and the pattern be the | |
34 | second operand, of one of the two binary operators C<=~> and C<!~>, | |
35 | listed in L<perlop/Binding Operators>; and the pattern will have been | |
36 | converted from an ordinary string by one of the operators in | |
37 | L<perlop/"Regexp Quote-Like Operators">, like so: | |
38 | ||
39 | $foo =~ m/abc/ | |
40 | ||
41 | This evaluates to true if and only if the string in the variable C<$foo> | |
42 | contains somewhere in it, the sequence of characters "a", "b", then "c". | |
43 | (The C<=~ m>, or match operator, is described in | |
44 | L<perlop/m/PATTERN/msixpodualngc>.) | |
45 | ||
46 | Patterns that aren't already stored in some variable must be delimitted, | |
47 | at both ends, by delimitter characters. These are often, as in the | |
48 | example above, forward slashes, and the typical way a pattern is written | |
49 | in documentation is with those slashes. In most cases, the delimitter | |
50 | is the same character, fore and aft, but there are a few cases where a | |
51 | character looks like it has a mirror-image mate, where the opening | |
52 | version is the beginning delimiter, and the closing one is the ending | |
53 | delimiter, like | |
54 | ||
55 | $foo =~ m<abc> | |
56 | ||
57 | Most times, the pattern is evaluated in double-quotish context, but it | |
58 | is possible to choose delimiters to force single-quotish, like | |
59 | ||
60 | $foo =~ m'abc' | |
61 | ||
62 | If the pattern contains its delimiter within it, that delimiter must be | |
63 | escaped. Prefixing it with a backslash (I<e.g.>, C<"/foo\/bar/">) | |
64 | serves this purpose. | |
65 | ||
66 | Any single character in a pattern matches that same character in the | |
67 | target string, unless the character is a I<metacharacter> with a special | |
68 | meaning described in this document. A sequence of non-metacharacters | |
69 | matches the same sequence in the target string, as we saw above with | |
70 | C<m/abc/>. | |
71 | ||
04c2d19c | 72 | Only a few characters (all of them being ASCII punctuation characters) |
c57a3350 KW |
73 | are metacharacters. The most commonly used one is a dot C<".">, which |
74 | normally matches almost any character (including a dot itself). | |
75 | ||
76 | You can cause characters that normally function as metacharacters to be | |
77 | interpreted literally by prefixing them with a C<"\">, just like the | |
78 | pattern's delimiter must be escaped if it also occurs within the | |
79 | pattern. Thus, C<"\."> matches just a literal dot, C<"."> instead of | |
80 | its normal meaning. This means that the backslash is also a | |
81 | metacharacter, so C<"\\"> matches a single C<"\">. And a sequence that | |
82 | contains an escaped metacharacter matches the same sequence (but without | |
83 | the escape) in the target string. So, the pattern C</blur\\fl/> would | |
84 | match any target string that contains the sequence C<"blur\fl">. | |
85 | ||
86 | The metacharacter C<"|"> is used to match one thing or another. Thus | |
87 | ||
88 | $foo =~ m/this|that/ | |
89 | ||
90 | is TRUE if and only if C<$foo> contains either the sequence C<"this"> or | |
91 | the sequence C<"that">. Like all metacharacters, prefixing the C<"|"> | |
92 | with a backslash makes it match the plain punctuation character; in its | |
93 | case, the VERTICAL LINE. | |
94 | ||
95 | $foo =~ m/this\|that/ | |
96 | ||
97 | is TRUE if and only if C<$foo> contains the sequence C<"this|that">. | |
98 | ||
99 | You aren't limited to just a single C<"|">. | |
100 | ||
101 | $foo =~ m/fee|fie|foe|fum/ | |
102 | ||
103 | is TRUE if and only if C<$foo> contains any of those 4 sequences from | |
104 | the children's story "Jack and the Beanstalk". | |
105 | ||
106 | As you can see, the C<"|"> binds less tightly than a sequence of | |
107 | ordinary characters. We can override this by using the grouping | |
108 | metacharacters, the parentheses C<"("> and C<")">. | |
109 | ||
110 | $foo =~ m/th(is|at) thing/ | |
111 | ||
112 | is TRUE if and only if C<$foo> contains either the sequence S<C<"this | |
113 | thing">> or the sequence S<C<"that thing">>. The portions of the string | |
114 | that match the portions of the pattern enclosed in parentheses are | |
115 | normally made available separately for use later in the pattern, | |
116 | substitution, or program. This is called "capturing", and it can get | |
117 | complicated. See L</Capture groups>. | |
118 | ||
119 | The first alternative includes everything from the last pattern | |
120 | delimiter (C<"(">, C<"(?:"> (described later), I<etc>. or the beginning | |
121 | of the pattern) up to the first C<"|">, and the last alternative | |
122 | contains everything from the last C<"|"> to the next closing pattern | |
123 | delimiter. That's why it's common practice to include alternatives in | |
124 | parentheses: to minimize confusion about where they start and end. | |
125 | ||
126 | Alternatives are tried from left to right, so the first | |
127 | alternative found for which the entire expression matches, is the one that | |
128 | is chosen. This means that alternatives are not necessarily greedy. For | |
129 | example: when matching C<foo|foot> against C<"barefoot">, only the C<"foo"> | |
130 | part will match, as that is the first alternative tried, and it successfully | |
131 | matches the target string. (This might not seem important, but it is | |
132 | important when you are capturing matched text using parentheses.) | |
133 | ||
134 | Besides taking away the special meaning of a metacharacter, a prefixed | |
135 | backslash changes some letter and digit characters away from matching | |
136 | just themselves to instead have special meaning. These are called | |
137 | "escape sequences", and all such are described in L<perlrebackslash>. A | |
138 | backslash sequence (of a letter or digit) that doesn't currently have | |
139 | special meaning to Perl will raise a warning if warnings are enabled, | |
140 | as those are reserved for potential future use. | |
141 | ||
142 | One such sequence is C<\b>, which matches a boundary of some sort. | |
143 | C<\b{wb}> and a few others give specialized types of boundaries. | |
144 | (They are all described in detail starting at | |
145 | L<perlrebackslash/\b{}, \b, \B{}, \B>.) Note that these don't match | |
146 | characters, but the zero-width spaces between characters. They are an | |
147 | example of a L<zero-width assertion|/Assertions>. Consider again, | |
148 | ||
149 | $foo =~ m/fee|fie|foe|fum/ | |
150 | ||
151 | It evaluates to TRUE if, besides those 4 words, any of the sequences | |
152 | "feed", "field", "Defoe", "fume", and many others are in C<$foo>. By | |
153 | judicious use of C<\b> (or better (because it is designed to handle | |
154 | natural language) C<\b{wb}>), we can make sure that only the Giant's | |
155 | words are matched: | |
156 | ||
157 | $foo =~ m/\b(fee|fie|foe|fum)\b/ | |
158 | $foo =~ m/\b{wb}(fee|fie|foe|fum)\b{wb}/ | |
159 | ||
160 | The final example shows that the characters C<"{"> and C<"}"> are | |
161 | metacharacters. | |
162 | ||
163 | Another use for escape sequences is to specify characters that cannot | |
164 | (or which you prefer not to) be written literally. These are described | |
165 | in detail in L<perlrebackslash/Character Escapes>, but the next three | |
166 | paragraphs briefly describe some of them. | |
167 | ||
168 | Various control characters can be written in C language style: C<"\n"> | |
169 | matches a newline, C<"\t"> a tab, C<"\r"> a carriage return, C<"\f"> a | |
170 | form feed, I<etc>. | |
171 | ||
172 | More generally, C<\I<nnn>>, where I<nnn> is a string of three octal | |
173 | digits, matches the character whose native code point is I<nnn>. You | |
174 | can easily run into trouble if you don't have exactly three digits. So | |
175 | always use three, or since Perl 5.14, you can use C<\o{...}> to specify | |
176 | any number of octal digits. | |
177 | ||
178 | Similarly, C<\xI<nn>>, where I<nn> are hexadecimal digits, matches the | |
179 | character whose native ordinal is I<nn>. Again, not using exactly two | |
180 | digits is a recipe for disaster, but you can use C<\x{...}> to specify | |
181 | any number of hex digits. | |
182 | ||
183 | Besides being a metacharacter, the C<"."> is an example of a "character | |
184 | class", something that can match any single character of a given set of | |
185 | them. In its case, the set is just about all possible characters. Perl | |
186 | predefines several character classes besides the C<".">; there is a | |
187 | separate reference page about just these, L<perlrecharclass>. | |
188 | ||
189 | You can define your own custom character classes, by putting into your | |
190 | pattern in the appropriate place(s), a list of all the characters you | |
191 | want in the set. You do this by enclosing the list within C<[]> bracket | |
192 | characters. These are called "bracketed character classes" when we are | |
193 | being precise, but often the word "bracketed" is dropped. (Dropping it | |
194 | usually doesn't cause confusion.) This means that the C<"["> character | |
dabde021 | 195 | is another metacharacter. It doesn't match anything just by itself; it |
c57a3350 KW |
196 | is used only to tell Perl that what follows it is a bracketed character |
197 | class. If you want to match a literal left square bracket, you must | |
198 | escape it, like C<"\[">. The matching C<"]"> is also a metacharacter; | |
199 | again it doesn't match anything by itself, but just marks the end of | |
200 | your custom class to Perl. It is an example of a "sometimes | |
201 | metacharacter". It isn't a metacharacter if there is no corresponding | |
202 | C<"[">, and matches its literal self: | |
203 | ||
204 | print "]" =~ /]/; # prints 1 | |
205 | ||
206 | The list of characters within the character class gives the set of | |
207 | characters matched by the class. C<"[abc]"> matches a single "a" or "b" | |
208 | or "c". But if the first character after the C<"["> is C<"^">, the | |
84fefe3f KW |
209 | class instead matches any character not in the list. Within a list, the |
210 | C<"-"> character specifies a range of characters, so that C<a-z> | |
211 | represents all characters between "a" and "z", inclusive. If you want | |
212 | either C<"-"> or C<"]"> itself to be a member of a class, put it at the | |
213 | start of the list (possibly after a C<"^">), or escape it with a | |
214 | backslash. C<"-"> is also taken literally when it is at the end of the | |
215 | list, just before the closing C<"]">. (The following all specify the | |
216 | same class of three characters: C<[-az]>, C<[az-]>, and C<[a\-z]>. All | |
217 | are different from C<[a-z]>, which specifies a class containing | |
218 | twenty-six characters, even on EBCDIC-based character sets.) | |
c57a3350 KW |
219 | |
220 | There is lots more to bracketed character classes; full details are in | |
221 | L<perlrecharclass/Bracketed Character Classes>. | |
222 | ||
223 | =head3 Metacharacters | |
224 | X<metacharacter> | |
225 | X<\> X<^> X<.> X<$> X<|> X<(> X<()> X<[> X<[]> | |
226 | ||
227 | L</The Basics> introduced some of the metacharacters. This section | |
228 | gives them all. Most of them have the same meaning as in the I<egrep> | |
229 | command. | |
230 | ||
231 | Only the C<"\"> is always a metacharacter. The others are metacharacters | |
232 | just sometimes. The following tables lists all of them, summarizes | |
233 | their use, and gives the contexts where they are metacharacters. | |
234 | Outside those contexts or if prefixed by a C<"\">, they match their | |
235 | corresponding punctuation character. In some cases, their meaning | |
236 | varies depending on various pattern modifiers that alter the default | |
237 | behaviors. See L</Modifiers>. | |
238 | ||
239 | ||
240 | PURPOSE WHERE | |
241 | \ Escape the next character Always, except when | |
242 | escaped by another \ | |
243 | ^ Match the beginning of the string Not in [] | |
244 | (or line, if /m is used) | |
245 | ^ Complement the [] class At the beginning of [] | |
246 | . Match any single character except newline Not in [] | |
247 | (under /s, includes newline) | |
248 | $ Match the end of the string Not in [], but can | |
249 | (or before newline at the end of the mean interpolate a | |
250 | string; or before any newline if /m is scalar | |
251 | used) | |
252 | | Alternation Not in [] | |
253 | () Grouping Not in [] | |
254 | [ Start Bracketed Character class Not in [] | |
255 | ] End Bracketed Character class Only in [], and | |
256 | not first | |
257 | * Matches the preceding element 0 or more Not in [] | |
258 | times | |
259 | + Matches the preceding element 1 or more Not in [] | |
260 | times | |
261 | ? Matches the preceding element 0 or 1 Not in [] | |
262 | times | |
263 | { Starts a sequence that gives number(s) Not in [] | |
264 | of times the preceding element can be | |
265 | matched | |
266 | { when following certain escape sequences | |
267 | starts a modifier to the meaning of the | |
268 | sequence | |
269 | } End sequence started by { | |
270 | - Indicates a range Only in [] interior | |
93c42b81 | 271 | # Beginning of comment, extends to line end Only with /x modifier |
c57a3350 KW |
272 | |
273 | Notice that most of the metacharacters lose their special meaning when | |
274 | they occur in a bracketed character class, except C<"^"> has a different | |
275 | meaning when it is at the beginning of such a class. And C<"-"> and C<"]"> | |
276 | are metacharacters only at restricted positions within bracketed | |
277 | character classes; while C<"}"> is a metacharacter only when closing a | |
278 | special construct started by C<"{">. | |
279 | ||
280 | In double-quotish context, as is usually the case, you need to be | |
281 | careful about C<"$"> and the non-metacharacter C<"@">. Those could | |
282 | interpolate variables, which may or may not be what you intended. | |
283 | ||
284 | These rules were designed for compactness of expression, rather than | |
285 | legibility and maintainability. The L</E<sol>x and E<sol>xx> pattern | |
286 | modifiers allow you to insert white space to improve readability. And | |
287 | use of S<C<L<re 'strict'|re/'strict' mode>>> adds extra checking to | |
288 | catch some typos that might silently compile into something unintended. | |
289 | ||
290 | By default, the C<"^"> character is guaranteed to match only the | |
291 | beginning of the string, the C<"$"> character only the end (or before the | |
292 | newline at the end), and Perl does certain optimizations with the | |
293 | assumption that the string contains only one line. Embedded newlines | |
294 | will not be matched by C<"^"> or C<"$">. You may, however, wish to treat a | |
295 | string as a multi-line buffer, such that the C<"^"> will match after any | |
296 | newline within the string (except if the newline is the last character in | |
297 | the string), and C<"$"> will match before any newline. At the | |
298 | cost of a little more overhead, you can do this by using the | |
eb992c6f | 299 | C<L</E<sol>m>> modifier on the pattern match operator. (Older programs |
c57a3350 KW |
300 | did this by setting C<$*>, but this option was removed in perl 5.10.) |
301 | X<^> X<$> X</m> | |
302 | ||
303 | To simplify multi-line substitutions, the C<"."> character never matches a | |
304 | newline unless you use the L<C<E<sol>s>|/s> modifier, which in effect tells | |
305 | Perl to pretend the string is a single line--even if it isn't. | |
306 | X<.> X</s> | |
307 | ||
0d017f4d WL |
308 | =head2 Modifiers |
309 | ||
4cb6b395 KW |
310 | =head3 Overview |
311 | ||
3644e48f KW |
312 | The default behavior for matching can be changed, using various |
313 | modifiers. Modifiers that relate to the interpretation of the pattern | |
314 | are listed just below. Modifiers that alter the way a pattern is used | |
315 | by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and | |
a19f6621 JK |
316 | L<perlop/"Gory details of parsing quoted constructs">. Modifiers can be added |
317 | dynamically; see L</Extended Patterns> below. | |
a0d0e21e | 318 | |
55497cff | 319 | =over 4 |
320 | ||
7711f978 | 321 | =item B<C<m>> |
d74e8afc | 322 | X</m> X<regex, multiline> X<regexp, multiline> X<regular expression, multiline> |
55497cff | 323 | |
3644e48f | 324 | Treat the string being matched against as multiple lines. That is, change C<"^"> and C<"$"> from matching |
1ca12bda RS |
325 | the start of the string's first line and the end of its last line to |
326 | matching the start and end of each line within the string. | |
55497cff | 327 | |
7711f978 | 328 | =item B<C<s>> |
d74e8afc ITB |
329 | X</s> X<regex, single-line> X<regexp, single-line> |
330 | X<regular expression, single-line> | |
55497cff | 331 | |
7711f978 | 332 | Treat the string as single line. That is, change C<"."> to match any character |
19799a22 | 333 | whatsoever, even a newline, which normally it would not match. |
55497cff | 334 | |
7711f978 KW |
335 | Used together, as C</ms>, they let the C<"."> match any character whatsoever, |
336 | while still allowing C<"^"> and C<"$"> to match, respectively, just after | |
19799a22 | 337 | and just before newlines within the string. |
7b8d334a | 338 | |
7711f978 | 339 | =item B<C<i>> |
87e95b7f YO |
340 | X</i> X<regex, case-insensitive> X<regexp, case-insensitive> |
341 | X<regular expression, case-insensitive> | |
342 | ||
7711f978 KW |
343 | Do case-insensitive pattern matching. For example, "A" will match "a" |
344 | under C</i>. | |
87e95b7f | 345 | |
5027a30b KW |
346 | If locale matching rules are in effect, the case map is taken from the |
347 | current | |
17580e7a | 348 | locale for code points less than 255, and from Unicode rules for larger |
ed7efc79 | 349 | code points. However, matches that would cross the Unicode |
7711f978 KW |
350 | rules/non-Unicode rules boundary (ords 255/256) will not succeed, unless |
351 | the locale is a UTF-8 one. See L<perllocale>. | |
ed7efc79 | 352 | |
7711f978 KW |
353 | There are a number of Unicode characters that match a sequence of |
354 | multiple characters under C</i>. For example, | |
355 | C<LATIN SMALL LIGATURE FI> should match the sequence C<fi>. Perl is not | |
ed7efc79 KW |
356 | currently able to do this when the multiple characters are in the pattern and |
357 | are split between groupings, or when one or more are quantified. Thus | |
358 | ||
359 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi/i; # Matches | |
360 | "\N{LATIN SMALL LIGATURE FI}" =~ /[fi][fi]/i; # Doesn't match! | |
361 | "\N{LATIN SMALL LIGATURE FI}" =~ /fi*/i; # Doesn't match! | |
362 | ||
363 | # The below doesn't match, and it isn't clear what $1 and $2 would | |
364 | # be even if it did!! | |
365 | "\N{LATIN SMALL LIGATURE FI}" =~ /(f)(i)/i; # Doesn't match! | |
366 | ||
9d53c457 KW |
367 | Perl doesn't match multiple characters in a bracketed |
368 | character class unless the character that maps to them is explicitly | |
369 | mentioned, and it doesn't match them at all if the character class is | |
370 | inverted, which otherwise could be highly confusing. See | |
371 | L<perlrecharclass/Bracketed Character Classes>, and | |
1f59b283 KW |
372 | L<perlrecharclass/Negation>. |
373 | ||
77c8f263 | 374 | =item B<C<x>> and B<C<xx>> |
d74e8afc | 375 | X</x> |
55497cff | 376 | |
377 | Extend your pattern's legibility by permitting whitespace and comments. | |
77c8f263 | 378 | Details in L</E<sol>x and E<sol>xx> |
55497cff | 379 | |
7711f978 | 380 | =item B<C<p>> |
87e95b7f YO |
381 | X</p> X<regex, preserve> X<regexp, preserve> |
382 | ||
7711f978 KW |
383 | Preserve the string matched such that C<${^PREMATCH}>, C<${^MATCH}>, and |
384 | C<${^POSTMATCH}> are available for use after matching. | |
87e95b7f | 385 | |
13b0f67d | 386 | In Perl 5.20 and higher this is ignored. Due to a new copy-on-write |
7711f978 | 387 | mechanism, C<${^PREMATCH}>, C<${^MATCH}>, and C<${^POSTMATCH}> will be available |
13b0f67d DM |
388 | after the match regardless of the modifier. |
389 | ||
7711f978 | 390 | =item B<C<a>>, B<C<d>>, B<C<l>>, and B<C<u>> |
b6fa137b FC |
391 | X</a> X</d> X</l> X</u> |
392 | ||
850b7ec9 | 393 | These modifiers, all new in 5.14, affect which character-set rules |
57fbbe96 | 394 | (Unicode, I<etc>.) are used, as described below in |
ed7efc79 | 395 | L</Character set modifiers>. |
b6fa137b | 396 | |
7711f978 | 397 | =item B<C<n>> |
33be4c61 MH |
398 | X</n> X<regex, non-capture> X<regexp, non-capture> |
399 | X<regular expression, non-capture> | |
400 | ||
401 | Prevent the grouping metacharacters C<()> from capturing. This modifier, | |
57fbbe96 | 402 | new in 5.22, will stop C<$1>, C<$2>, I<etc>... from being filled in. |
33be4c61 MH |
403 | |
404 | "hello" =~ /(hi|hello)/; # $1 is "hello" | |
405 | "hello" =~ /(hi|hello)/n; # $1 is undef | |
406 | ||
25941dca | 407 | This is equivalent to putting C<?:> at the beginning of every capturing group: |
33be4c61 MH |
408 | |
409 | "hello" =~ /(?:hi|hello)/; # $1 is undef | |
410 | ||
411 | C</n> can be negated on a per-group basis. Alternatively, named captures | |
412 | may still be used. | |
413 | ||
414 | "hello" =~ /(?-n:(hi|hello))/n; # $1 is "hello" | |
bdafd784 KW |
415 | "hello" =~ /(?<greet>hi|hello)/n; # $1 is "hello", $+{greet} is |
416 | # "hello" | |
33be4c61 | 417 | |
7cf040c1 RS |
418 | =item Other Modifiers |
419 | ||
420 | There are a number of flags that can be found at the end of regular | |
421 | expression constructs that are I<not> generic regular expression flags, but | |
422 | apply to the operation being performed, like matching or substitution (C<m//> | |
423 | or C<s///> respectively). | |
424 | ||
425 | Flags described further in | |
426 | L<perlretut/"Using regular expressions in Perl"> are: | |
427 | ||
428 | c - keep the current position during repeated matching | |
429 | g - globally match the pattern repeatedly in the string | |
430 | ||
431 | Substitution-specific modifiers described in | |
33be4c61 | 432 | L<perlop/"s/PATTERN/REPLACEMENT/msixpodualngcer"> are: |
171e7319 | 433 | |
7cf040c1 RS |
434 | e - evaluate the right-hand side as an expression |
435 | ee - evaluate the right side as a string then eval the result | |
436 | o - pretend to optimize your code, but actually introduce bugs | |
437 | r - perform non-destructive substitution and return the new value | |
171e7319 | 438 | |
55497cff | 439 | =back |
a0d0e21e | 440 | |
516074bb | 441 | Regular expression modifiers are usually written in documentation |
57fbbe96 | 442 | as I<e.g.>, "the C</x> modifier", even though the delimiter |
4cb6b395 | 443 | in question might not really be a slash. The modifiers C</imnsxadlup> |
ab7bb42d | 444 | may also be embedded within the regular expression itself using |
ed7efc79 | 445 | the C<(?...)> construct, see L</Extended Patterns> below. |
b6fa137b | 446 | |
4cb6b395 KW |
447 | =head3 Details on some modifiers |
448 | ||
449 | Some of the modifiers require more explanation than given in the | |
450 | L</Overview> above. | |
451 | ||
77c8f263 | 452 | =head4 C</x> and C</xx> |
ed7efc79 | 453 | |
77c8f263 | 454 | A single C</x> tells |
7b059540 | 455 | the regular expression parser to ignore most whitespace that is neither |
7c688e65 | 456 | backslashed nor within a bracketed character class. You can use this to |
2ab07670 | 457 | break up your regular expression into more readable parts. |
7711f978 | 458 | Also, the C<"#"> character is treated as a metacharacter introducing a |
7c688e65 KW |
459 | comment that runs up to the pattern's closing delimiter, or to the end |
460 | of the current line if the pattern extends onto the next line. Hence, | |
461 | this is very much like an ordinary Perl code comment. (You can include | |
462 | the closing delimiter within the comment only if you precede it with a | |
463 | backslash, so be careful!) | |
464 | ||
465 | Use of C</x> means that if you want real | |
7711f978 | 466 | whitespace or C<"#"> characters in the pattern (outside a bracketed character |
7c688e65 | 467 | class, which is unaffected by C</x>), then you'll either have to |
7b059540 | 468 | escape them (using backslashes or C<\Q...\E>) or encode them using octal, |
673c254b | 469 | hex, or C<\N{}> or C<\p{name=...}> escapes. |
8be3c4ca KW |
470 | It is ineffective to try to continue a comment onto the next line by |
471 | escaping the C<\n> with a backslash or C<\Q>. | |
7c688e65 KW |
472 | |
473 | You can use L</(?#text)> to create a comment that ends earlier than the | |
474 | end of the current line, but C<text> also can't contain the closing | |
475 | delimiter unless escaped with a backslash. | |
476 | ||
57fbbe96 | 477 | A common pitfall is to forget that C<"#"> characters begin a comment under |
2ab07670 KW |
478 | C</x> and are not matched literally. Just keep that in mind when trying |
479 | to puzzle out why a particular C</x> pattern isn't working as expected. | |
480 | ||
57fbbe96 | 481 | Starting in Perl v5.26, if the modifier has a second C<"x"> within it, |
77c8f263 KW |
482 | it does everything that a single C</x> does, but additionally |
483 | non-backslashed SPACE and TAB characters within bracketed character | |
484 | classes are also generally ignored, and hence can be added to make the | |
485 | classes more readable. | |
486 | ||
487 | / [d-e g-i 3-7]/xx | |
488 | /[ ! @ " # $ % ^ & * () = ? <> ' ]/xx | |
489 | ||
490 | may be easier to grasp than the squashed equivalents | |
491 | ||
492 | /[d-eg-i3-7]/ | |
493 | /[!@"#$%^&*()=?<>']/ | |
494 | ||
7c688e65 KW |
495 | Taken together, these features go a long way towards |
496 | making Perl's regular expressions more readable. Here's an example: | |
497 | ||
498 | # Delete (most) C comments. | |
499 | $program =~ s { | |
500 | /\* # Match the opening delimiter. | |
501 | .*? # Match a minimal number of characters. | |
502 | \*/ # Match the closing delimiter. | |
503 | } []gsx; | |
504 | ||
505 | Note that anything inside | |
7651b971 | 506 | a C<\Q...\E> stays unaffected by C</x>. And note that C</x> doesn't affect |
0b928c2f | 507 | space interpretation within a single multi-character construct. For |
1b2f32d5 KW |
508 | example C<(?:...)> can't have a space between the C<"(">, |
509 | C<"?">, and C<":">. Within any delimiters for such a construct, allowed | |
510 | spaces are not affected by C</x>, and depend on the construct. For | |
511 | example, all constructs using curly braces as delimiters, such as | |
512 | C<\x{...}> can have blanks within but adjacent to the braces, but not | |
513 | elsewhere, and no non-blank space characters. An exception are Unicode | |
514 | properties which follow Unicode rules, for which see | |
9bb1f947 | 515 | L<perluniprops/Properties accessible through \p{} and \P{}>. |
d74e8afc | 516 | X</x> |
a0d0e21e | 517 | |
8373491a KW |
518 | The set of characters that are deemed whitespace are those that Unicode |
519 | calls "Pattern White Space", namely: | |
520 | ||
521 | U+0009 CHARACTER TABULATION | |
522 | U+000A LINE FEED | |
523 | U+000B LINE TABULATION | |
524 | U+000C FORM FEED | |
525 | U+000D CARRIAGE RETURN | |
526 | U+0020 SPACE | |
527 | U+0085 NEXT LINE | |
528 | U+200E LEFT-TO-RIGHT MARK | |
529 | U+200F RIGHT-TO-LEFT MARK | |
530 | U+2028 LINE SEPARATOR | |
531 | U+2029 PARAGRAPH SEPARATOR | |
532 | ||
4cb6b395 | 533 | =head4 Character set modifiers |
ed7efc79 KW |
534 | |
535 | C</d>, C</u>, C</a>, and C</l>, available starting in 5.14, are called | |
850b7ec9 | 536 | the character set modifiers; they affect the character set rules |
ed7efc79 KW |
537 | used for the regular expression. |
538 | ||
808432af KW |
539 | The C</d>, C</u>, and C</l> modifiers are not likely to be of much use |
540 | to you, and so you need not worry about them very much. They exist for | |
541 | Perl's internal use, so that complex regular expression data structures | |
542 | can be automatically serialized and later exactly reconstituted, | |
543 | including all their nuances. But, since Perl can't keep a secret, and | |
544 | there may be rare instances where they are useful, they are documented | |
545 | here. | |
ed7efc79 | 546 | |
808432af KW |
547 | The C</a> modifier, on the other hand, may be useful. Its purpose is to |
548 | allow code that is to work mostly on ASCII data to not have to concern | |
549 | itself with Unicode. | |
ca9560b2 | 550 | |
808432af KW |
551 | Briefly, C</l> sets the character set to that of whatever B<L>ocale is in |
552 | effect at the time of the execution of the pattern match. | |
ca9560b2 | 553 | |
808432af KW |
554 | C</u> sets the character set to B<U>nicode. |
555 | ||
556 | C</a> also sets the character set to Unicode, BUT adds several | |
557 | restrictions for B<A>SCII-safe matching. | |
558 | ||
559 | C</d> is the old, problematic, pre-5.14 B<D>efault character set | |
560 | behavior. Its only use is to force that old behavior. | |
561 | ||
562 | At any given time, exactly one of these modifiers is in effect. Their | |
563 | existence allows Perl to keep the originally compiled behavior of a | |
564 | regular expression, regardless of what rules are in effect when it is | |
565 | actually executed. And if it is interpolated into a larger regex, the | |
447fcf49 KW |
566 | original's rules continue to apply to it, and don't affect the other |
567 | parts. | |
808432af KW |
568 | |
569 | The C</l> and C</u> modifiers are automatically selected for | |
570 | regular expressions compiled within the scope of various pragmas, | |
571 | and we recommend that in general, you use those pragmas instead of | |
572 | specifying these modifiers explicitly. For one thing, the modifiers | |
573 | affect only pattern matching, and do not extend to even any replacement | |
7711f978 | 574 | done, whereas using the pragmas gives consistent results for all |
808432af KW |
575 | appropriate operations within their scopes. For example, |
576 | ||
577 | s/foo/\Ubar/il | |
578 | ||
579 | will match "foo" using the locale's rules for case-insensitive matching, | |
580 | but the C</l> does not affect how the C<\U> operates. Most likely you | |
581 | want both of them to use locale rules. To do this, instead compile the | |
582 | regular expression within the scope of C<use locale>. This both | |
ce4fe27b KW |
583 | implicitly adds the C</l>, and applies locale rules to the C<\U>. The |
584 | lesson is to C<use locale>, and not C</l> explicitly. | |
808432af KW |
585 | |
586 | Similarly, it would be better to use C<use feature 'unicode_strings'> | |
587 | instead of, | |
588 | ||
589 | s/foo/\Lbar/iu | |
590 | ||
591 | to get Unicode rules, as the C<\L> in the former (but not necessarily | |
592 | the latter) would also use Unicode rules. | |
593 | ||
594 | More detail on each of the modifiers follows. Most likely you don't | |
595 | need to know this detail for C</l>, C</u>, and C</d>, and can skip ahead | |
596 | to L<E<sol>a|/E<sol>a (and E<sol>aa)>. | |
ca9560b2 | 597 | |
ed7efc79 KW |
598 | =head4 /l |
599 | ||
600 | means to use the current locale's rules (see L<perllocale>) when pattern | |
601 | matching. For example, C<\w> will match the "word" characters of that | |
602 | locale, and C<"/i"> case-insensitive matching will match according to | |
603 | the locale's case folding rules. The locale used will be the one in | |
604 | effect at the time of execution of the pattern match. This may not be | |
605 | the same as the compilation-time locale, and can differ from one match | |
606 | to another if there is an intervening call of the | |
b6fa137b | 607 | L<setlocale() function|perllocale/The setlocale function>. |
ed7efc79 | 608 | |
7711f978 KW |
609 | Prior to v5.20, Perl did not support multi-byte locales. Starting then, |
610 | UTF-8 locales are supported. No other multi byte locales are ever | |
611 | likely to be supported. However, in all locales, one can have code | |
612 | points above 255 and these will always be treated as Unicode no matter | |
613 | what locale is in effect. | |
31f05a37 | 614 | |
ed7efc79 | 615 | Under Unicode rules, there are a few case-insensitive matches that cross |
31f05a37 KW |
616 | the 255/256 boundary. Except for UTF-8 locales in Perls v5.20 and |
617 | later, these are disallowed under C</l>. For example, 0xFF (on ASCII | |
618 | platforms) does not caselessly match the character at 0x178, C<LATIN | |
619 | CAPITAL LETTER Y WITH DIAERESIS>, because 0xFF may not be C<LATIN SMALL | |
620 | LETTER Y WITH DIAERESIS> in the current locale, and Perl has no way of | |
621 | knowing if that character even exists in the locale, much less what code | |
622 | point it is. | |
623 | ||
624 | In a UTF-8 locale in v5.20 and later, the only visible difference | |
625 | between locale and non-locale in regular expressions should be tainting | |
626 | (see L<perlsec>). | |
ed7efc79 KW |
627 | |
628 | This modifier may be specified to be the default by C<use locale>, but | |
629 | see L</Which character set modifier is in effect?>. | |
b6fa137b FC |
630 | X</l> |
631 | ||
ed7efc79 KW |
632 | =head4 /u |
633 | ||
634 | means to use Unicode rules when pattern matching. On ASCII platforms, | |
635 | this means that the code points between 128 and 255 take on their | |
808432af KW |
636 | Latin-1 (ISO-8859-1) meanings (which are the same as Unicode's). |
637 | (Otherwise Perl considers their meanings to be undefined.) Thus, | |
638 | under this modifier, the ASCII platform effectively becomes a Unicode | |
639 | platform; and hence, for example, C<\w> will match any of the more than | |
640 | 100_000 word characters in Unicode. | |
ed7efc79 KW |
641 | |
642 | Unlike most locales, which are specific to a language and country pair, | |
516074bb KW |
643 | Unicode classifies all the characters that are letters I<somewhere> in |
644 | the world as | |
ed7efc79 KW |
645 | C<\w>. For example, your locale might not think that C<LATIN SMALL |
646 | LETTER ETH> is a letter (unless you happen to speak Icelandic), but | |
647 | Unicode does. Similarly, all the characters that are decimal digits | |
648 | somewhere in the world will match C<\d>; this is hundreds, not 10, | |
649 | possible matches. And some of those digits look like some of the 10 | |
650 | ASCII digits, but mean a different number, so a human could easily think | |
651 | a number is a different quantity than it really is. For example, | |
652 | C<BENGALI DIGIT FOUR> (U+09EA) looks very much like an | |
8350b274 KW |
653 | C<ASCII DIGIT EIGHT> (U+0038), and C<LEPCHA DIGIT SIX> (U+1C46) looks |
654 | very much like an C<ASCII DIGIT FIVE> (U+0035). And, C<\d+>, may match | |
655 | strings of digits that are a mixture from different writing systems, | |
656 | creating a security issue. A fraudulent website, for example, could | |
657 | display the price of something using U+1C46, and it would appear to the | |
658 | user that something cost 500 units, but it really costs 600. A browser | |
659 | that enforced script runs (L</Script Runs>) would prevent that | |
660 | fraudulent display. L<Unicode::UCD/num()> can also be used to sort this | |
661 | out. Or the C</a> modifier can be used to force C<\d> to match just the | |
662 | ASCII 0 through 9. | |
ed7efc79 | 663 | |
516074bb KW |
664 | Also, under this modifier, case-insensitive matching works on the full |
665 | set of Unicode | |
ed7efc79 KW |
666 | characters. The C<KELVIN SIGN>, for example matches the letters "k" and |
667 | "K"; and C<LATIN SMALL LIGATURE FF> matches the sequence "ff", which, | |
668 | if you're not prepared, might make it look like a hexadecimal constant, | |
669 | presenting another potential security issue. See | |
e2176993 | 670 | L<https://unicode.org/reports/tr36> for a detailed discussion of Unicode |
ed7efc79 KW |
671 | security issues. |
672 | ||
ed7efc79 | 673 | This modifier may be specified to be the default by C<use feature |
66cbab2c KW |
674 | 'unicode_strings>, C<use locale ':not_characters'>, or |
675 | C<L<use 5.012|perlfunc/use VERSION>> (or higher), | |
808432af | 676 | but see L</Which character set modifier is in effect?>. |
b6fa137b FC |
677 | X</u> |
678 | ||
ed7efc79 KW |
679 | =head4 /d |
680 | ||
681 | This modifier means to use the "Default" native rules of the platform | |
682 | except when there is cause to use Unicode rules instead, as follows: | |
683 | ||
684 | =over 4 | |
685 | ||
686 | =item 1 | |
687 | ||
688 | the target string is encoded in UTF-8; or | |
689 | ||
690 | =item 2 | |
691 | ||
692 | the pattern is encoded in UTF-8; or | |
693 | ||
694 | =item 3 | |
695 | ||
696 | the pattern explicitly mentions a code point that is above 255 (say by | |
697 | C<\x{100}>); or | |
698 | ||
699 | =item 4 | |
b6fa137b | 700 | |
ed7efc79 KW |
701 | the pattern uses a Unicode name (C<\N{...}>); or |
702 | ||
703 | =item 5 | |
704 | ||
ce4fe27b | 705 | the pattern uses a Unicode property (C<\p{...}> or C<\P{...}>); or |
9d1a5160 KW |
706 | |
707 | =item 6 | |
708 | ||
64935bc6 KW |
709 | the pattern uses a Unicode break (C<\b{...}> or C<\B{...}>); or |
710 | ||
711 | =item 7 | |
712 | ||
eb992c6f | 713 | the pattern uses C<L</(?[ ])>> |
ed7efc79 | 714 | |
034602eb KW |
715 | =item 8 |
716 | ||
d9790612 | 717 | the pattern uses L<C<(*script_run: ...)>|/Script Runs> |
034602eb | 718 | |
ed7efc79 KW |
719 | =back |
720 | ||
721 | Another mnemonic for this modifier is "Depends", as the rules actually | |
722 | used depend on various things, and as a result you can get unexpected | |
808432af | 723 | results. See L<perlunicode/The "Unicode Bug">. The Unicode Bug has |
557ff1b2 KW |
724 | become rather infamous, leading to yet other (without swearing) names |
725 | for this modifier, "Dicey" and "Dodgy". | |
ed7efc79 | 726 | |
4b9734bf KW |
727 | Unless the pattern or string are encoded in UTF-8, only ASCII characters |
728 | can match positively. | |
ed7efc79 KW |
729 | |
730 | Here are some examples of how that works on an ASCII platform: | |
731 | ||
732 | $str = "\xDF"; # $str is not in UTF-8 format. | |
733 | $str =~ /^\w/; # No match, as $str isn't in UTF-8 format. | |
734 | $str .= "\x{0e0b}"; # Now $str is in UTF-8 format. | |
735 | $str =~ /^\w/; # Match! $str is now in UTF-8 format. | |
736 | chop $str; | |
737 | $str =~ /^\w/; # Still a match! $str remains in UTF-8 format. | |
738 | ||
808432af KW |
739 | This modifier is automatically selected by default when none of the |
740 | others are, so yet another name for it is "Default". | |
741 | ||
742 | Because of the unexpected behaviors associated with this modifier, you | |
7711f978 KW |
743 | probably should only explicitly use it to maintain weird backward |
744 | compatibilities. | |
808432af KW |
745 | |
746 | =head4 /a (and /aa) | |
747 | ||
3644e48f KW |
748 | This modifier stands for ASCII-restrict (or ASCII-safe). This modifier |
749 | may be doubled-up to increase its effect. | |
808432af KW |
750 | |
751 | When it appears singly, it causes the sequences C<\d>, C<\s>, C<\w>, and | |
752 | the Posix character classes to match only in the ASCII range. They thus | |
753 | revert to their pre-5.6, pre-Unicode meanings. Under C</a>, C<\d> | |
754 | always means precisely the digits C<"0"> to C<"9">; C<\s> means the five | |
779cf272 KW |
755 | characters C<[ \f\n\r\t]>, and starting in Perl v5.18, the vertical tab; |
756 | C<\w> means the 63 characters | |
808432af KW |
757 | C<[A-Za-z0-9_]>; and likewise, all the Posix classes such as |
758 | C<[[:print:]]> match only the appropriate ASCII-range characters. | |
759 | ||
760 | This modifier is useful for people who only incidentally use Unicode, | |
761 | and who do not wish to be burdened with its complexities and security | |
762 | concerns. | |
763 | ||
764 | With C</a>, one can write C<\d> with confidence that it will only match | |
765 | ASCII characters, and should the need arise to match beyond ASCII, you | |
766 | can instead use C<\p{Digit}> (or C<\p{Word}> for C<\w>). There are | |
767 | similar C<\p{...}> constructs that can match beyond ASCII both white | |
768 | space (see L<perlrecharclass/Whitespace>), and Posix classes (see | |
769 | L<perlrecharclass/POSIX Character Classes>). Thus, this modifier | |
770 | doesn't mean you can't use Unicode, it means that to get Unicode | |
771 | matching you must explicitly use a construct (C<\p{}>, C<\P{}>) that | |
772 | signals Unicode. | |
773 | ||
774 | As you would expect, this modifier causes, for example, C<\D> to mean | |
775 | the same thing as C<[^0-9]>; in fact, all non-ASCII characters match | |
776 | C<\D>, C<\S>, and C<\W>. C<\b> still means to match at the boundary | |
777 | between C<\w> and C<\W>, using the C</a> definitions of them (similarly | |
778 | for C<\B>). | |
779 | ||
780 | Otherwise, C</a> behaves like the C</u> modifier, in that | |
850b7ec9 | 781 | case-insensitive matching uses Unicode rules; for example, "k" will |
808432af KW |
782 | match the Unicode C<\N{KELVIN SIGN}> under C</i> matching, and code |
783 | points in the Latin1 range, above ASCII will have Unicode rules when it | |
784 | comes to case-insensitive matching. | |
785 | ||
786 | To forbid ASCII/non-ASCII matches (like "k" with C<\N{KELVIN SIGN}>), | |
7711f978 | 787 | specify the C<"a"> twice, for example C</aai> or C</aia>. (The first |
57fbbe96 | 788 | occurrence of C<"a"> restricts the C<\d>, I<etc>., and the second occurrence |
808432af KW |
789 | adds the C</i> restrictions.) But, note that code points outside the |
790 | ASCII range will use Unicode rules for C</i> matching, so the modifier | |
791 | doesn't really restrict things to just ASCII; it just forbids the | |
792 | intermixing of ASCII and non-ASCII. | |
793 | ||
794 | To summarize, this modifier provides protection for applications that | |
795 | don't wish to be exposed to all of Unicode. Specifying it twice | |
796 | gives added protection. | |
797 | ||
798 | This modifier may be specified to be the default by C<use re '/a'> | |
799 | or C<use re '/aa'>. If you do so, you may actually have occasion to use | |
31dc26d6 | 800 | the C</u> modifier explicitly if there are a few regular expressions |
808432af KW |
801 | where you do want full Unicode rules (but even here, it's best if |
802 | everything were under feature C<"unicode_strings">, along with the | |
803 | C<use re '/aa'>). Also see L</Which character set modifier is in | |
804 | effect?>. | |
805 | X</a> | |
806 | X</aa> | |
807 | ||
ed7efc79 KW |
808 | =head4 Which character set modifier is in effect? |
809 | ||
810 | Which of these modifiers is in effect at any given point in a regular | |
808432af KW |
811 | expression depends on a fairly complex set of interactions. These have |
812 | been designed so that in general you don't have to worry about it, but | |
813 | this section gives the gory details. As | |
ed7efc79 KW |
814 | explained below in L</Extended Patterns> it is possible to explicitly |
815 | specify modifiers that apply only to portions of a regular expression. | |
816 | The innermost always has priority over any outer ones, and one applying | |
6368643f KW |
817 | to the whole expression has priority over any of the default settings that are |
818 | described in the remainder of this section. | |
ed7efc79 | 819 | |
916cec3f | 820 | The C<L<use re 'E<sol>foo'|re/"'/flags' mode">> pragma can be used to set |
ed7efc79 KW |
821 | default modifiers (including these) for regular expressions compiled |
822 | within its scope. This pragma has precedence over the other pragmas | |
516074bb | 823 | listed below that also change the defaults. |
ed7efc79 KW |
824 | |
825 | Otherwise, C<L<use locale|perllocale>> sets the default modifier to C</l>; | |
66cbab2c | 826 | and C<L<use feature 'unicode_strings|feature>>, or |
ed7efc79 KW |
827 | C<L<use 5.012|perlfunc/use VERSION>> (or higher) set the default to |
828 | C</u> when not in the same scope as either C<L<use locale|perllocale>> | |
66cbab2c KW |
829 | or C<L<use bytes|bytes>>. |
830 | (C<L<use locale ':not_characters'|perllocale/Unicode and UTF-8>> also | |
831 | sets the default to C</u>, overriding any plain C<use locale>.) | |
832 | Unlike the mechanisms mentioned above, these | |
6368643f KW |
833 | affect operations besides regular expressions pattern matching, and so |
834 | give more consistent results with other operators, including using | |
57fbbe96 | 835 | C<\U>, C<\l>, I<etc>. in substitution replacements. |
ed7efc79 KW |
836 | |
837 | If none of the above apply, for backwards compatibility reasons, the | |
838 | C</d> modifier is the one in effect by default. As this can lead to | |
839 | unexpected results, it is best to specify which other rule set should be | |
840 | used. | |
841 | ||
842 | =head4 Character set modifier behavior prior to Perl 5.14 | |
843 | ||
844 | Prior to 5.14, there were no explicit modifiers, but C</l> was implied | |
845 | for regexes compiled within the scope of C<use locale>, and C</d> was | |
846 | implied otherwise. However, interpolating a regex into a larger regex | |
847 | would ignore the original compilation in favor of whatever was in effect | |
848 | at the time of the second compilation. There were a number of | |
849 | inconsistencies (bugs) with the C</d> modifier, where Unicode rules | |
850 | would be used when inappropriate, and vice versa. C<\p{}> did not imply | |
851 | Unicode rules, and neither did all occurrences of C<\N{}>, until 5.12. | |
b6fa137b | 852 | |
a0d0e21e LW |
853 | =head2 Regular Expressions |
854 | ||
04838cea RGS |
855 | =head3 Quantifiers |
856 | ||
3644e48f KW |
857 | Quantifiers are used when a particular portion of a pattern needs to |
858 | match a certain number (or numbers) of times. If there isn't a | |
859 | quantifier the number of times to match is exactly one. The following | |
860 | standard quantifiers are recognized: | |
d74e8afc | 861 | X<metacharacter> X<quantifier> X<*> X<+> X<?> X<{n}> X<{n,}> X<{n,m}> |
a0d0e21e | 862 | |
f793d64a KW |
863 | * Match 0 or more times |
864 | + Match 1 or more times | |
865 | ? Match 1 or 0 times | |
866 | {n} Match exactly n times | |
867 | {n,} Match at least n times | |
20420ba9 | 868 | {,n} Match at most n times |
f793d64a | 869 | {n,m} Match at least n but not more than m times |
a0d0e21e | 870 | |
3644e48f KW |
871 | (If a non-escaped curly bracket occurs in a context other than one of |
872 | the quantifiers listed above, where it does not form part of a | |
873 | backslashed sequence like C<\x{...}>, it is either a fatal syntax error, | |
874 | or treated as a regular character, generally with a deprecation warning | |
875 | raised. To escape it, you can precede it with a backslash (C<"\{">) or | |
876 | enclose it within square brackets (C<"[{]">). | |
877 | This change will allow for future syntax extensions (like making the | |
878 | lower bound of a quantifier optional), and better error checking of | |
879 | quantifiers). | |
9af81bfe | 880 | |
7711f978 KW |
881 | The C<"*"> quantifier is equivalent to C<{0,}>, the C<"+"> |
882 | quantifier to C<{1,}>, and the C<"?"> quantifier to C<{0,1}>. I<n> and I<m> are limited | |
d0b16107 | 883 | to non-negative integral values less than a preset limit defined when perl is built. |
80ac1a60 | 884 | This is usually 65534 on the most common platforms. The actual limit can |
9c79236d GS |
885 | be seen in the error message generated by code such as this: |
886 | ||
820475bd | 887 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 888 | |
54310121 | 889 | By default, a quantified subpattern is "greedy", that is, it will match as |
890 | many times as possible (given a particular starting location) while still | |
891 | allowing the rest of the pattern to match. If you want it to match the | |
7711f978 | 892 | minimum number of times possible, follow the quantifier with a C<"?">. Note |
54310121 | 893 | that the meanings don't change, just the "greediness": |
0d017f4d | 894 | X<metacharacter> X<greedy> X<greediness> |
20420ba9 | 895 | X<?> X<*?> X<+?> X<??> X<{n}?> X<{n,}?> X<{,n}?> X<{n,m}?> |
a0d0e21e | 896 | |
f793d64a KW |
897 | *? Match 0 or more times, not greedily |
898 | +? Match 1 or more times, not greedily | |
899 | ?? Match 0 or 1 time, not greedily | |
0b928c2f | 900 | {n}? Match exactly n times, not greedily (redundant) |
f793d64a | 901 | {n,}? Match at least n times, not greedily |
20420ba9 | 902 | {,n}? Match at most n times, not greedily |
f793d64a | 903 | {n,m}? Match at least n but not more than m times, not greedily |
a0d0e21e | 904 | |
5f3789aa | 905 | Normally when a quantified subpattern does not allow the rest of the |
b9b4dddf | 906 | overall pattern to match, Perl will backtrack. However, this behaviour is |
0d017f4d | 907 | sometimes undesirable. Thus Perl provides the "possessive" quantifier form |
b9b4dddf YO |
908 | as well. |
909 | ||
f793d64a KW |
910 | *+ Match 0 or more times and give nothing back |
911 | ++ Match 1 or more times and give nothing back | |
912 | ?+ Match 0 or 1 time and give nothing back | |
913 | {n}+ Match exactly n times and give nothing back (redundant) | |
914 | {n,}+ Match at least n times and give nothing back | |
20420ba9 | 915 | {,n}+ Match at most n times and give nothing back |
f793d64a | 916 | {n,m}+ Match at least n but not more than m times and give nothing back |
b9b4dddf YO |
917 | |
918 | For instance, | |
919 | ||
920 | 'aaaa' =~ /a++a/ | |
921 | ||
57fbbe96 | 922 | will never match, as the C<a++> will gobble up all the C<"a">'s in the |
b9b4dddf YO |
923 | string and won't leave any for the remaining part of the pattern. This |
924 | feature can be extremely useful to give perl hints about where it | |
925 | shouldn't backtrack. For instance, the typical "match a double-quoted | |
926 | string" problem can be most efficiently performed when written as: | |
927 | ||
928 | /"(?:[^"\\]++|\\.)*+"/ | |
929 | ||
0d017f4d | 930 | as we know that if the final quote does not match, backtracking will not |
0b928c2f | 931 | help. See the independent subexpression |
eb992c6f | 932 | C<L</(?E<gt>I<pattern>)>> for more details; |
b9b4dddf YO |
933 | possessive quantifiers are just syntactic sugar for that construct. For |
934 | instance the above example could also be written as follows: | |
935 | ||
936 | /"(?>(?:(?>[^"\\]+)|\\.)*)"/ | |
937 | ||
775b8798 | 938 | Note that the possessive quantifier modifier can not be combined |
5f3789aa YO |
939 | with the non-greedy modifier. This is because it would make no sense. |
940 | Consider the follow equivalency table: | |
941 | ||
942 | Illegal Legal | |
943 | ------------ ------ | |
944 | X??+ X{0} | |
945 | X+?+ X{1} | |
946 | X{min,max}?+ X{min} | |
947 | ||
04838cea RGS |
948 | =head3 Escape sequences |
949 | ||
0b928c2f | 950 | Because patterns are processed as double-quoted strings, the following |
a0d0e21e LW |
951 | also work: |
952 | ||
f793d64a KW |
953 | \t tab (HT, TAB) |
954 | \n newline (LF, NL) | |
955 | \r return (CR) | |
956 | \f form feed (FF) | |
957 | \a alarm (bell) (BEL) | |
958 | \e escape (think troff) (ESC) | |
f793d64a | 959 | \cK control char (example: VT) |
dc0d9c48 | 960 | \x{}, \x00 character whose ordinal is the given hexadecimal number |
fb121860 | 961 | \N{name} named Unicode character or character sequence |
f793d64a | 962 | \N{U+263D} Unicode character (example: FIRST QUARTER MOON) |
f0a2b745 | 963 | \o{}, \000 character whose ordinal is the given octal number |
f793d64a KW |
964 | \l lowercase next char (think vi) |
965 | \u uppercase next char (think vi) | |
ad81d09f KE |
966 | \L lowercase until \E (think vi) |
967 | \U uppercase until \E (think vi) | |
968 | \Q quote (disable) pattern metacharacters until \E | |
f793d64a | 969 | \E end either case modification or quoted section, think vi |
a0d0e21e | 970 | |
9bb1f947 | 971 | Details are in L<perlop/Quote and Quote-like Operators>. |
1d2dff63 | 972 | |
e1d1eefb | 973 | =head3 Character Classes and other Special Escapes |
04838cea | 974 | |
a0d0e21e | 975 | In addition, Perl defines the following: |
d0b16107 | 976 | X<\g> X<\k> X<\K> X<backreference> |
a0d0e21e | 977 | |
f793d64a KW |
978 | Sequence Note Description |
979 | [...] [1] Match a character according to the rules of the | |
980 | bracketed character class defined by the "...". | |
981 | Example: [a-z] matches "a" or "b" or "c" ... or "z" | |
982 | [[:...:]] [2] Match a character according to the rules of the POSIX | |
983 | character class "..." within the outer bracketed | |
984 | character class. Example: [[:upper:]] matches any | |
985 | uppercase character. | |
572224ce | 986 | (?[...]) [8] Extended bracketed character class |
d35dd6c6 KW |
987 | \w [3] Match a "word" character (alphanumeric plus "_", plus |
988 | other connector punctuation chars plus Unicode | |
0b928c2f | 989 | marks) |
f793d64a KW |
990 | \W [3] Match a non-"word" character |
991 | \s [3] Match a whitespace character | |
992 | \S [3] Match a non-whitespace character | |
993 | \d [3] Match a decimal digit character | |
994 | \D [3] Match a non-digit character | |
995 | \pP [3] Match P, named property. Use \p{Prop} for longer names | |
996 | \PP [3] Match non-P | |
997 | \X [4] Match Unicode "eXtended grapheme cluster" | |
c27a5cfe | 998 | \1 [5] Backreference to a specific capture group or buffer. |
f793d64a KW |
999 | '1' may actually be any positive integer. |
1000 | \g1 [5] Backreference to a specific or previous group, | |
1001 | \g{-1} [5] The number may be negative indicating a relative | |
c27a5cfe | 1002 | previous group and may optionally be wrapped in |
f793d64a KW |
1003 | curly brackets for safer parsing. |
1004 | \g{name} [5] Named backreference | |
1005 | \k<name> [5] Named backreference | |
a44b2be7 KW |
1006 | \k'name' [5] Named backreference |
1007 | \k{name} [5] Named backreference | |
f793d64a | 1008 | \K [6] Keep the stuff left of the \K, don't include it in $& |
2171640d | 1009 | \N [7] Any character but \n. Not affected by /s modifier |
f793d64a KW |
1010 | \v [3] Vertical whitespace |
1011 | \V [3] Not vertical whitespace | |
1012 | \h [3] Horizontal whitespace | |
1013 | \H [3] Not horizontal whitespace | |
1014 | \R [4] Linebreak | |
e1d1eefb | 1015 | |
9bb1f947 KW |
1016 | =over 4 |
1017 | ||
1018 | =item [1] | |
1019 | ||
1020 | See L<perlrecharclass/Bracketed Character Classes> for details. | |
df225385 | 1021 | |
9bb1f947 | 1022 | =item [2] |
b8c5462f | 1023 | |
9bb1f947 | 1024 | See L<perlrecharclass/POSIX Character Classes> for details. |
b8c5462f | 1025 | |
9bb1f947 | 1026 | =item [3] |
5496314a | 1027 | |
1532347b | 1028 | See L<perlunicode/Unicode Character Properties> for details |
5496314a | 1029 | |
9bb1f947 | 1030 | =item [4] |
5496314a | 1031 | |
9bb1f947 | 1032 | See L<perlrebackslash/Misc> for details. |
d0b16107 | 1033 | |
9bb1f947 | 1034 | =item [5] |
b8c5462f | 1035 | |
c27a5cfe | 1036 | See L</Capture groups> below for details. |
93733859 | 1037 | |
9bb1f947 | 1038 | =item [6] |
b8c5462f | 1039 | |
9bb1f947 KW |
1040 | See L</Extended Patterns> below for details. |
1041 | ||
1042 | =item [7] | |
1043 | ||
407fecf1 KW |
1044 | Note that C<\N> has two meanings. When of the form C<\N{I<NAME>}>, it |
1045 | matches the character or character sequence whose name is I<NAME>; and | |
1046 | similarly | |
fb121860 KW |
1047 | when of the form C<\N{U+I<hex>}>, it matches the character whose Unicode |
1048 | code point is I<hex>. Otherwise it matches any character but C<\n>. | |
9bb1f947 | 1049 | |
572224ce KW |
1050 | =item [8] |
1051 | ||
1052 | See L<perlrecharclass/Extended Bracketed Character Classes> for details. | |
1053 | ||
9bb1f947 | 1054 | =back |
d0b16107 | 1055 | |
04838cea RGS |
1056 | =head3 Assertions |
1057 | ||
c57a3350 KW |
1058 | Besides L<C<"^"> and C<"$">|/Metacharacters>, Perl defines the following |
1059 | zero-width assertions: | |
d74e8afc ITB |
1060 | X<zero-width assertion> X<assertion> X<regex, zero-width assertion> |
1061 | X<regexp, zero-width assertion> | |
1062 | X<regular expression, zero-width assertion> | |
1063 | X<\b> X<\B> X<\A> X<\Z> X<\z> X<\G> | |
a0d0e21e | 1064 | |
c57a3350 KW |
1065 | \b{} Match at Unicode boundary of specified type |
1066 | \B{} Match where corresponding \b{} doesn't match | |
1067 | \b Match a \w\W or \W\w boundary | |
1068 | \B Match except at a \w\W or \W\w boundary | |
1069 | \A Match only at beginning of string | |
1070 | \Z Match only at end of string, or before newline at the end | |
1071 | \z Match only at end of string | |
1072 | \G Match only at pos() (e.g. at the end-of-match position | |
9da458fc | 1073 | of prior m//g) |
a0d0e21e | 1074 | |
64935bc6 KW |
1075 | A Unicode boundary (C<\b{}>), available starting in v5.22, is a spot |
1076 | between two characters, or before the first character in the string, or | |
1077 | after the final character in the string where certain criteria defined | |
1078 | by Unicode are met. See L<perlrebackslash/\b{}, \b, \B{}, \B> for | |
1079 | details. | |
1080 | ||
14218588 | 1081 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
1082 | that has a C<\w> on one side of it and a C<\W> on the other side |
1083 | of it (in either order), counting the imaginary characters off the | |
1084 | beginning and end of the string as matching a C<\W>. (Within | |
1085 | character classes C<\b> represents backspace rather than a word | |
1086 | boundary, just as it normally does in any double-quoted string.) | |
7711f978 | 1087 | The C<\A> and C<\Z> are just like C<"^"> and C<"$">, except that they |
19799a22 | 1088 | won't match multiple times when the C</m> modifier is used, while |
7711f978 | 1089 | C<"^"> and C<"$"> will match at every internal line boundary. To match |
19799a22 GS |
1090 | the actual end of the string and not ignore an optional trailing |
1091 | newline, use C<\z>. | |
d74e8afc | 1092 | X<\b> X<\A> X<\Z> X<\z> X</m> |
19799a22 GS |
1093 | |
1094 | The C<\G> assertion can be used to chain global matches (using | |
1095 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
1096 | It is also useful when writing C<lex>-like scanners, when you have | |
1097 | several patterns that you want to match against consequent substrings | |
0b928c2f | 1098 | of your string; see the previous reference. The actual location |
19799a22 | 1099 | where C<\G> will match can also be influenced by using C<pos()> as |
58e23c8d | 1100 | an lvalue: see L<perlfunc/pos>. Note that the rule for zero-length |
0b928c2f FC |
1101 | matches (see L</"Repeated Patterns Matching a Zero-length Substring">) |
1102 | is modified somewhat, in that contents to the left of C<\G> are | |
58e23c8d YO |
1103 | not counted when determining the length of the match. Thus the following |
1104 | will not match forever: | |
d74e8afc | 1105 | X<\G> |
c47ff5f1 | 1106 | |
e761bb84 CO |
1107 | my $string = 'ABC'; |
1108 | pos($string) = 1; | |
1109 | while ($string =~ /(.\G)/g) { | |
1110 | print $1; | |
1111 | } | |
58e23c8d YO |
1112 | |
1113 | It will print 'A' and then terminate, as it considers the match to | |
1114 | be zero-width, and thus will not match at the same position twice in a | |
1115 | row. | |
1116 | ||
1117 | It is worth noting that C<\G> improperly used can result in an infinite | |
1118 | loop. Take care when using patterns that include C<\G> in an alternation. | |
1119 | ||
d5e7783a DM |
1120 | Note also that C<s///> will refuse to overwrite part of a substitution |
1121 | that has already been replaced; so for example this will stop after the | |
1122 | first iteration, rather than iterating its way backwards through the | |
1123 | string: | |
1124 | ||
1125 | $_ = "123456789"; | |
1126 | pos = 6; | |
1127 | s/.(?=.\G)/X/g; | |
1128 | print; # prints 1234X6789, not XXXXX6789 | |
1129 | ||
1130 | ||
c27a5cfe | 1131 | =head3 Capture groups |
04838cea | 1132 | |
3644e48f | 1133 | The grouping construct C<( ... )> creates capture groups (also referred to as |
c27a5cfe | 1134 | capture buffers). To refer to the current contents of a group later on, within |
d8b950dc KW |
1135 | the same pattern, use C<\g1> (or C<\g{1}>) for the first, C<\g2> (or C<\g{2}>) |
1136 | for the second, and so on. | |
1137 | This is called a I<backreference>. | |
d74e8afc | 1138 | X<regex, capture buffer> X<regexp, capture buffer> |
c27a5cfe | 1139 | X<regex, capture group> X<regexp, capture group> |
d74e8afc | 1140 | X<regular expression, capture buffer> X<backreference> |
c27a5cfe | 1141 | X<regular expression, capture group> X<backreference> |
1f1031fe | 1142 | X<\g{1}> X<\g{-1}> X<\g{name}> X<relative backreference> X<named backreference> |
d8b950dc KW |
1143 | X<named capture buffer> X<regular expression, named capture buffer> |
1144 | X<named capture group> X<regular expression, named capture group> | |
1145 | X<%+> X<$+{name}> X<< \k<name> >> | |
1146 | There is no limit to the number of captured substrings that you may use. | |
57fbbe96 | 1147 | Groups are numbered with the leftmost open parenthesis being number 1, I<etc>. If |
d8b950dc KW |
1148 | a group did not match, the associated backreference won't match either. (This |
1149 | can happen if the group is optional, or in a different branch of an | |
1150 | alternation.) | |
57fbbe96 | 1151 | You can omit the C<"g">, and write C<"\1">, I<etc>, but there are some issues with |
d8b950dc KW |
1152 | this form, described below. |
1153 | ||
1154 | You can also refer to capture groups relatively, by using a negative number, so | |
1155 | that C<\g-1> and C<\g{-1}> both refer to the immediately preceding capture | |
1156 | group, and C<\g-2> and C<\g{-2}> both refer to the group before it. For | |
1157 | example: | |
5624f11d YO |
1158 | |
1159 | / | |
c27a5cfe KW |
1160 | (Y) # group 1 |
1161 | ( # group 2 | |
1162 | (X) # group 3 | |
1163 | \g{-1} # backref to group 3 | |
1164 | \g{-3} # backref to group 1 | |
5624f11d YO |
1165 | ) |
1166 | /x | |
1167 | ||
d8b950dc KW |
1168 | would match the same as C</(Y) ( (X) \g3 \g1 )/x>. This allows you to |
1169 | interpolate regexes into larger regexes and not have to worry about the | |
1170 | capture groups being renumbered. | |
1171 | ||
1172 | You can dispense with numbers altogether and create named capture groups. | |
1173 | The notation is C<(?E<lt>I<name>E<gt>...)> to declare and C<\g{I<name>}> to | |
1174 | reference. (To be compatible with .Net regular expressions, C<\g{I<name>}> may | |
1175 | also be written as C<\k{I<name>}>, C<\kE<lt>I<name>E<gt>> or C<\k'I<name>'>.) | |
1176 | I<name> must not begin with a number, nor contain hyphens. | |
1177 | When different groups within the same pattern have the same name, any reference | |
1178 | to that name assumes the leftmost defined group. Named groups count in | |
1179 | absolute and relative numbering, and so can also be referred to by those | |
1180 | numbers. | |
1181 | (It's possible to do things with named capture groups that would otherwise | |
1182 | require C<(??{})>.) | |
1183 | ||
1184 | Capture group contents are dynamically scoped and available to you outside the | |
1185 | pattern until the end of the enclosing block or until the next successful | |
1186 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
1187 | You can refer to them by absolute number (using C<"$1"> instead of C<"\g1">, | |
57fbbe96 | 1188 | I<etc>); or by name via the C<%+> hash, using C<"$+{I<name>}">. |
d8b950dc KW |
1189 | |
1190 | Braces are required in referring to named capture groups, but are optional for | |
1191 | absolute or relative numbered ones. Braces are safer when creating a regex by | |
1192 | concatenating smaller strings. For example if you have C<qr/$a$b/>, and C<$a> | |
1193 | contained C<"\g1">, and C<$b> contained C<"37">, you would get C</\g137/> which | |
1194 | is probably not what you intended. | |
1195 | ||
1b2f32d5 KW |
1196 | If you use braces, you may also optionally add any number of blank |
1197 | (space or tab) characters within but adjacent to the braces, like | |
1198 | S<C<\g{ -1 }>>, or S<C<\k{ I<name> }>>. | |
1199 | ||
d8b950dc KW |
1200 | The C<\g> and C<\k> notations were introduced in Perl 5.10.0. Prior to that |
1201 | there were no named nor relative numbered capture groups. Absolute numbered | |
0b928c2f | 1202 | groups were referred to using C<\1>, |
57fbbe96 | 1203 | C<\2>, I<etc>., and this notation is still |
d8b950dc KW |
1204 | accepted (and likely always will be). But it leads to some ambiguities if |
1205 | there are more than 9 capture groups, as C<\10> could mean either the tenth | |
1206 | capture group, or the character whose ordinal in octal is 010 (a backspace in | |
1207 | ASCII). Perl resolves this ambiguity by interpreting C<\10> as a backreference | |
1208 | only if at least 10 left parentheses have opened before it. Likewise C<\11> is | |
1209 | a backreference only if at least 11 left parentheses have opened before it. | |
e1f120a9 KW |
1210 | And so on. C<\1> through C<\9> are always interpreted as backreferences. |
1211 | There are several examples below that illustrate these perils. You can avoid | |
1212 | the ambiguity by always using C<\g{}> or C<\g> if you mean capturing groups; | |
1213 | and for octal constants always using C<\o{}>, or for C<\077> and below, using 3 | |
1214 | digits padded with leading zeros, since a leading zero implies an octal | |
1215 | constant. | |
d8b950dc KW |
1216 | |
1217 | The C<\I<digit>> notation also works in certain circumstances outside | |
ed7efc79 | 1218 | the pattern. See L</Warning on \1 Instead of $1> below for details. |
81714fb9 | 1219 | |
14218588 | 1220 | Examples: |
a0d0e21e LW |
1221 | |
1222 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
1223 | ||
d8b950dc | 1224 | /(.)\g1/ # find first doubled char |
81714fb9 YO |
1225 | and print "'$1' is the first doubled character\n"; |
1226 | ||
1227 | /(?<char>.)\k<char>/ # ... a different way | |
1228 | and print "'$+{char}' is the first doubled character\n"; | |
1229 | ||
d8b950dc | 1230 | /(?'char'.)\g1/ # ... mix and match |
81714fb9 | 1231 | and print "'$1' is the first doubled character\n"; |
c47ff5f1 | 1232 | |
14218588 | 1233 | if (/Time: (..):(..):(..)/) { # parse out values |
f793d64a KW |
1234 | $hours = $1; |
1235 | $minutes = $2; | |
1236 | $seconds = $3; | |
a0d0e21e | 1237 | } |
c47ff5f1 | 1238 | |
9d860678 KW |
1239 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\g10/ # \g10 is a backreference |
1240 | /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/ # \10 is octal | |
1241 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/ # \10 is a backreference | |
1242 | /((.)(.)(.)(.)(.)(.)(.)(.)(.))\010/ # \010 is octal | |
1243 | ||
1244 | $a = '(.)\1'; # Creates problems when concatenated. | |
1245 | $b = '(.)\g{1}'; # Avoids the problems. | |
1246 | "aa" =~ /${a}/; # True | |
1247 | "aa" =~ /${b}/; # True | |
1248 | "aa0" =~ /${a}0/; # False! | |
1249 | "aa0" =~ /${b}0/; # True | |
dc0d9c48 KW |
1250 | "aa\x08" =~ /${a}0/; # True! |
1251 | "aa\x08" =~ /${b}0/; # False | |
9d860678 | 1252 | |
14218588 GS |
1253 | Several special variables also refer back to portions of the previous |
1254 | match. C<$+> returns whatever the last bracket match matched. | |
1255 | C<$&> returns the entire matched string. (At one point C<$0> did | |
1256 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
1257 | everything before the matched string. C<$'> returns everything |
1258 | after the matched string. And C<$^N> contains whatever was matched by | |
1259 | the most-recently closed group (submatch). C<$^N> can be used in | |
1260 | extended patterns (see below), for example to assign a submatch to a | |
81714fb9 | 1261 | variable. |
d74e8afc | 1262 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
14218588 | 1263 | |
d8b950dc | 1264 | These special variables, like the C<%+> hash and the numbered match variables |
57fbbe96 | 1265 | (C<$1>, C<$2>, C<$3>, I<etc>.) are dynamically scoped |
14218588 GS |
1266 | until the end of the enclosing block or until the next successful |
1267 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
d74e8afc ITB |
1268 | X<$+> X<$^N> X<$&> X<$`> X<$'> |
1269 | X<$1> X<$2> X<$3> X<$4> X<$5> X<$6> X<$7> X<$8> X<$9> | |
1270 | ||
0d017f4d | 1271 | B<NOTE>: Failed matches in Perl do not reset the match variables, |
5146ce24 | 1272 | which makes it easier to write code that tests for a series of more |
665e98b9 JH |
1273 | specific cases and remembers the best match. |
1274 | ||
13b0f67d DM |
1275 | B<WARNING>: If your code is to run on Perl 5.16 or earlier, |
1276 | beware that once Perl sees that you need one of C<$&>, C<$`>, or | |
14218588 | 1277 | C<$'> anywhere in the program, it has to provide them for every |
13b0f67d DM |
1278 | pattern match. This may substantially slow your program. |
1279 | ||
57fbbe96 | 1280 | Perl uses the same mechanism to produce C<$1>, C<$2>, I<etc>, so you also |
13b0f67d DM |
1281 | pay a price for each pattern that contains capturing parentheses. |
1282 | (To avoid this cost while retaining the grouping behaviour, use the | |
14218588 GS |
1283 | extended regular expression C<(?: ... )> instead.) But if you never |
1284 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
1285 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
1286 | if you can, but if you can't (and some algorithms really appreciate | |
1287 | them), once you've used them once, use them at will, because you've | |
13b0f67d | 1288 | already paid the price. |
d74e8afc | 1289 | X<$&> X<$`> X<$'> |
68dc0745 | 1290 | |
13b0f67d DM |
1291 | Perl 5.16 introduced a slightly more efficient mechanism that notes |
1292 | separately whether each of C<$`>, C<$&>, and C<$'> have been seen, and | |
1293 | thus may only need to copy part of the string. Perl 5.20 introduced a | |
1294 | much more efficient copy-on-write mechanism which eliminates any slowdown. | |
1295 | ||
1296 | As another workaround for this problem, Perl 5.10.0 introduced C<${^PREMATCH}>, | |
cde0cee5 YO |
1297 | C<${^MATCH}> and C<${^POSTMATCH}>, which are equivalent to C<$`>, C<$&> |
1298 | and C<$'>, B<except> that they are only guaranteed to be defined after a | |
87e95b7f | 1299 | successful match that was executed with the C</p> (preserve) modifier. |
cde0cee5 | 1300 | The use of these variables incurs no global performance penalty, unlike |
7711f978 | 1301 | their punctuation character equivalents, however at the trade-off that you |
13b0f67d DM |
1302 | have to tell perl when you want to use them. As of Perl 5.20, these three |
1303 | variables are equivalent to C<$`>, C<$&> and C<$'>, and C</p> is ignored. | |
87e95b7f | 1304 | X</p> X<p modifier> |
cde0cee5 | 1305 | |
9d727203 KW |
1306 | =head2 Quoting metacharacters |
1307 | ||
19799a22 GS |
1308 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
1309 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
1310 | are no backslashed symbols that aren't alphanumeric. So anything | |
7711f978 KW |
1311 | that looks like C<\\>, C<\(>, C<\)>, C<\[>, C<\]>, C<\{>, or C<\}> is |
1312 | always | |
19799a22 GS |
1313 | interpreted as a literal character, not a metacharacter. This was |
1314 | once used in a common idiom to disable or quote the special meanings | |
1315 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 1316 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
1317 | |
1318 | $pattern =~ s/(\W)/\\$1/g; | |
1319 | ||
f1cbbd6e | 1320 | (If C<use locale> is set, then this depends on the current locale.) |
7711f978 KW |
1321 | Today it is more common to use the C<L<quotemeta()|perlfunc/quotemeta>> |
1322 | function or the C<\Q> metaquoting escape sequence to disable all | |
1323 | metacharacters' special meanings like this: | |
a0d0e21e LW |
1324 | |
1325 | /$unquoted\Q$quoted\E$unquoted/ | |
1326 | ||
9da458fc IZ |
1327 | Beware that if you put literal backslashes (those not inside |
1328 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
1329 | backslash interpolation may lead to confusing results. If you | |
1330 | I<need> to use literal backslashes within C<\Q...\E>, | |
1331 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
1332 | ||
736fe711 KW |
1333 | C<quotemeta()> and C<\Q> are fully described in L<perlfunc/quotemeta>. |
1334 | ||
19799a22 GS |
1335 | =head2 Extended Patterns |
1336 | ||
14218588 | 1337 | Perl also defines a consistent extension syntax for features not |
0b928c2f FC |
1338 | found in standard tools like B<awk> and |
1339 | B<lex>. The syntax for most of these is a | |
14218588 GS |
1340 | pair of parentheses with a question mark as the first thing within |
1341 | the parentheses. The character after the question mark indicates | |
1342 | the extension. | |
19799a22 | 1343 | |
14218588 GS |
1344 | A question mark was chosen for this and for the minimal-matching |
1345 | construct because 1) question marks are rare in older regular | |
1346 | expressions, and 2) whenever you see one, you should stop and | |
0b928c2f | 1347 | "question" exactly what is going on. That's psychology.... |
a0d0e21e | 1348 | |
70ca8714 | 1349 | =over 4 |
a0d0e21e | 1350 | |
407fecf1 | 1351 | =item C<(?#I<text>)> |
d74e8afc | 1352 | X<(?#)> |
a0d0e21e | 1353 | |
407fecf1 | 1354 | A comment. The I<text> is ignored. |
7c688e65 | 1355 | Note that Perl closes |
7711f978 KW |
1356 | the comment as soon as it sees a C<")">, so there is no way to put a literal |
1357 | C<")"> in the comment. The pattern's closing delimiter must be escaped by | |
7c688e65 KW |
1358 | a backslash if it appears in the comment. |
1359 | ||
1360 | See L</E<sol>x> for another way to have comments in patterns. | |
a0d0e21e | 1361 | |
2ab07670 KW |
1362 | Note that a comment can go just about anywhere, except in the middle of |
1363 | an escape sequence. Examples: | |
1364 | ||
1365 | qr/foo(?#comment)bar/' # Matches 'foobar' | |
1366 | ||
1367 | # The pattern below matches 'abcd', 'abccd', or 'abcccd' | |
1368 | qr/abc(?#comment between literal and its quantifier){1,3}d/ | |
1369 | ||
1370 | # The pattern below generates a syntax error, because the '\p' must | |
1371 | # be followed immediately by a '{'. | |
1372 | qr/\p(?#comment between \p and its property name){Any}/ | |
1373 | ||
1374 | # The pattern below generates a syntax error, because the initial | |
1375 | # '\(' is a literal opening parenthesis, and so there is nothing | |
1376 | # for the closing ')' to match | |
1377 | qr/\(?#the backslash means this isn't a comment)p{Any}/ | |
1378 | ||
e4a1c03b KW |
1379 | # Comments can be used to fold long patterns into multiple lines |
1380 | qr/First part of a long regex(?# | |
1381 | )remaining part/ | |
1382 | ||
4cb6b395 | 1383 | =item C<(?adlupimnsx-imnsx)> |
fb85c044 | 1384 | |
4cb6b395 | 1385 | =item C<(?^alupimnsx)> |
fb85c044 | 1386 | X<(?)> X<(?^)> |
19799a22 | 1387 | |
d9a91485 | 1388 | Zero or more embedded pattern-match modifiers, to be turned on (or |
2ab07670 | 1389 | turned off if preceded by C<"-">) for the remainder of the pattern or |
fb85c044 KW |
1390 | the remainder of the enclosing pattern group (if any). |
1391 | ||
a95b7a20 AC |
1392 | This is particularly useful for dynamically-generated patterns, |
1393 | such as those read in from a | |
0d017f4d | 1394 | configuration file, taken from an argument, or specified in a table |
0b928c2f FC |
1395 | somewhere. Consider the case where some patterns want to be |
1396 | case-sensitive and some do not: The case-insensitive ones merely need to | |
0d017f4d | 1397 | include C<(?i)> at the front of the pattern. For example: |
19799a22 GS |
1398 | |
1399 | $pattern = "foobar"; | |
5d458dd8 | 1400 | if ( /$pattern/i ) { } |
19799a22 GS |
1401 | |
1402 | # more flexible: | |
1403 | ||
1404 | $pattern = "(?i)foobar"; | |
5d458dd8 | 1405 | if ( /$pattern/ ) { } |
19799a22 | 1406 | |
0b6d1084 | 1407 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 | 1408 | |
d8b950dc | 1409 | ( (?i) blah ) \s+ \g1 |
19799a22 | 1410 | |
0d017f4d WL |
1411 | will match C<blah> in any case, some spaces, and an exact (I<including the case>!) |
1412 | repetition of the previous word, assuming the C</x> modifier, and no C</i> | |
1413 | modifier outside this group. | |
19799a22 | 1414 | |
8eb5594e | 1415 | These modifiers do not carry over into named subpatterns called in the |
407fecf1 KW |
1416 | enclosing group. In other words, a pattern such as C<((?i)(?&I<NAME>))> does not |
1417 | change the case-sensitivity of the I<NAME> pattern. | |
8eb5594e | 1418 | |
2ab07670 KW |
1419 | A modifier is overridden by later occurrences of this construct in the |
1420 | same scope containing the same modifier, so that | |
1421 | ||
1422 | /((?im)foo(?-m)bar)/ | |
1423 | ||
1424 | matches all of C<foobar> case insensitively, but uses C</m> rules for | |
57fbbe96 KW |
1425 | only the C<foo> portion. The C<"a"> flag overrides C<aa> as well; |
1426 | likewise C<aa> overrides C<"a">. The same goes for C<"x"> and C<xx>. | |
77c8f263 KW |
1427 | Hence, in |
1428 | ||
1429 | /(?-x)foo/xx | |
1430 | ||
1431 | both C</x> and C</xx> are turned off during matching C<foo>. And in | |
1432 | ||
1433 | /(?x)foo/x | |
1434 | ||
1435 | C</x> but NOT C</xx> is turned on for matching C<foo>. (One might | |
1436 | mistakenly think that since the inner C<(?x)> is already in the scope of | |
1437 | C</x>, that the result would effectively be the sum of them, yielding | |
1438 | C</xx>. It doesn't work that way.) Similarly, doing something like | |
57fbbe96 KW |
1439 | C<(?xx-x)foo> turns off all C<"x"> behavior for matching C<foo>, it is not |
1440 | that you subtract 1 C<"x"> from 2 to get 1 C<"x"> remaining. | |
2ab07670 | 1441 | |
dc925305 KW |
1442 | Any of these modifiers can be set to apply globally to all regular |
1443 | expressions compiled within the scope of a C<use re>. See | |
a0bbd6ff | 1444 | L<re/"'/flags' mode">. |
dc925305 | 1445 | |
9de15fec | 1446 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
4cb6b395 | 1447 | after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Flags (except |
9de15fec KW |
1448 | C<"d">) may follow the caret to override it. |
1449 | But a minus sign is not legal with it. | |
1450 | ||
57fbbe96 KW |
1451 | Note that the C<"a">, C<"d">, C<"l">, C<"p">, and C<"u"> modifiers are special in |
1452 | that they can only be enabled, not disabled, and the C<"a">, C<"d">, C<"l">, and | |
1453 | C<"u"> modifiers are mutually exclusive: specifying one de-specifies the | |
1454 | others, and a maximum of one (or two C<"a">'s) may appear in the | |
ed7efc79 | 1455 | construct. Thus, for |
0b928c2f | 1456 | example, C<(?-p)> will warn when compiled under C<use warnings>; |
b6fa137b | 1457 | C<(?-d:...)> and C<(?dl:...)> are fatal errors. |
9de15fec | 1458 | |
57fbbe96 | 1459 | Note also that the C<"p"> modifier is special in that its presence |
9de15fec | 1460 | anywhere in a pattern has a global effect. |
cde0cee5 | 1461 | |
d9a91485 KW |
1462 | Having zero modifiers makes this a no-op (so why did you specify it, |
1463 | unless it's generated code), and starting in v5.30, warns under L<C<use | |
1464 | re 'strict'>|re/'strict' mode>. | |
1465 | ||
407fecf1 | 1466 | =item C<(?:I<pattern>)> |
d74e8afc | 1467 | X<(?:)> |
a0d0e21e | 1468 | |
407fecf1 | 1469 | =item C<(?adluimnsx-imnsx:I<pattern>)> |
ca9dfc88 | 1470 | |
407fecf1 | 1471 | =item C<(?^aluimnsx:I<pattern>)> |
fb85c044 KW |
1472 | X<(?^:)> |
1473 | ||
5a964f20 | 1474 | This is for clustering, not capturing; it groups subexpressions like |
7711f978 | 1475 | C<"()">, but doesn't make backreferences as C<"()"> does. So |
a0d0e21e | 1476 | |
5a964f20 | 1477 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e | 1478 | |
a95b7a20 | 1479 | matches the same field delimiters as |
a0d0e21e | 1480 | |
5a964f20 | 1481 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 1482 | |
a95b7a20 AC |
1483 | but doesn't spit out the delimiters themselves as extra fields (even though |
1484 | that's the behaviour of L<perlfunc/split> when its pattern contains capturing | |
1485 | groups). It's also cheaper not to capture | |
19799a22 | 1486 | characters if you don't need to. |
a0d0e21e | 1487 | |
7711f978 | 1488 | Any letters between C<"?"> and C<":"> act as flags modifiers as with |
4cb6b395 | 1489 | C<(?adluimnsx-imnsx)>. For example, |
ca9dfc88 IZ |
1490 | |
1491 | /(?s-i:more.*than).*million/i | |
1492 | ||
14218588 | 1493 | is equivalent to the more verbose |
ca9dfc88 IZ |
1494 | |
1495 | /(?:(?s-i)more.*than).*million/i | |
1496 | ||
7711f978 | 1497 | Note that any C<()> constructs enclosed within this one will still |
4cb6b395 KW |
1498 | capture unless the C</n> modifier is in effect. |
1499 | ||
57fbbe96 KW |
1500 | Like the L</(?adlupimnsx-imnsx)> construct, C<aa> and C<"a"> override each |
1501 | other, as do C<xx> and C<"x">. They are not additive. So, doing | |
1502 | something like C<(?xx-x:foo)> turns off all C<"x"> behavior for matching | |
77c8f263 | 1503 | C<foo>. |
2ab07670 | 1504 | |
fb85c044 | 1505 | Starting in Perl 5.14, a C<"^"> (caret or circumflex accent) immediately |
4cb6b395 | 1506 | after the C<"?"> is a shorthand equivalent to C<d-imnsx>. Any positive |
9de15fec | 1507 | flags (except C<"d">) may follow the caret, so |
fb85c044 KW |
1508 | |
1509 | (?^x:foo) | |
1510 | ||
1511 | is equivalent to | |
1512 | ||
4cb6b395 | 1513 | (?x-imns:foo) |
fb85c044 KW |
1514 | |
1515 | The caret tells Perl that this cluster doesn't inherit the flags of any | |
4cb6b395 | 1516 | surrounding pattern, but uses the system defaults (C<d-imnsx>), |
fb85c044 KW |
1517 | modified by any flags specified. |
1518 | ||
1519 | The caret allows for simpler stringification of compiled regular | |
1520 | expressions. These look like | |
1521 | ||
1522 | (?^:pattern) | |
1523 | ||
1524 | with any non-default flags appearing between the caret and the colon. | |
1525 | A test that looks at such stringification thus doesn't need to have the | |
1526 | system default flags hard-coded in it, just the caret. If new flags are | |
1527 | added to Perl, the meaning of the caret's expansion will change to include | |
1528 | the default for those flags, so the test will still work, unchanged. | |
1529 | ||
1530 | Specifying a negative flag after the caret is an error, as the flag is | |
1531 | redundant. | |
1532 | ||
1533 | Mnemonic for C<(?^...)>: A fresh beginning since the usual use of a caret is | |
1534 | to match at the beginning. | |
1535 | ||
407fecf1 | 1536 | =item C<(?|I<pattern>)> |
594d7033 YO |
1537 | X<(?|)> X<Branch reset> |
1538 | ||
1539 | This is the "branch reset" pattern, which has the special property | |
c27a5cfe | 1540 | that the capture groups are numbered from the same starting point |
99d59c4d | 1541 | in each alternation branch. It is available starting from perl 5.10.0. |
4deaaa80 | 1542 | |
c27a5cfe | 1543 | Capture groups are numbered from left to right, but inside this |
693596a8 | 1544 | construct the numbering is restarted for each branch. |
4deaaa80 | 1545 | |
c27a5cfe | 1546 | The numbering within each branch will be as normal, and any groups |
4deaaa80 PJ |
1547 | following this construct will be numbered as though the construct |
1548 | contained only one branch, that being the one with the most capture | |
c27a5cfe | 1549 | groups in it. |
4deaaa80 | 1550 | |
0b928c2f | 1551 | This construct is useful when you want to capture one of a |
4deaaa80 PJ |
1552 | number of alternative matches. |
1553 | ||
1554 | Consider the following pattern. The numbers underneath show in | |
c27a5cfe | 1555 | which group the captured content will be stored. |
594d7033 YO |
1556 | |
1557 | ||
57fbbe96 | 1558 | # before ---------------branch-reset----------- after |
594d7033 | 1559 | / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x |
57fbbe96 | 1560 | # 1 2 2 3 2 3 4 |
594d7033 | 1561 | |
57fbbe96 KW |
1562 | Be careful when using the branch reset pattern in combination with |
1563 | named captures. Named captures are implemented as being aliases to | |
c27a5cfe | 1564 | numbered groups holding the captures, and that interferes with the |
ab106183 A |
1565 | implementation of the branch reset pattern. If you are using named |
1566 | captures in a branch reset pattern, it's best to use the same names, | |
1567 | in the same order, in each of the alternations: | |
1568 | ||
1569 | /(?| (?<a> x ) (?<b> y ) | |
1570 | | (?<a> z ) (?<b> w )) /x | |
1571 | ||
1572 | Not doing so may lead to surprises: | |
1573 | ||
1574 | "12" =~ /(?| (?<a> \d+ ) | (?<b> \D+))/x; | |
a95b7a20 AC |
1575 | say $+{a}; # Prints '12' |
1576 | say $+{b}; # *Also* prints '12'. | |
ab106183 | 1577 | |
c27a5cfe KW |
1578 | The problem here is that both the group named C<< a >> and the group |
1579 | named C<< b >> are aliases for the group belonging to C<< $1 >>. | |
90a18110 | 1580 | |
f67a5002 | 1581 | =item Lookaround Assertions |
ee9b8eae YO |
1582 | X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround> |
1583 | ||
f67a5002 | 1584 | Lookaround assertions are zero-width patterns which match a specific |
ee9b8eae YO |
1585 | pattern without including it in C<$&>. Positive assertions match when |
1586 | their subpattern matches, negative assertions match when their subpattern | |
f67a5002 EA |
1587 | fails. Lookbehind matches text up to the current match position, |
1588 | lookahead matches text following the current match position. | |
ee9b8eae YO |
1589 | |
1590 | =over 4 | |
1591 | ||
407fecf1 | 1592 | =item C<(?=I<pattern>)> |
e7206367 | 1593 | |
407fecf1 | 1594 | =item C<(*pla:I<pattern>)> |
e7206367 | 1595 | |
407fecf1 | 1596 | =item C<(*positive_lookahead:I<pattern>)> |
e7206367 KW |
1597 | X<(?=)> |
1598 | X<(*pla> | |
1599 | X<(*positive_lookahead> | |
1600 | X<look-ahead, positive> X<lookahead, positive> | |
a0d0e21e | 1601 | |
f67a5002 | 1602 | A zero-width positive lookahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
1603 | matches a word followed by a tab, without including the tab in C<$&>. |
1604 | ||
407fecf1 | 1605 | =item C<(?!I<pattern>)> |
e7206367 | 1606 | |
407fecf1 | 1607 | =item C<(*nla:I<pattern>)> |
e7206367 | 1608 | |
407fecf1 | 1609 | =item C<(*negative_lookahead:I<pattern>)> |
e7206367 KW |
1610 | X<(?!)> |
1611 | X<(*nla> | |
1612 | X<(*negative_lookahead> | |
1613 | X<look-ahead, negative> X<lookahead, negative> | |
a0d0e21e | 1614 | |
f67a5002 | 1615 | A zero-width negative lookahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 1616 | matches any occurrence of "foo" that isn't followed by "bar". Note |
f67a5002 EA |
1617 | however that lookahead and lookbehind are NOT the same thing. You cannot |
1618 | use this for lookbehind. | |
7b8d334a | 1619 | |
5a964f20 | 1620 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
1621 | will not do what you want. That's because the C<(?!foo)> is just saying that |
1622 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
f67a5002 | 1623 | match. Use lookbehind instead (see below). |
c277df42 | 1624 | |
407fecf1 | 1625 | =item C<(?<=I<pattern>)> |
a8f2f5fa AC |
1626 | |
1627 | =item C<\K> | |
e7206367 | 1628 | |
407fecf1 | 1629 | =item C<(*plb:I<pattern>)> |
e7206367 | 1630 | |
407fecf1 | 1631 | =item C<(*positive_lookbehind:I<pattern>)> |
e7206367 KW |
1632 | X<(?<=)> |
1633 | X<(*plb> | |
1634 | X<(*positive_lookbehind> | |
1635 | X<look-behind, positive> X<lookbehind, positive> X<\K> | |
c277df42 | 1636 | |
f67a5002 | 1637 | A zero-width positive lookbehind assertion. For example, C</(?<=\t)\w+/> |
19799a22 | 1638 | matches a word that follows a tab, without including the tab in C<$&>. |
c277df42 | 1639 | |
2abbd513 KW |
1640 | Prior to Perl 5.30, it worked only for fixed-width lookbehind, but |
1641 | starting in that release, it can handle variable lengths from 1 to 255 | |
1642 | characters as an experimental feature. The feature is enabled | |
1643 | automatically if you use a variable length lookbehind assertion, but | |
1644 | will raise a warning at pattern compilation time, unless turned off, in | |
1645 | the C<experimental::vlb> category. This is to warn you that the exact | |
1646 | behavior is subject to change should feedback from actual use in the | |
1647 | field indicate to do so; or even complete removal if the problems found | |
1648 | are not practically surmountable. You can achieve close to pre-5.30 | |
1649 | behavior by fatalizing warnings in this category. | |
1650 | ||
1651 | There is a special form of this construct, called C<\K> | |
fa690ddc | 1652 | (available since Perl 5.10.0), which causes the |
ee9b8eae | 1653 | regex engine to "keep" everything it had matched prior to the C<\K> and |
2abbd513 KW |
1654 | not include it in C<$&>. This effectively provides non-experimental |
1655 | variable-length lookbehind of any length. | |
1656 | ||
1657 | And, there is a technique that can be used to handle variable length | |
1658 | lookbehinds on earlier releases, and longer than 255 characters. It is | |
1659 | described in | |
1660 | L<http://www.drregex.com/2019/02/variable-length-lookbehinds-actually.html>. | |
1661 | ||
1662 | Note that under C</i>, a few single characters match two or three other | |
1663 | characters. This makes them variable length, and the 255 length applies | |
1664 | to the maximum number of characters in the match. For | |
1665 | example C<qr/\N{LATIN SMALL LETTER SHARP S}/i> matches the sequence | |
1666 | C<"ss">. Your lookbehind assertion could contain 127 Sharp S | |
1667 | characters under C</i>, but adding a 128th would generate a compilation | |
1668 | error, as that could match 256 C<"s"> characters in a row. | |
fa690ddc KW |
1669 | |
1670 | The use of C<\K> inside of another lookaround assertion | |
ee9b8eae YO |
1671 | is allowed, but the behaviour is currently not well defined. |
1672 | ||
c62285ac | 1673 | For various reasons C<\K> may be significantly more efficient than the |
ee9b8eae YO |
1674 | equivalent C<< (?<=...) >> construct, and it is especially useful in |
1675 | situations where you want to efficiently remove something following | |
1676 | something else in a string. For instance | |
1677 | ||
1678 | s/(foo)bar/$1/g; | |
1679 | ||
1680 | can be rewritten as the much more efficient | |
1681 | ||
1682 | s/foo\Kbar//g; | |
1683 | ||
2abbd513 KW |
1684 | Use of the non-greedy modifier C<"?"> may not give you the expected |
1685 | results if it is within a capturing group within the construct. | |
1686 | ||
407fecf1 | 1687 | =item C<(?<!I<pattern>)> |
e7206367 | 1688 | |
407fecf1 | 1689 | =item C<(*nlb:I<pattern>)> |
e7206367 | 1690 | |
407fecf1 | 1691 | =item C<(*negative_lookbehind:I<pattern>)> |
e7206367 KW |
1692 | X<(?<!)> |
1693 | X<(*nlb> | |
1694 | X<(*negative_lookbehind> | |
1695 | X<look-behind, negative> X<lookbehind, negative> | |
c277df42 | 1696 | |
f67a5002 | 1697 | A zero-width negative lookbehind assertion. For example C</(?<!bar)foo/> |
2abbd513 KW |
1698 | matches any occurrence of "foo" that does not follow "bar". |
1699 | ||
1700 | Prior to Perl 5.30, it worked only for fixed-width lookbehind, but | |
1701 | starting in that release, it can handle variable lengths from 1 to 255 | |
1702 | characters as an experimental feature. The feature is enabled | |
1703 | automatically if you use a variable length lookbehind assertion, but | |
1704 | will raise a warning at pattern compilation time, unless turned off, in | |
1705 | the C<experimental::vlb> category. This is to warn you that the exact | |
1706 | behavior is subject to change should feedback from actual use in the | |
1707 | field indicate to do so; or even complete removal if the problems found | |
1708 | are not practically surmountable. You can achieve close to pre-5.30 | |
1709 | behavior by fatalizing warnings in this category. | |
1710 | ||
1711 | There is a technique that can be used to handle variable length | |
1712 | lookbehinds on earlier releases, and longer than 255 characters. It is | |
1713 | described in | |
fa690ddc | 1714 | L<http://www.drregex.com/2019/02/variable-length-lookbehinds-actually.html>. |
c277df42 | 1715 | |
2abbd513 KW |
1716 | Note that under C</i>, a few single characters match two or three other |
1717 | characters. This makes them variable length, and the 255 length applies | |
1718 | to the maximum number of characters in the match. For | |
1719 | example C<qr/\N{LATIN SMALL LETTER SHARP S}/i> matches the sequence | |
1720 | C<"ss">. Your lookbehind assertion could contain 127 Sharp S | |
1721 | characters under C</i>, but adding a 128th would generate a compilation | |
1722 | error, as that could match 256 C<"s"> characters in a row. | |
1723 | ||
1724 | Use of the non-greedy modifier C<"?"> may not give you the expected | |
1725 | results if it is within a capturing group within the construct. | |
1726 | ||
ee9b8eae YO |
1727 | =back |
1728 | ||
407fecf1 | 1729 | =item C<< (?<I<NAME>>I<pattern>) >> |
a8f2f5fa | 1730 | |
407fecf1 | 1731 | =item C<(?'I<NAME>'I<pattern>)> |
81714fb9 YO |
1732 | X<< (?<NAME>) >> X<(?'NAME')> X<named capture> X<capture> |
1733 | ||
c27a5cfe | 1734 | A named capture group. Identical in every respect to normal capturing |
0b928c2f FC |
1735 | parentheses C<()> but for the additional fact that the group |
1736 | can be referred to by name in various regular expression | |
407fecf1 | 1737 | constructs (like C<\g{I<NAME>}>) and can be accessed by name |
0b928c2f | 1738 | after a successful match via C<%+> or C<%->. See L<perlvar> |
90a18110 | 1739 | for more details on the C<%+> and C<%-> hashes. |
81714fb9 | 1740 | |
33727e0f | 1741 | If multiple distinct capture groups have the same name, then |
407fecf1 | 1742 | C<$+{I<NAME>}> will refer to the leftmost defined group in the match. |
81714fb9 | 1743 | |
407fecf1 KW |
1744 | The forms C<(?'I<NAME>'I<pattern>)> and C<< (?<I<NAME>>I<pattern>) >> |
1745 | are equivalent. | |
81714fb9 YO |
1746 | |
1747 | B<NOTE:> While the notation of this construct is the same as the similar | |
c27a5cfe | 1748 | function in .NET regexes, the behavior is not. In Perl the groups are |
81714fb9 YO |
1749 | numbered sequentially regardless of being named or not. Thus in the |
1750 | pattern | |
1751 | ||
1752 | /(x)(?<foo>y)(z)/ | |
1753 | ||
407fecf1 | 1754 | C<$+{foo}> will be the same as C<$2>, and C<$3> will contain 'z' instead of |
81714fb9 YO |
1755 | the opposite which is what a .NET regex hacker might expect. |
1756 | ||
7711f978 | 1757 | Currently I<NAME> is restricted to simple identifiers only. |
1f1031fe YO |
1758 | In other words, it must match C</^[_A-Za-z][_A-Za-z0-9]*\z/> or |
1759 | its Unicode extension (see L<utf8>), | |
1760 | though it isn't extended by the locale (see L<perllocale>). | |
81714fb9 | 1761 | |
1f1031fe | 1762 | B<NOTE:> In order to make things easier for programmers with experience |
407fecf1 KW |
1763 | with the Python or PCRE regex engines, the pattern C<< |
1764 | (?PE<lt>I<NAME>E<gt>I<pattern>) >> | |
1765 | may be used instead of C<< (?<I<NAME>>I<pattern>) >>; however this form does not | |
64c5a566 | 1766 | support the use of single quotes as a delimiter for the name. |
81714fb9 | 1767 | |
407fecf1 | 1768 | =item C<< \k<I<NAME>> >> |
1f1031fe | 1769 | |
407fecf1 | 1770 | =item C<< \k'I<NAME>' >> |
81714fb9 | 1771 | |
a44b2be7 KW |
1772 | =item C<< \k{I<NAME>} >> |
1773 | ||
81714fb9 YO |
1774 | Named backreference. Similar to numeric backreferences, except that |
1775 | the group is designated by name and not number. If multiple groups | |
1776 | have the same name then it refers to the leftmost defined group in | |
1777 | the current match. | |
1778 | ||
407fecf1 | 1779 | It is an error to refer to a name not defined by a C<< (?<I<NAME>>) >> |
81714fb9 YO |
1780 | earlier in the pattern. |
1781 | ||
1b2f32d5 KW |
1782 | All three forms are equivalent, although with C<< \k{ I<NAME> } >>, |
1783 | you may optionally have blanks within but adjacent to the braces, as | |
1784 | shown. | |
81714fb9 | 1785 | |
1f1031fe | 1786 | B<NOTE:> In order to make things easier for programmers with experience |
407fecf1 KW |
1787 | with the Python or PCRE regex engines, the pattern C<< (?P=I<NAME>) >> |
1788 | may be used instead of C<< \k<I<NAME>> >>. | |
1f1031fe | 1789 | |
407fecf1 | 1790 | =item C<(?{ I<code> })> |
d74e8afc | 1791 | X<(?{})> X<regex, code in> X<regexp, code in> X<regular expression, code in> |
c277df42 | 1792 | |
83f32aba RS |
1793 | B<WARNING>: Using this feature safely requires that you understand its |
1794 | limitations. Code executed that has side effects may not perform identically | |
1795 | from version to version due to the effect of future optimisations in the regex | |
1796 | engine. For more information on this, see L</Embedded Code Execution | |
1797 | Frequency>. | |
c277df42 | 1798 | |
e128ab2c DM |
1799 | This zero-width assertion executes any embedded Perl code. It always |
1800 | succeeds, and its return value is set as C<$^R>. | |
19799a22 | 1801 | |
e128ab2c DM |
1802 | In literal patterns, the code is parsed at the same time as the |
1803 | surrounding code. While within the pattern, control is passed temporarily | |
1804 | back to the perl parser, until the logically-balancing closing brace is | |
1805 | encountered. This is similar to the way that an array index expression in | |
1806 | a literal string is handled, for example | |
77ea4f6d | 1807 | |
e128ab2c DM |
1808 | "abc$array[ 1 + f('[') + g()]def" |
1809 | ||
1810 | In particular, braces do not need to be balanced: | |
1811 | ||
576fa024 | 1812 | s/abc(?{ f('{'); })/def/ |
e128ab2c DM |
1813 | |
1814 | Even in a pattern that is interpolated and compiled at run-time, literal | |
1815 | code blocks will be compiled once, at perl compile time; the following | |
1816 | prints "ABCD": | |
1817 | ||
1818 | print "D"; | |
1819 | my $qr = qr/(?{ BEGIN { print "A" } })/; | |
1820 | my $foo = "foo"; | |
1821 | /$foo$qr(?{ BEGIN { print "B" } })/; | |
1822 | BEGIN { print "C" } | |
1823 | ||
1824 | In patterns where the text of the code is derived from run-time | |
1825 | information rather than appearing literally in a source code /pattern/, | |
1826 | the code is compiled at the same time that the pattern is compiled, and | |
5771dda0 | 1827 | for reasons of security, C<use re 'eval'> must be in scope. This is to |
e128ab2c DM |
1828 | stop user-supplied patterns containing code snippets from being |
1829 | executable. | |
1830 | ||
5771dda0 | 1831 | In situations where you need to enable this with C<use re 'eval'>, you should |
e128ab2c DM |
1832 | also have taint checking enabled. Better yet, use the carefully |
1833 | constrained evaluation within a Safe compartment. See L<perlsec> for | |
1834 | details about both these mechanisms. | |
1835 | ||
1836 | From the viewpoint of parsing, lexical variable scope and closures, | |
1837 | ||
1838 | /AAA(?{ BBB })CCC/ | |
1839 | ||
1840 | behaves approximately like | |
1841 | ||
1842 | /AAA/ && do { BBB } && /CCC/ | |
1843 | ||
1844 | Similarly, | |
1845 | ||
1846 | qr/AAA(?{ BBB })CCC/ | |
1847 | ||
1848 | behaves approximately like | |
77ea4f6d | 1849 | |
e128ab2c DM |
1850 | sub { /AAA/ && do { BBB } && /CCC/ } |
1851 | ||
1852 | In particular: | |
1853 | ||
1854 | { my $i = 1; $r = qr/(?{ print $i })/ } | |
1855 | my $i = 2; | |
1856 | /$r/; # prints "1" | |
1857 | ||
1858 | Inside a C<(?{...})> block, C<$_> refers to the string the regular | |
754091cb | 1859 | expression is matching against. You can also use C<pos()> to know what is |
fa11829f | 1860 | the current position of matching within this string. |
754091cb | 1861 | |
e128ab2c DM |
1862 | The code block introduces a new scope from the perspective of lexical |
1863 | variable declarations, but B<not> from the perspective of C<local> and | |
1864 | similar localizing behaviours. So later code blocks within the same | |
1865 | pattern will still see the values which were localized in earlier blocks. | |
1866 | These accumulated localizations are undone either at the end of a | |
1867 | successful match, or if the assertion is backtracked (compare | |
5a0de581 | 1868 | L</"Backtracking">). For example, |
b9ac3b5b GS |
1869 | |
1870 | $_ = 'a' x 8; | |
5d458dd8 | 1871 | m< |
d1fbf752 | 1872 | (?{ $cnt = 0 }) # Initialize $cnt. |
b9ac3b5b | 1873 | ( |
5d458dd8 | 1874 | a |
b9ac3b5b | 1875 | (?{ |
d1fbf752 KW |
1876 | local $cnt = $cnt + 1; # Update $cnt, |
1877 | # backtracking-safe. | |
b9ac3b5b | 1878 | }) |
5d458dd8 | 1879 | )* |
b9ac3b5b | 1880 | aaaa |
d1fbf752 KW |
1881 | (?{ $res = $cnt }) # On success copy to |
1882 | # non-localized location. | |
b9ac3b5b GS |
1883 | >x; |
1884 | ||
e128ab2c DM |
1885 | will initially increment C<$cnt> up to 8; then during backtracking, its |
1886 | value will be unwound back to 4, which is the value assigned to C<$res>. | |
7711f978 | 1887 | At the end of the regex execution, C<$cnt> will be wound back to its initial |
e128ab2c DM |
1888 | value of 0. |
1889 | ||
1890 | This assertion may be used as the condition in a | |
b9ac3b5b | 1891 | |
e128ab2c DM |
1892 | (?(condition)yes-pattern|no-pattern) |
1893 | ||
407fecf1 | 1894 | switch. If I<not> used in this way, the result of evaluation of I<code> |
e128ab2c | 1895 | is put into the special variable C<$^R>. This happens immediately, so |
407fecf1 | 1896 | C<$^R> can be used from other C<(?{ I<code> })> assertions inside the same |
e128ab2c | 1897 | regular expression. |
b9ac3b5b | 1898 | |
19799a22 GS |
1899 | The assignment to C<$^R> above is properly localized, so the old |
1900 | value of C<$^R> is restored if the assertion is backtracked; compare | |
5a0de581 | 1901 | L</"Backtracking">. |
b9ac3b5b | 1902 | |
e128ab2c DM |
1903 | Note that the special variable C<$^N> is particularly useful with code |
1904 | blocks to capture the results of submatches in variables without having to | |
1905 | keep track of the number of nested parentheses. For example: | |
1906 | ||
1907 | $_ = "The brown fox jumps over the lazy dog"; | |
1908 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
1909 | print "color = $color, animal = $animal\n"; | |
1910 | ||
8988a1bb | 1911 | |
407fecf1 | 1912 | =item C<(??{ I<code> })> |
d74e8afc ITB |
1913 | X<(??{})> |
1914 | X<regex, postponed> X<regexp, postponed> X<regular expression, postponed> | |
0f5d15d6 | 1915 | |
83f32aba RS |
1916 | B<WARNING>: Using this feature safely requires that you understand its |
1917 | limitations. Code executed that has side effects may not perform | |
1918 | identically from version to version due to the effect of future | |
1919 | optimisations in the regex engine. For more information on this, see | |
1920 | L</Embedded Code Execution Frequency>. | |
0f5d15d6 | 1921 | |
e128ab2c | 1922 | This is a "postponed" regular subexpression. It behaves in I<exactly> the |
407fecf1 | 1923 | same way as a C<(?{ I<code> })> code block as described above, except that |
e128ab2c DM |
1924 | its return value, rather than being assigned to C<$^R>, is treated as a |
1925 | pattern, compiled if it's a string (or used as-is if its a qr// object), | |
1926 | then matched as if it were inserted instead of this construct. | |
6bda09f9 | 1927 | |
e128ab2c DM |
1928 | During the matching of this sub-pattern, it has its own set of |
1929 | captures which are valid during the sub-match, but are discarded once | |
1930 | control returns to the main pattern. For example, the following matches, | |
1931 | with the inner pattern capturing "B" and matching "BB", while the outer | |
1932 | pattern captures "A"; | |
1933 | ||
1934 | my $inner = '(.)\1'; | |
1935 | "ABBA" =~ /^(.)(??{ $inner })\1/; | |
1936 | print $1; # prints "A"; | |
6bda09f9 | 1937 | |
e128ab2c DM |
1938 | Note that this means that there is no way for the inner pattern to refer |
1939 | to a capture group defined outside. (The code block itself can use C<$1>, | |
57fbbe96 | 1940 | I<etc>., to refer to the enclosing pattern's capture groups.) Thus, although |
0f5d15d6 | 1941 | |
e128ab2c DM |
1942 | ('a' x 100)=~/(??{'(.)' x 100})/ |
1943 | ||
7711f978 | 1944 | I<will> match, it will I<not> set C<$1> on exit. |
19799a22 GS |
1945 | |
1946 | The following pattern matches a parenthesized group: | |
0f5d15d6 | 1947 | |
d1fbf752 KW |
1948 | $re = qr{ |
1949 | \( | |
1950 | (?: | |
1951 | (?> [^()]+ ) # Non-parens without backtracking | |
1952 | | | |
1953 | (??{ $re }) # Group with matching parens | |
1954 | )* | |
1955 | \) | |
1956 | }x; | |
0f5d15d6 | 1957 | |
93f313ef | 1958 | See also |
407fecf1 | 1959 | L<C<(?I<PARNO>)>|/(?I<PARNO>) (?-I<PARNO>) (?+I<PARNO>) (?R) (?0)> |
93f313ef | 1960 | for a different, more efficient way to accomplish |
6bda09f9 YO |
1961 | the same task. |
1962 | ||
7e7dae31 KW |
1963 | Executing a postponed regular expression too many times without |
1964 | consuming any input string will also result in a fatal error. The depth | |
1965 | at which that happens is compiled into perl, so it can be changed with a | |
1966 | custom build. | |
6bda09f9 | 1967 | |
93f313ef | 1968 | =item C<(?I<PARNO>)> C<(?-I<PARNO>)> C<(?+I<PARNO>)> C<(?R)> C<(?0)> |
542fa716 | 1969 | X<(?PARNO)> X<(?1)> X<(?R)> X<(?0)> X<(?-1)> X<(?+1)> X<(?-PARNO)> X<(?+PARNO)> |
6bda09f9 | 1970 | X<regex, recursive> X<regexp, recursive> X<regular expression, recursive> |
d1b2014a YO |
1971 | X<regex, relative recursion> X<GOSUB> X<GOSTART> |
1972 | ||
1973 | Recursive subpattern. Treat the contents of a given capture buffer in the | |
1974 | current pattern as an independent subpattern and attempt to match it at | |
1975 | the current position in the string. Information about capture state from | |
1976 | the caller for things like backreferences is available to the subpattern, | |
1977 | but capture buffers set by the subpattern are not visible to the caller. | |
6bda09f9 | 1978 | |
407fecf1 | 1979 | Similar to C<(??{ I<code> })> except that it does not involve executing any |
e128ab2c DM |
1980 | code or potentially compiling a returned pattern string; instead it treats |
1981 | the part of the current pattern contained within a specified capture group | |
d1b2014a | 1982 | as an independent pattern that must match at the current position. Also |
407fecf1 | 1983 | different is the treatment of capture buffers, unlike C<(??{ I<code> })> |
dd407255 | 1984 | recursive patterns have access to their caller's match state, so one can |
d1b2014a | 1985 | use backreferences safely. |
6bda09f9 | 1986 | |
93f313ef | 1987 | I<PARNO> is a sequence of digits (not starting with 0) whose value reflects |
c27a5cfe | 1988 | the paren-number of the capture group to recurse to. C<(?R)> recurses to |
894be9b7 | 1989 | the beginning of the whole pattern. C<(?0)> is an alternate syntax for |
93f313ef | 1990 | C<(?R)>. If I<PARNO> is preceded by a plus or minus sign then it is assumed |
c27a5cfe | 1991 | to be relative, with negative numbers indicating preceding capture groups |
542fa716 | 1992 | and positive ones following. Thus C<(?-1)> refers to the most recently |
c27a5cfe | 1993 | declared group, and C<(?+1)> indicates the next group to be declared. |
c74340f9 | 1994 | Note that the counting for relative recursion differs from that of |
c27a5cfe | 1995 | relative backreferences, in that with recursion unclosed groups B<are> |
c74340f9 | 1996 | included. |
6bda09f9 | 1997 | |
7711f978 | 1998 | The following pattern matches a function C<foo()> which may contain |
f145b7e9 | 1999 | balanced parentheses as the argument. |
6bda09f9 | 2000 | |
d1fbf752 | 2001 | $re = qr{ ( # paren group 1 (full function) |
81714fb9 | 2002 | foo |
d1fbf752 | 2003 | ( # paren group 2 (parens) |
6bda09f9 | 2004 | \( |
d1fbf752 | 2005 | ( # paren group 3 (contents of parens) |
6bda09f9 | 2006 | (?: |
d1fbf752 | 2007 | (?> [^()]+ ) # Non-parens without backtracking |
6bda09f9 | 2008 | | |
d1fbf752 | 2009 | (?2) # Recurse to start of paren group 2 |
6bda09f9 YO |
2010 | )* |
2011 | ) | |
2012 | \) | |
2013 | ) | |
2014 | ) | |
2015 | }x; | |
2016 | ||
2017 | If the pattern was used as follows | |
2018 | ||
2019 | 'foo(bar(baz)+baz(bop))'=~/$re/ | |
2020 | and print "\$1 = $1\n", | |
2021 | "\$2 = $2\n", | |
2022 | "\$3 = $3\n"; | |
2023 | ||
2024 | the output produced should be the following: | |
2025 | ||
2026 | $1 = foo(bar(baz)+baz(bop)) | |
2027 | $2 = (bar(baz)+baz(bop)) | |
81714fb9 | 2028 | $3 = bar(baz)+baz(bop) |
6bda09f9 | 2029 | |
c27a5cfe | 2030 | If there is no corresponding capture group defined, then it is a |
7e7dae31 KW |
2031 | fatal error. Recursing deeply without consuming any input string will |
2032 | also result in a fatal error. The depth at which that happens is | |
2033 | compiled into perl, so it can be changed with a custom build. | |
6bda09f9 | 2034 | |
542fa716 YO |
2035 | The following shows how using negative indexing can make it |
2036 | easier to embed recursive patterns inside of a C<qr//> construct | |
2037 | for later use: | |
2038 | ||
2039 | my $parens = qr/(\((?:[^()]++|(?-1))*+\))/; | |
c77257ed | 2040 | if (/foo $parens \s+ \+ \s+ bar $parens/x) { |
542fa716 YO |
2041 | # do something here... |
2042 | } | |
2043 | ||
81714fb9 | 2044 | B<Note> that this pattern does not behave the same way as the equivalent |
0d017f4d | 2045 | PCRE or Python construct of the same form. In Perl you can backtrack into |
6bda09f9 | 2046 | a recursed group, in PCRE and Python the recursed into group is treated |
542fa716 | 2047 | as atomic. Also, modifiers are resolved at compile time, so constructs |
7711f978 | 2048 | like C<(?i:(?1))> or C<(?:(?i)(?1))> do not affect how the sub-pattern will |
542fa716 | 2049 | be processed. |
6bda09f9 | 2050 | |
407fecf1 | 2051 | =item C<(?&I<NAME>)> |
894be9b7 YO |
2052 | X<(?&NAME)> |
2053 | ||
93f313ef | 2054 | Recurse to a named subpattern. Identical to C<(?I<PARNO>)> except that the |
0d017f4d | 2055 | parenthesis to recurse to is determined by name. If multiple parentheses have |
894be9b7 YO |
2056 | the same name, then it recurses to the leftmost. |
2057 | ||
2058 | It is an error to refer to a name that is not declared somewhere in the | |
2059 | pattern. | |
2060 | ||
1f1031fe | 2061 | B<NOTE:> In order to make things easier for programmers with experience |
407fecf1 KW |
2062 | with the Python or PCRE regex engines the pattern C<< (?P>I<NAME>) >> |
2063 | may be used instead of C<< (?&I<NAME>) >>. | |
1f1031fe | 2064 | |
407fecf1 | 2065 | =item C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)> |
e2e6a0f1 | 2066 | X<(?()> |
286f584a | 2067 | |
407fecf1 | 2068 | =item C<(?(I<condition>)I<yes-pattern>)> |
286f584a | 2069 | |
407fecf1 KW |
2070 | Conditional expression. Matches I<yes-pattern> if I<condition> yields |
2071 | a true value, matches I<no-pattern> otherwise. A missing pattern always | |
41ef34de ML |
2072 | matches. |
2073 | ||
407fecf1 | 2074 | C<(I<condition>)> should be one of: |
7711f978 KW |
2075 | |
2076 | =over 4 | |
2077 | ||
2078 | =item an integer in parentheses | |
2079 | ||
2080 | (which is valid if the corresponding pair of parentheses | |
2081 | matched); | |
2082 | ||
f67a5002 | 2083 | =item a lookahead/lookbehind/evaluate zero-width assertion; |
7711f978 KW |
2084 | |
2085 | =item a name in angle brackets or single quotes | |
2086 | ||
2087 | (which is valid if a group with the given name matched); | |
2088 | ||
2089 | =item the special symbol C<(R)> | |
2090 | ||
2091 | (true when evaluated inside of recursion or eval). Additionally the | |
57fbbe96 | 2092 | C<"R"> may be |
e2e6a0f1 | 2093 | followed by a number, (which will be true when evaluated when recursing |
407fecf1 | 2094 | inside of the appropriate group), or by C<&I<NAME>>, in which case it will |
e2e6a0f1 YO |
2095 | be true only when evaluated during recursion in the named group. |
2096 | ||
7711f978 KW |
2097 | =back |
2098 | ||
e2e6a0f1 YO |
2099 | Here's a summary of the possible predicates: |
2100 | ||
2101 | =over 4 | |
2102 | ||
7711f978 | 2103 | =item C<(1)> C<(2)> ... |
e2e6a0f1 | 2104 | |
c27a5cfe | 2105 | Checks if the numbered capturing group has matched something. |
9240ab7d | 2106 | Full syntax: C<< (?(1)then|else) >> |
e2e6a0f1 | 2107 | |
7711f978 | 2108 | =item C<(E<lt>I<NAME>E<gt>)> C<('I<NAME>')> |
e2e6a0f1 | 2109 | |
c27a5cfe | 2110 | Checks if a group with the given name has matched something. |
9240ab7d | 2111 | Full syntax: C<< (?(<name>)then|else) >> |
e2e6a0f1 | 2112 | |
7711f978 | 2113 | =item C<(?=...)> C<(?!...)> C<(?<=...)> C<(?<!...)> |
f01cd190 | 2114 | |
7711f978 | 2115 | Checks whether the pattern matches (or does not match, for the C<"!"> |
f01cd190 | 2116 | variants). |
fdec910a | 2117 | Full syntax: C<< (?(?=I<lookahead>)I<then>|I<else>) >> |
f01cd190 | 2118 | |
7711f978 | 2119 | =item C<(?{ I<CODE> })> |
e2e6a0f1 | 2120 | |
f01cd190 | 2121 | Treats the return value of the code block as the condition. |
407fecf1 | 2122 | Full syntax: C<< (?(?{ I<code> })I<then>|I<else>) >> |
e2e6a0f1 | 2123 | |
7711f978 | 2124 | =item C<(R)> |
e2e6a0f1 YO |
2125 | |
2126 | Checks if the expression has been evaluated inside of recursion. | |
407fecf1 | 2127 | Full syntax: C<< (?(R)I<then>|I<else>) >> |
e2e6a0f1 | 2128 | |
7711f978 | 2129 | =item C<(R1)> C<(R2)> ... |
e2e6a0f1 YO |
2130 | |
2131 | Checks if the expression has been evaluated while executing directly | |
2132 | inside of the n-th capture group. This check is the regex equivalent of | |
2133 | ||
2134 | if ((caller(0))[3] eq 'subname') { ... } | |
2135 | ||
2136 | In other words, it does not check the full recursion stack. | |
2137 | ||
407fecf1 | 2138 | Full syntax: C<< (?(R1)I<then>|I<else>) >> |
9240ab7d | 2139 | |
7711f978 | 2140 | =item C<(R&I<NAME>)> |
e2e6a0f1 YO |
2141 | |
2142 | Similar to C<(R1)>, this predicate checks to see if we're executing | |
2143 | directly inside of the leftmost group with a given name (this is the same | |
7711f978 | 2144 | logic used by C<(?&I<NAME>)> to disambiguate). It does not check the full |
e2e6a0f1 | 2145 | stack, but only the name of the innermost active recursion. |
407fecf1 | 2146 | Full syntax: C<< (?(R&I<name>)I<then>|I<else>) >> |
e2e6a0f1 | 2147 | |
7711f978 | 2148 | =item C<(DEFINE)> |
e2e6a0f1 YO |
2149 | |
2150 | In this case, the yes-pattern is never directly executed, and no | |
2151 | no-pattern is allowed. Similar in spirit to C<(?{0})> but more efficient. | |
2152 | See below for details. | |
407fecf1 | 2153 | Full syntax: C<< (?(DEFINE)I<definitions>...) >> |
e2e6a0f1 YO |
2154 | |
2155 | =back | |
2156 | ||
2157 | For example: | |
2158 | ||
2159 | m{ ( \( )? | |
2160 | [^()]+ | |
2161 | (?(1) \) ) | |
2162 | }x | |
2163 | ||
2164 | matches a chunk of non-parentheses, possibly included in parentheses | |
2165 | themselves. | |
2166 | ||
0b928c2f FC |
2167 | A special form is the C<(DEFINE)> predicate, which never executes its |
2168 | yes-pattern directly, and does not allow a no-pattern. This allows one to | |
2169 | define subpatterns which will be executed only by the recursion mechanism. | |
e2e6a0f1 YO |
2170 | This way, you can define a set of regular expression rules that can be |
2171 | bundled into any pattern you choose. | |
2172 | ||
2173 | It is recommended that for this usage you put the DEFINE block at the | |
2174 | end of the pattern, and that you name any subpatterns defined within it. | |
2175 | ||
2176 | Also, it's worth noting that patterns defined this way probably will | |
31dc26d6 | 2177 | not be as efficient, as the optimizer is not very clever about |
e2e6a0f1 YO |
2178 | handling them. |
2179 | ||
2180 | An example of how this might be used is as follows: | |
2181 | ||
2bf803e2 | 2182 | /(?<NAME>(?&NAME_PAT))(?<ADDR>(?&ADDRESS_PAT)) |
e2e6a0f1 | 2183 | (?(DEFINE) |
2bf803e2 | 2184 | (?<NAME_PAT>....) |
22dc6719 | 2185 | (?<ADDRESS_PAT>....) |
e2e6a0f1 YO |
2186 | )/x |
2187 | ||
c27a5cfe KW |
2188 | Note that capture groups matched inside of recursion are not accessible |
2189 | after the recursion returns, so the extra layer of capturing groups is | |
e2e6a0f1 YO |
2190 | necessary. Thus C<$+{NAME_PAT}> would not be defined even though |
2191 | C<$+{NAME}> would be. | |
286f584a | 2192 | |
51a1303c BF |
2193 | Finally, keep in mind that subpatterns created inside a DEFINE block |
2194 | count towards the absolute and relative number of captures, so this: | |
2195 | ||
2196 | my @captures = "a" =~ /(.) # First capture | |
2197 | (?(DEFINE) | |
2198 | (?<EXAMPLE> 1 ) # Second capture | |
2199 | )/x; | |
2200 | say scalar @captures; | |
2201 | ||
2202 | Will output 2, not 1. This is particularly important if you intend to | |
2203 | compile the definitions with the C<qr//> operator, and later | |
2204 | interpolate them in another pattern. | |
2205 | ||
407fecf1 | 2206 | =item C<< (?>I<pattern>) >> |
e7206367 | 2207 | |
407fecf1 | 2208 | =item C<< (*atomic:I<pattern>) >> |
e7206367 KW |
2209 | X<(?E<gt>pattern)> |
2210 | X<(*atomic> | |
6bda09f9 | 2211 | X<backtrack> X<backtracking> X<atomic> X<possessive> |
5a964f20 | 2212 | |
19799a22 | 2213 | An "independent" subexpression, one which matches the substring |
407fecf1 | 2214 | that a standalone I<pattern> would match if anchored at the given |
9da458fc | 2215 | position, and it matches I<nothing other than this substring>. This |
19799a22 | 2216 | construct is useful for optimizations of what would otherwise be |
5a0de581 | 2217 | "eternal" matches, because it will not backtrack (see L</"Backtracking">). |
9da458fc IZ |
2218 | It may also be useful in places where the "grab all you can, and do not |
2219 | give anything back" semantic is desirable. | |
19799a22 | 2220 | |
c47ff5f1 | 2221 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 | 2222 | (anchored at the beginning of string, as above) will match I<all> |
57fbbe96 | 2223 | characters C<"a"> at the beginning of string, leaving no C<"a"> for |
19799a22 GS |
2224 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, |
2225 | since the match of the subgroup C<a*> is influenced by the following | |
5a0de581 | 2226 | group C<ab> (see L</"Backtracking">). In particular, C<a*> inside |
19799a22 GS |
2227 | C<a*ab> will match fewer characters than a standalone C<a*>, since |
2228 | this makes the tail match. | |
2229 | ||
407fecf1 | 2230 | C<< (?>I<pattern>) >> does not disable backtracking altogether once it has |
0b928c2f FC |
2231 | matched. It is still possible to backtrack past the construct, but not |
2232 | into it. So C<< ((?>a*)|(?>b*))ar >> will still match "bar". | |
2233 | ||
407fecf1 KW |
2234 | An effect similar to C<< (?>I<pattern>) >> may be achieved by writing |
2235 | C<(?=(I<pattern>))\g{-1}>. This matches the same substring as a standalone | |
0b928c2f | 2236 | C<a+>, and the following C<\g{-1}> eats the matched string; it therefore |
c47ff5f1 | 2237 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
2238 | (The difference between these two constructs is that the second one |
2239 | uses a capturing group, thus shifting ordinals of backreferences | |
2240 | in the rest of a regular expression.) | |
2241 | ||
2242 | Consider this pattern: | |
c277df42 | 2243 | |
871b0233 | 2244 | m{ \( |
e2e6a0f1 | 2245 | ( |
f793d64a | 2246 | [^()]+ # x+ |
e2e6a0f1 | 2247 | | |
871b0233 IZ |
2248 | \( [^()]* \) |
2249 | )+ | |
e2e6a0f1 | 2250 | \) |
871b0233 | 2251 | }x |
5a964f20 | 2252 | |
19799a22 GS |
2253 | That will efficiently match a nonempty group with matching parentheses |
2254 | two levels deep or less. However, if there is no such group, it | |
2255 | will take virtually forever on a long string. That's because there | |
2256 | are so many different ways to split a long string into several | |
2257 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
2258 | to a subpattern of the above pattern. Consider how the pattern | |
2259 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
2260 | seconds, but that each extra letter doubles this time. This | |
2261 | exponential performance will make it appear that your program has | |
14218588 | 2262 | hung. However, a tiny change to this pattern |
5a964f20 | 2263 | |
e2e6a0f1 YO |
2264 | m{ \( |
2265 | ( | |
f793d64a | 2266 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
e2e6a0f1 | 2267 | | |
871b0233 IZ |
2268 | \( [^()]* \) |
2269 | )+ | |
e2e6a0f1 | 2270 | \) |
871b0233 | 2271 | }x |
c277df42 | 2272 | |
c47ff5f1 | 2273 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 | 2274 | this yourself would be a productive exercise), but finishes in a fourth |
57fbbe96 | 2275 | the time when used on a similar string with 1000000 C<"a">s. Be aware, |
0b928c2f FC |
2276 | however, that, when this construct is followed by a |
2277 | quantifier, it currently triggers a warning message under | |
9f1b1f2d | 2278 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 2279 | C<"matches null string many times in regex">. |
c277df42 | 2280 | |
c47ff5f1 | 2281 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
f67a5002 | 2282 | effect may be achieved by negative lookahead, as in C<[^()]+ (?! [^()] )>. |
57fbbe96 | 2283 | This was only 4 times slower on a string with 1000000 C<"a">s. |
c277df42 | 2284 | |
9da458fc IZ |
2285 | The "grab all you can, and do not give anything back" semantic is desirable |
2286 | in many situations where on the first sight a simple C<()*> looks like | |
2287 | the correct solution. Suppose we parse text with comments being delimited | |
7711f978 | 2288 | by C<"#"> followed by some optional (horizontal) whitespace. Contrary to |
4375e838 | 2289 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
2290 | the comment delimiter, because it may "give up" some whitespace if |
2291 | the remainder of the pattern can be made to match that way. The correct | |
2292 | answer is either one of these: | |
2293 | ||
2294 | (?>#[ \t]*) | |
2295 | #[ \t]*(?![ \t]) | |
2296 | ||
7711f978 | 2297 | For example, to grab non-empty comments into C<$1>, one should use either |
9da458fc IZ |
2298 | one of these: |
2299 | ||
2300 | / (?> \# [ \t]* ) ( .+ ) /x; | |
2301 | / \# [ \t]* ( [^ \t] .* ) /x; | |
2302 | ||
2303 | Which one you pick depends on which of these expressions better reflects | |
2304 | the above specification of comments. | |
2305 | ||
6bda09f9 YO |
2306 | In some literature this construct is called "atomic matching" or |
2307 | "possessive matching". | |
2308 | ||
b9b4dddf YO |
2309 | Possessive quantifiers are equivalent to putting the item they are applied |
2310 | to inside of one of these constructs. The following equivalences apply: | |
2311 | ||
2312 | Quantifier Form Bracketing Form | |
2313 | --------------- --------------- | |
2314 | PAT*+ (?>PAT*) | |
2315 | PAT++ (?>PAT+) | |
2316 | PAT?+ (?>PAT?) | |
2317 | PAT{min,max}+ (?>PAT{min,max}) | |
2318 | ||
30457add KW |
2319 | Nested C<(?E<gt>...)> constructs are not no-ops, even if at first glance |
2320 | they might seem to be. This is because the nested C<(?E<gt>...)> can | |
2321 | restrict internal backtracking that otherwise might occur. For example, | |
2322 | ||
2323 | "abc" =~ /(?>a[bc]*c)/ | |
2324 | ||
2325 | matches, but | |
2326 | ||
2327 | "abc" =~ /(?>a(?>[bc]*)c)/ | |
2328 | ||
2329 | does not. | |
2330 | ||
9d1a5160 | 2331 | =item C<(?[ ])> |
f4f5fe57 | 2332 | |
572224ce | 2333 | See L<perlrecharclass/Extended Bracketed Character Classes>. |
9d1a5160 | 2334 | |
ff8bb468 AC |
2335 | Note that this feature is currently L<experimental|perlpolicy/experimental>; |
2336 | using it yields a warning in the C<experimental::regex_sets> category. | |
2337 | ||
e2e6a0f1 YO |
2338 | =back |
2339 | ||
1e0a6411 RGS |
2340 | =head2 Backtracking |
2341 | X<backtrack> X<backtracking> | |
2342 | ||
2343 | NOTE: This section presents an abstract approximation of regular | |
2344 | expression behavior. For a more rigorous (and complicated) view of | |
2345 | the rules involved in selecting a match among possible alternatives, | |
2346 | see L</Combining RE Pieces>. | |
2347 | ||
2348 | A fundamental feature of regular expression matching involves the | |
2349 | notion called I<backtracking>, which is currently used (when needed) | |
20420ba9 KW |
2350 | by all regular non-possessive expression quantifiers, namely C<"*">, |
2351 | C<*?>, C<"+">, C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often | |
2352 | optimized internally, but the general principle outlined here is valid. | |
1e0a6411 RGS |
2353 | |
2354 | For a regular expression to match, the I<entire> regular expression must | |
2355 | match, not just part of it. So if the beginning of a pattern containing a | |
2356 | quantifier succeeds in a way that causes later parts in the pattern to | |
2357 | fail, the matching engine backs up and recalculates the beginning | |
2358 | part--that's why it's called backtracking. | |
2359 | ||
2360 | Here is an example of backtracking: Let's say you want to find the | |
2361 | word following "foo" in the string "Food is on the foo table.": | |
2362 | ||
2363 | $_ = "Food is on the foo table."; | |
2364 | if ( /\b(foo)\s+(\w+)/i ) { | |
2365 | print "$2 follows $1.\n"; | |
2366 | } | |
2367 | ||
2368 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
2369 | finds a possible match right at the beginning of the string, and loads up | |
2370 | C<$1> with "Foo". However, as soon as the matching engine sees that there's | |
2371 | no whitespace following the "Foo" that it had saved in C<$1>, it realizes its | |
2372 | mistake and starts over again one character after where it had the | |
2373 | tentative match. This time it goes all the way until the next occurrence | |
2374 | of "foo". The complete regular expression matches this time, and you get | |
2375 | the expected output of "table follows foo." | |
2376 | ||
2377 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
2378 | everything between "foo" and "bar". Initially, you write something | |
2379 | like this: | |
2380 | ||
2381 | $_ = "The food is under the bar in the barn."; | |
2382 | if ( /foo(.*)bar/ ) { | |
2383 | print "got <$1>\n"; | |
2384 | } | |
2385 | ||
2386 | Which perhaps unexpectedly yields: | |
2387 | ||
2388 | got <d is under the bar in the > | |
2389 | ||
2390 | That's because C<.*> was greedy, so you get everything between the | |
2391 | I<first> "foo" and the I<last> "bar". Here it's more effective | |
2392 | to use minimal matching to make sure you get the text between a "foo" | |
2393 | and the first "bar" thereafter. | |
2394 | ||
2395 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
2396 | got <d is under the > | |
2397 | ||
2398 | Here's another example. Let's say you'd like to match a number at the end | |
2399 | of a string, and you also want to keep the preceding part of the match. | |
2400 | So you write this: | |
2401 | ||
2402 | $_ = "I have 2 numbers: 53147"; | |
2403 | if ( /(.*)(\d*)/ ) { # Wrong! | |
2404 | print "Beginning is <$1>, number is <$2>.\n"; | |
2405 | } | |
2406 | ||
2407 | That won't work at all, because C<.*> was greedy and gobbled up the | |
2408 | whole string. As C<\d*> can match on an empty string the complete | |
2409 | regular expression matched successfully. | |
2410 | ||
2411 | Beginning is <I have 2 numbers: 53147>, number is <>. | |
2412 | ||
2413 | Here are some variants, most of which don't work: | |
2414 | ||
2415 | $_ = "I have 2 numbers: 53147"; | |
2416 | @pats = qw{ | |
2417 | (.*)(\d*) | |
2418 | (.*)(\d+) | |
2419 | (.*?)(\d*) | |
2420 | (.*?)(\d+) | |
2421 | (.*)(\d+)$ | |
2422 | (.*?)(\d+)$ | |
2423 | (.*)\b(\d+)$ | |
2424 | (.*\D)(\d+)$ | |
2425 | }; | |
2426 | ||
2427 | for $pat (@pats) { | |
2428 | printf "%-12s ", $pat; | |
2429 | if ( /$pat/ ) { | |
2430 | print "<$1> <$2>\n"; | |
2431 | } else { | |
2432 | print "FAIL\n"; | |
2433 | } | |
2434 | } | |
2435 | ||
2436 | That will print out: | |
2437 | ||
2438 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
2439 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
2440 | (.*?)(\d*) <> <> | |
2441 | (.*?)(\d+) <I have > <2> | |
2442 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
2443 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
2444 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
2445 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
2446 | ||
2447 | As you see, this can be a bit tricky. It's important to realize that a | |
2448 | regular expression is merely a set of assertions that gives a definition | |
2449 | of success. There may be 0, 1, or several different ways that the | |
2450 | definition might succeed against a particular string. And if there are | |
2451 | multiple ways it might succeed, you need to understand backtracking to | |
2452 | know which variety of success you will achieve. | |
2453 | ||
2454 | When using lookahead assertions and negations, this can all get even | |
2455 | trickier. Imagine you'd like to find a sequence of non-digits not | |
2456 | followed by "123". You might try to write that as | |
2457 | ||
2458 | $_ = "ABC123"; | |
2459 | if ( /^\D*(?!123)/ ) { # Wrong! | |
2460 | print "Yup, no 123 in $_\n"; | |
2461 | } | |
2462 | ||
2463 | But that isn't going to match; at least, not the way you're hoping. It | |
2464 | claims that there is no 123 in the string. Here's a clearer picture of | |
2465 | why that pattern matches, contrary to popular expectations: | |
2466 | ||
2467 | $x = 'ABC123'; | |
2468 | $y = 'ABC445'; | |
2469 | ||
2470 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/; | |
2471 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/; | |
2472 | ||
2473 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/; | |
2474 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/; | |
2475 | ||
2476 | This prints | |
2477 | ||
2478 | 2: got ABC | |
2479 | 3: got AB | |
2480 | 4: got ABC | |
2481 | ||
2482 | You might have expected test 3 to fail because it seems to a more | |
2483 | general purpose version of test 1. The important difference between | |
2484 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
2485 | backtracking, whereas test 1 will not. What's happening is | |
2486 | that you've asked "Is it true that at the start of C<$x>, following 0 or more | |
2487 | non-digits, you have something that's not 123?" If the pattern matcher had | |
2488 | let C<\D*> expand to "ABC", this would have caused the whole pattern to | |
2489 | fail. | |
2490 | ||
2491 | The search engine will initially match C<\D*> with "ABC". Then it will | |
2492 | try to match C<(?!123)> with "123", which fails. But because | |
2493 | a quantifier (C<\D*>) has been used in the regular expression, the | |
2494 | search engine can backtrack and retry the match differently | |
2495 | in the hope of matching the complete regular expression. | |
2496 | ||
2497 | The pattern really, I<really> wants to succeed, so it uses the | |
2498 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
2499 | time. Now there's indeed something following "AB" that is not | |
2500 | "123". It's "C123", which suffices. | |
2501 | ||
2502 | We can deal with this by using both an assertion and a negation. | |
2503 | We'll say that the first part in C<$1> must be followed both by a digit | |
2504 | and by something that's not "123". Remember that the lookaheads | |
2505 | are zero-width expressions--they only look, but don't consume any | |
2506 | of the string in their match. So rewriting this way produces what | |
2507 | you'd expect; that is, case 5 will fail, but case 6 succeeds: | |
2508 | ||
2509 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/; | |
2510 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/; | |
2511 | ||
2512 | 6: got ABC | |
2513 | ||
2514 | In other words, the two zero-width assertions next to each other work as though | |
2515 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> | |
2516 | matches only if you're at the beginning of the line AND the end of the | |
2517 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
2518 | regular expressions always means AND, except when you write an explicit OR | |
2519 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
2520 | although the attempted matches are made at different positions because "a" | |
2521 | is not a zero-width assertion, but a one-width assertion. | |
2522 | ||
2523 | B<WARNING>: Particularly complicated regular expressions can take | |
2524 | exponential time to solve because of the immense number of possible | |
2525 | ways they can use backtracking to try for a match. For example, without | |
2526 | internal optimizations done by the regular expression engine, this will | |
2527 | take a painfully long time to run: | |
2528 | ||
2529 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ | |
2530 | ||
2531 | And if you used C<"*">'s in the internal groups instead of limiting them | |
2532 | to 0 through 5 matches, then it would take forever--or until you ran | |
2533 | out of stack space. Moreover, these internal optimizations are not | |
2534 | always applicable. For example, if you put C<{0,5}> instead of C<"*"> | |
2535 | on the external group, no current optimization is applicable, and the | |
2536 | match takes a long time to finish. | |
2537 | ||
2538 | A powerful tool for optimizing such beasts is what is known as an | |
2539 | "independent group", | |
2a920e30 | 2540 | which does not backtrack (see C<L</(?E<gt>pattern)>>). Note also that |
1e0a6411 RGS |
2541 | zero-length lookahead/lookbehind assertions will not backtrack to make |
2542 | the tail match, since they are in "logical" context: only | |
2543 | whether they match is considered relevant. For an example | |
2544 | where side-effects of lookahead I<might> have influenced the | |
2a920e30 | 2545 | following match, see C<L</(?E<gt>pattern)>>. |
1e0a6411 | 2546 | |
034602eb | 2547 | =head2 Script Runs |
d9790612 | 2548 | X<(*script_run:...)> X<(sr:...)> |
3337229c | 2549 | X<(*atomic_script_run:...)> X<(asr:...)> |
034602eb KW |
2550 | |
2551 | A script run is basically a sequence of characters, all from the same | |
2552 | Unicode script (see L<perlunicode/Scripts>), such as Latin or Greek. In | |
2553 | most places a single word would never be written in multiple scripts, | |
2554 | unless it is a spoofing attack. An infamous example, is | |
2555 | ||
2556 | paypal.com | |
2557 | ||
2558 | Those letters could all be Latin (as in the example just above), or they | |
2559 | could be all Cyrillic (except for the dot), or they could be a mixture | |
2560 | of the two. In the case of an internet address the C<.com> would be in | |
2561 | Latin, And any Cyrillic ones would cause it to be a mixture, not a | |
2562 | script run. Someone clicking on such a link would not be directed to | |
2563 | the real Paypal website, but an attacker would craft a look-alike one to | |
2564 | attempt to gather sensitive information from the person. | |
2565 | ||
2566 | Starting in Perl 5.28, it is now easy to detect strings that aren't | |
d9790612 KW |
2567 | script runs. Simply enclose just about any pattern like either of |
2568 | these: | |
034602eb | 2569 | |
d9790612 KW |
2570 | (*script_run:pattern) |
2571 | (*sr:pattern) | |
034602eb KW |
2572 | |
2573 | What happens is that after I<pattern> succeeds in matching, it is | |
2574 | subjected to the additional criterion that every character in it must be | |
2575 | from the same script (see exceptions below). If this isn't true, | |
2576 | backtracking occurs until something all in the same script is found that | |
2577 | matches, or all possibilities are exhausted. This can cause a lot of | |
2578 | backtracking, but generally, only malicious input will result in this, | |
2579 | though the slow down could cause a denial of service attack. If your | |
4a1d9640 KW |
2580 | needs permit, it is best to make the pattern atomic to cut down on the |
2581 | amount of backtracking. This is so likely to be what you want, that | |
2582 | instead of writing this: | |
034602eb | 2583 | |
d9790612 | 2584 | (*script_run:(?>pattern)) |
034602eb | 2585 | |
3337229c KW |
2586 | you can write either of these: |
2587 | ||
2588 | (*atomic_script_run:pattern) | |
2589 | (*asr:pattern) | |
2590 | ||
eb992c6f | 2591 | (See C<L</(?E<gt>I<pattern>)>>.) |
034602eb KW |
2592 | |
2593 | In Taiwan, Japan, and Korea, it is common for text to have a mixture of | |
2594 | characters from their native scripts and base Chinese. Perl follows | |
e2176993 | 2595 | Unicode's UTS 39 (L<https://unicode.org/reports/tr39/>) Unicode Security |
4a1d9640 KW |
2596 | Mechanisms in allowing such mixtures. For example, the Japanese scripts |
2597 | Katakana and Hiragana are commonly mixed together in practice, along | |
2598 | with some Chinese characters, and hence are treated as being in a single | |
2599 | script run by Perl. | |
034602eb | 2600 | |
f4e61fc0 | 2601 | The rules used for matching decimal digits are slightly stricter. Many |
034602eb KW |
2602 | scripts have their own sets of digits equivalent to the Western C<0> |
2603 | through C<9> ones. A few, such as Arabic, have more than one set. For | |
2604 | a string to be considered a script run, all digits in it must come from | |
f4e61fc0 KW |
2605 | the same set of ten, as determined by the first digit encountered. |
2606 | As an example, | |
034602eb | 2607 | |
d9790612 | 2608 | qr/(*script_run: \d+ \b )/x |
034602eb KW |
2609 | |
2610 | guarantees that the digits matched will all be from the same set of 10. | |
2611 | You won't get a look-alike digit from a different script that has a | |
2612 | different value than what it appears to be. | |
2613 | ||
2614 | Unicode has three pseudo scripts that are handled specially. | |
2615 | ||
2616 | "Unknown" is applied to code points whose meaning has yet to be | |
2617 | determined. Perl currently will match as a script run, any single | |
2618 | character string consisting of one of these code points. But any string | |
2619 | longer than one code point containing one of these will not be | |
2620 | considered a script run. | |
2621 | ||
2622 | "Inherited" is applied to characters that modify another, such as an | |
2623 | accent of some type. These are considered to be in the script of the | |
2624 | master character, and so never cause a script run to not match. | |
2625 | ||
2626 | The other one is "Common". This consists of mostly punctuation, emoji, | |
f4e61fc0 KW |
2627 | and characters used in mathematics and music, the ASCII digits C<0> |
2628 | through C<9>, and full-width forms of these digits. These characters | |
2629 | can appear intermixed in text in many of the world's scripts. These | |
2630 | also don't cause a script run to not match. But like other scripts, all | |
2631 | digits in a run must come from the same set of 10. | |
034602eb KW |
2632 | |
2633 | This construct is non-capturing. You can add parentheses to I<pattern> | |
2634 | to capture, if desired. You will have to do this if you plan to use | |
2635 | L</(*ACCEPT) (*ACCEPT:arg)> and not have it bypass the script run | |
2636 | checking. | |
2637 | ||
4a1d9640 | 2638 | The C<Script_Extensions> property as modified by UTS 39 |
e2176993 | 2639 | (L<https://unicode.org/reports/tr39/>) is used as the basis for this |
4a1d9640 KW |
2640 | feature. |
2641 | ||
2642 | To summarize, | |
2643 | ||
2644 | =over 4 | |
2645 | ||
2646 | =item * | |
2647 | ||
2648 | All length 0 or length 1 sequences are script runs. | |
2649 | ||
2650 | =item * | |
2651 | ||
2652 | A longer sequence is a script run if and only if B<all> of the following | |
2653 | conditions are met: | |
2654 | ||
2655 | Z<> | |
2656 | ||
2657 | =over | |
2658 | ||
2659 | =item 1 | |
2660 | ||
2661 | No code point in the sequence has the C<Script_Extension> property of | |
2662 | C<Unknown>. | |
2663 | ||
3f4fa0ec | 2664 | This currently means that all code points in the sequence have been |
4a1d9640 KW |
2665 | assigned by Unicode to be characters that aren't private use nor |
2666 | surrogate code points. | |
2667 | ||
2668 | =item 2 | |
2669 | ||
2670 | All characters in the sequence come from the Common script and/or the | |
2671 | Inherited script and/or a single other script. | |
2672 | ||
2673 | The script of a character is determined by the C<Script_Extensions> | |
e2176993 | 2674 | property as modified by UTS 39 (L<https://unicode.org/reports/tr39/>), as |
4a1d9640 KW |
2675 | described above. |
2676 | ||
2677 | =item 3 | |
2678 | ||
2679 | All decimal digits in the sequence come from the same block of 10 | |
2680 | consecutive digits. | |
2681 | ||
2682 | =back | |
2683 | ||
2684 | =back | |
034602eb | 2685 | |
e2e6a0f1 YO |
2686 | =head2 Special Backtracking Control Verbs |
2687 | ||
5fcc329e KW |
2688 | These special patterns are generally of the form C<(*I<VERB>:I<arg>)>. Unless |
2689 | otherwise stated the I<arg> argument is optional; in some cases, it is | |
fee50582 | 2690 | mandatory. |
e2e6a0f1 YO |
2691 | |
2692 | Any pattern containing a special backtracking verb that allows an argument | |
e1020413 | 2693 | has the special behaviour that when executed it sets the current package's |
5d458dd8 YO |
2694 | C<$REGERROR> and C<$REGMARK> variables. When doing so the following |
2695 | rules apply: | |
e2e6a0f1 | 2696 | |
5fcc329e | 2697 | On failure, the C<$REGERROR> variable will be set to the I<arg> value of the |
5d458dd8 | 2698 | verb pattern, if the verb was involved in the failure of the match. If the |
5fcc329e | 2699 | I<arg> part of the pattern was omitted, then C<$REGERROR> will be set to the |
407fecf1 | 2700 | name of the last C<(*MARK:I<NAME>)> pattern executed, or to TRUE if there was |
5d458dd8 | 2701 | none. Also, the C<$REGMARK> variable will be set to FALSE. |
e2e6a0f1 | 2702 | |
5d458dd8 YO |
2703 | On a successful match, the C<$REGERROR> variable will be set to FALSE, and |
2704 | the C<$REGMARK> variable will be set to the name of the last | |
407fecf1 KW |
2705 | C<(*MARK:I<NAME>)> pattern executed. See the explanation for the |
2706 | C<(*MARK:I<NAME>)> verb below for more details. | |
e2e6a0f1 | 2707 | |
5d458dd8 | 2708 | B<NOTE:> C<$REGERROR> and C<$REGMARK> are not magic variables like C<$1> |
0b928c2f | 2709 | and most other regex-related variables. They are not local to a scope, nor |
5d458dd8 | 2710 | readonly, but instead are volatile package variables similar to C<$AUTOLOAD>. |
ddb07903 AC |
2711 | They are set in the package containing the code that I<executed> the regex |
2712 | (rather than the one that compiled it, where those differ). If necessary, you | |
2713 | can use C<local> to localize changes to these variables to a specific scope | |
2714 | before executing a regex. | |
e2e6a0f1 YO |
2715 | |
2716 | If a pattern does not contain a special backtracking verb that allows an | |
5d458dd8 | 2717 | argument, then C<$REGERROR> and C<$REGMARK> are not touched at all. |
e2e6a0f1 | 2718 | |
70ca8714 | 2719 | =over 3 |
e2e6a0f1 | 2720 | |
fee50582 | 2721 | =item Verbs |
e2e6a0f1 YO |
2722 | |
2723 | =over 4 | |
2724 | ||
407fecf1 | 2725 | =item C<(*PRUNE)> C<(*PRUNE:I<NAME>)> |
f7819f85 | 2726 | X<(*PRUNE)> X<(*PRUNE:NAME)> |
54612592 | 2727 | |
5d458dd8 | 2728 | This zero-width pattern prunes the backtracking tree at the current point |
a95b7a20 | 2729 | when backtracked into on failure. Consider the pattern C</I<A> (*PRUNE) I<B>/>, |
7711f978 KW |
2730 | where I<A> and I<B> are complex patterns. Until the C<(*PRUNE)> verb is reached, |
2731 | I<A> may backtrack as necessary to match. Once it is reached, matching | |
2732 | continues in I<B>, which may also backtrack as necessary; however, should B | |
5d458dd8 YO |
2733 | not match, then no further backtracking will take place, and the pattern |
2734 | will fail outright at the current starting position. | |
54612592 YO |
2735 | |
2736 | The following example counts all the possible matching strings in a | |
2737 | pattern (without actually matching any of them). | |
2738 | ||
e2e6a0f1 | 2739 | 'aaab' =~ /a+b?(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
2740 | print "Count=$count\n"; |
2741 | ||
2742 | which produces: | |
2743 | ||
2744 | aaab | |
2745 | aaa | |
2746 | aa | |
2747 | a | |
2748 | aab | |
2749 | aa | |
2750 | a | |
2751 | ab | |
2752 | a | |
2753 | Count=9 | |
2754 | ||
5d458dd8 | 2755 | If we add a C<(*PRUNE)> before the count like the following |
54612592 | 2756 | |
5d458dd8 | 2757 | 'aaab' =~ /a+b?(*PRUNE)(?{print "$&\n"; $count++})(*FAIL)/; |
54612592 YO |
2758 | print "Count=$count\n"; |
2759 | ||
0b928c2f | 2760 | we prevent backtracking and find the count of the longest matching string |
353c6505 | 2761 | at each matching starting point like so: |
54612592 YO |
2762 | |
2763 | aaab | |
2764 | aab | |
2765 | ab | |
2766 | Count=3 | |
2767 | ||
5d458dd8 | 2768 | Any number of C<(*PRUNE)> assertions may be used in a pattern. |
54612592 | 2769 | |
407fecf1 | 2770 | See also C<<< L<< /(?>I<pattern>) >> >>> and possessive quantifiers for |
4b05bc8e | 2771 | other ways to |
5d458dd8 YO |
2772 | control backtracking. In some cases, the use of C<(*PRUNE)> can be |
2773 | replaced with a C<< (?>pattern) >> with no functional difference; however, | |
2774 | C<(*PRUNE)> can be used to handle cases that cannot be expressed using a | |
2775 | C<< (?>pattern) >> alone. | |
54612592 | 2776 | |
407fecf1 | 2777 | =item C<(*SKIP)> C<(*SKIP:I<NAME>)> |
5d458dd8 | 2778 | X<(*SKIP)> |
e2e6a0f1 | 2779 | |
5d458dd8 | 2780 | This zero-width pattern is similar to C<(*PRUNE)>, except that on |
e2e6a0f1 | 2781 | failure it also signifies that whatever text that was matched leading up |
5d458dd8 YO |
2782 | to the C<(*SKIP)> pattern being executed cannot be part of I<any> match |
2783 | of this pattern. This effectively means that the regex engine "skips" forward | |
2784 | to this position on failure and tries to match again, (assuming that | |
2785 | there is sufficient room to match). | |
2786 | ||
407fecf1 KW |
2787 | The name of the C<(*SKIP:I<NAME>)> pattern has special significance. If a |
2788 | C<(*MARK:I<NAME>)> was encountered while matching, then it is that position | |
5d458dd8 YO |
2789 | which is used as the "skip point". If no C<(*MARK)> of that name was |
2790 | encountered, then the C<(*SKIP)> operator has no effect. When used | |
2791 | without a name the "skip point" is where the match point was when | |
7711f978 | 2792 | executing the C<(*SKIP)> pattern. |
5d458dd8 | 2793 | |
0b928c2f | 2794 | Compare the following to the examples in C<(*PRUNE)>; note the string |
24b23f37 YO |
2795 | is twice as long: |
2796 | ||
d1fbf752 KW |
2797 | 'aaabaaab' =~ /a+b?(*SKIP)(?{print "$&\n"; $count++})(*FAIL)/; |
2798 | print "Count=$count\n"; | |
24b23f37 YO |
2799 | |
2800 | outputs | |
2801 | ||
2802 | aaab | |
2803 | aaab | |
2804 | Count=2 | |
2805 | ||
5d458dd8 | 2806 | Once the 'aaab' at the start of the string has matched, and the C<(*SKIP)> |
353c6505 | 2807 | executed, the next starting point will be where the cursor was when the |
5d458dd8 YO |
2808 | C<(*SKIP)> was executed. |
2809 | ||
407fecf1 | 2810 | =item C<(*MARK:I<NAME>)> C<(*:I<NAME>)> |
b16db30f | 2811 | X<(*MARK)> X<(*MARK:NAME)> X<(*:NAME)> |
5d458dd8 YO |
2812 | |
2813 | This zero-width pattern can be used to mark the point reached in a string | |
2814 | when a certain part of the pattern has been successfully matched. This | |
2815 | mark may be given a name. A later C<(*SKIP)> pattern will then skip | |
2816 | forward to that point if backtracked into on failure. Any number of | |
7711f978 | 2817 | C<(*MARK)> patterns are allowed, and the I<NAME> portion may be duplicated. |
5d458dd8 | 2818 | |
407fecf1 | 2819 | In addition to interacting with the C<(*SKIP)> pattern, C<(*MARK:I<NAME>)> |
5d458dd8 YO |
2820 | can be used to "label" a pattern branch, so that after matching, the |
2821 | program can determine which branches of the pattern were involved in the | |
2822 | match. | |
2823 | ||
2824 | When a match is successful, the C<$REGMARK> variable will be set to the | |
407fecf1 | 2825 | name of the most recently executed C<(*MARK:I<NAME>)> that was involved |
5d458dd8 YO |
2826 | in the match. |
2827 | ||
2828 | This can be used to determine which branch of a pattern was matched | |
c27a5cfe | 2829 | without using a separate capture group for each branch, which in turn |
5d458dd8 YO |
2830 | can result in a performance improvement, as perl cannot optimize |
2831 | C</(?:(x)|(y)|(z))/> as efficiently as something like | |
2832 | C</(?:x(*MARK:x)|y(*MARK:y)|z(*MARK:z))/>. | |
2833 | ||
2834 | When a match has failed, and unless another verb has been involved in | |
2835 | failing the match and has provided its own name to use, the C<$REGERROR> | |
2836 | variable will be set to the name of the most recently executed | |
407fecf1 | 2837 | C<(*MARK:I<NAME>)>. |
5d458dd8 | 2838 | |
42ac7c82 | 2839 | See L</(*SKIP)> for more details. |
5d458dd8 | 2840 | |
407fecf1 | 2841 | As a shortcut C<(*MARK:I<NAME>)> can be written C<(*:I<NAME>)>. |
b62d2d15 | 2842 | |
407fecf1 | 2843 | =item C<(*THEN)> C<(*THEN:I<NAME>)> |
5d458dd8 | 2844 | |
17b35041 | 2845 | This is similar to the "cut group" operator C<::> from Raku. Like |
5d458dd8 YO |
2846 | C<(*PRUNE)>, this verb always matches, and when backtracked into on |
2847 | failure, it causes the regex engine to try the next alternation in the | |
ac9d8485 | 2848 | innermost enclosing group (capturing or otherwise) that has alternations. |
407fecf1 | 2849 | The two branches of a C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)> do not |
ac9d8485 | 2850 | count as an alternation, as far as C<(*THEN)> is concerned. |
5d458dd8 YO |
2851 | |
2852 | Its name comes from the observation that this operation combined with the | |
7711f978 | 2853 | alternation operator (C<"|">) can be used to create what is essentially a |
5d458dd8 YO |
2854 | pattern-based if/then/else block: |
2855 | ||
2856 | ( COND (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) | |
2857 | ||
2858 | Note that if this operator is used and NOT inside of an alternation then | |
2859 | it acts exactly like the C<(*PRUNE)> operator. | |
2860 | ||
2861 | / A (*PRUNE) B / | |
2862 | ||
2863 | is the same as | |
2864 | ||
2865 | / A (*THEN) B / | |
2866 | ||
2867 | but | |
2868 | ||
25e26d77 | 2869 | / ( A (*THEN) B | C ) / |
5d458dd8 YO |
2870 | |
2871 | is not the same as | |
2872 | ||
25e26d77 | 2873 | / ( A (*PRUNE) B | C ) / |
5d458dd8 | 2874 | |
7711f978 KW |
2875 | as after matching the I<A> but failing on the I<B> the C<(*THEN)> verb will |
2876 | backtrack and try I<C>; but the C<(*PRUNE)> verb will simply fail. | |
24b23f37 | 2877 | |
5fcc329e | 2878 | =item C<(*COMMIT)> C<(*COMMIT:I<arg>)> |
e2e6a0f1 | 2879 | X<(*COMMIT)> |
24b23f37 | 2880 | |
17b35041 | 2881 | This is the Raku "commit pattern" C<< <commit> >> or C<:::>. It's a |
5d458dd8 YO |
2882 | zero-width pattern similar to C<(*SKIP)>, except that when backtracked |
2883 | into on failure it causes the match to fail outright. No further attempts | |
2884 | to find a valid match by advancing the start pointer will occur again. | |
2885 | For example, | |
24b23f37 | 2886 | |
d1fbf752 KW |
2887 | 'aaabaaab' =~ /a+b?(*COMMIT)(?{print "$&\n"; $count++})(*FAIL)/; |
2888 | print "Count=$count\n"; | |
24b23f37 YO |
2889 | |
2890 | outputs | |
2891 | ||
2892 | aaab | |
2893 | Count=1 | |
2894 | ||
e2e6a0f1 YO |
2895 | In other words, once the C<(*COMMIT)> has been entered, and if the pattern |
2896 | does not match, the regex engine will not try any further matching on the | |
2897 | rest of the string. | |
c277df42 | 2898 | |
407fecf1 | 2899 | =item C<(*FAIL)> C<(*F)> C<(*FAIL:I<arg>)> |
e2e6a0f1 | 2900 | X<(*FAIL)> X<(*F)> |
9af228c6 | 2901 | |
e2e6a0f1 YO |
2902 | This pattern matches nothing and always fails. It can be used to force the |
2903 | engine to backtrack. It is equivalent to C<(?!)>, but easier to read. In | |
fee50582 | 2904 | fact, C<(?!)> gets optimised into C<(*FAIL)> internally. You can provide |
7711f978 KW |
2905 | an argument so that if the match fails because of this C<FAIL> directive |
2906 | the argument can be obtained from C<$REGERROR>. | |
9af228c6 | 2907 | |
e2e6a0f1 | 2908 | It is probably useful only when combined with C<(?{})> or C<(??{})>. |
9af228c6 | 2909 | |
407fecf1 | 2910 | =item C<(*ACCEPT)> C<(*ACCEPT:I<arg>)> |
e2e6a0f1 | 2911 | X<(*ACCEPT)> |
9af228c6 | 2912 | |
e2e6a0f1 YO |
2913 | This pattern matches nothing and causes the end of successful matching at |
2914 | the point at which the C<(*ACCEPT)> pattern was encountered, regardless of | |
2915 | whether there is actually more to match in the string. When inside of a | |
0d017f4d | 2916 | nested pattern, such as recursion, or in a subpattern dynamically generated |
e2e6a0f1 | 2917 | via C<(??{})>, only the innermost pattern is ended immediately. |
9af228c6 | 2918 | |
c27a5cfe | 2919 | If the C<(*ACCEPT)> is inside of capturing groups then the groups are |
e2e6a0f1 YO |
2920 | marked as ended at the point at which the C<(*ACCEPT)> was encountered. |
2921 | For instance: | |
9af228c6 | 2922 | |
e2e6a0f1 | 2923 | 'AB' =~ /(A (A|B(*ACCEPT)|C) D)(E)/x; |
9af228c6 | 2924 | |
57fbbe96 | 2925 | will match, and C<$1> will be C<AB> and C<$2> will be C<"B">, C<$3> will not |
0b928c2f | 2926 | be set. If another branch in the inner parentheses was matched, such as in the |
57fbbe96 | 2927 | string 'ACDE', then the C<"D"> and C<"E"> would have to be matched as well. |
9af228c6 | 2928 | |
7711f978 KW |
2929 | You can provide an argument, which will be available in the var |
2930 | C<$REGMARK> after the match completes. | |
fee50582 | 2931 | |
9af228c6 | 2932 | =back |
c277df42 | 2933 | |
a0d0e21e LW |
2934 | =back |
2935 | ||
7711f978 | 2936 | =head2 Warning on C<\1> Instead of C<$1> |
cb1a09d0 | 2937 | |
5a964f20 | 2938 | Some people get too used to writing things like: |
cb1a09d0 AD |
2939 | |
2940 | $pattern =~ s/(\W)/\\\1/g; | |
2941 | ||
3ff1c45a KW |
2942 | This is grandfathered (for \1 to \9) for the RHS of a substitute to avoid |
2943 | shocking the | |
cb1a09d0 | 2944 | B<sed> addicts, but it's a dirty habit to get into. That's because in |
d1be9408 | 2945 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
2946 | the usual double-quoted string means a control-A. The customary Unix |
2947 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
2948 | of doing that, you get yourself into trouble if you then add an C</e> | |
2949 | modifier. | |
2950 | ||
f793d64a | 2951 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
2952 | |
2953 | Or if you try to do | |
2954 | ||
2955 | s/(\d+)/\1000/; | |
2956 | ||
2957 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 2958 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
2959 | with the operation of matching a backreference. Certainly they mean two |
2960 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 2961 | |
0d017f4d | 2962 | =head2 Repeated Patterns Matching a Zero-length Substring |
c84d73f1 | 2963 | |
19799a22 | 2964 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
2965 | |
2966 | Regular expressions provide a terse and powerful programming language. As | |
2967 | with most other power tools, power comes together with the ability | |
2968 | to wreak havoc. | |
2969 | ||
2970 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 2971 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
2972 | |
2973 | 'foo' =~ m{ ( o? )* }x; | |
2974 | ||
57fbbe96 | 2975 | The C<o?> matches at the beginning of "C<foo>", and since the position |
c84d73f1 | 2976 | in the string is not moved by the match, C<o?> would match again and again |
7711f978 | 2977 | because of the C<"*"> quantifier. Another common way to create a similar cycle |
f1dc5bb2 | 2978 | is with the looping modifier C</g>: |
c84d73f1 IZ |
2979 | |
2980 | @matches = ( 'foo' =~ m{ o? }xg ); | |
2981 | ||
2982 | or | |
2983 | ||
2984 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
2985 | ||
7711f978 | 2986 | or the loop implied by C<split()>. |
c84d73f1 IZ |
2987 | |
2988 | However, long experience has shown that many programming tasks may | |
14218588 GS |
2989 | be significantly simplified by using repeated subexpressions that |
2990 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 | 2991 | |
d1fbf752 | 2992 | @chars = split //, $string; # // is not magic in split |
c84d73f1 IZ |
2993 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / |
2994 | ||
9da458fc | 2995 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 | 2996 | the infinite loop>. The rules for this are different for lower-level |
527e91da | 2997 | loops given by the greedy quantifiers C<*+{}>, and for higher-level |
7711f978 | 2998 | ones like the C</g> modifier or C<split()> operator. |
c84d73f1 | 2999 | |
19799a22 GS |
3000 | The lower-level loops are I<interrupted> (that is, the loop is |
3001 | broken) when Perl detects that a repeated expression matched a | |
3002 | zero-length substring. Thus | |
c84d73f1 IZ |
3003 | |
3004 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
3005 | ||
5d458dd8 | 3006 | is made equivalent to |
c84d73f1 | 3007 | |
0b928c2f FC |
3008 | m{ (?: NON_ZERO_LENGTH )* (?: ZERO_LENGTH )? }x; |
3009 | ||
3010 | For example, this program | |
3011 | ||
3012 | #!perl -l | |
3013 | "aaaaab" =~ / | |
3014 | (?: | |
3015 | a # non-zero | |
3016 | | # or | |
3017 | (?{print "hello"}) # print hello whenever this | |
3018 | # branch is tried | |
3019 | (?=(b)) # zero-width assertion | |
3020 | )* # any number of times | |
3021 | /x; | |
3022 | print $&; | |
3023 | print $1; | |
c84d73f1 | 3024 | |
0b928c2f FC |
3025 | prints |
3026 | ||
3027 | hello | |
3028 | aaaaa | |
3029 | b | |
3030 | ||
3031 | Notice that "hello" is only printed once, as when Perl sees that the sixth | |
3032 | iteration of the outermost C<(?:)*> matches a zero-length string, it stops | |
7711f978 | 3033 | the C<"*">. |
0b928c2f FC |
3034 | |
3035 | The higher-level loops preserve an additional state between iterations: | |
5d458dd8 | 3036 | whether the last match was zero-length. To break the loop, the following |
c84d73f1 | 3037 | match after a zero-length match is prohibited to have a length of zero. |
5a0de581 | 3038 | This prohibition interacts with backtracking (see L</"Backtracking">), |
c84d73f1 IZ |
3039 | and so the I<second best> match is chosen if the I<best> match is of |
3040 | zero length. | |
3041 | ||
19799a22 | 3042 | For example: |
c84d73f1 IZ |
3043 | |
3044 | $_ = 'bar'; | |
3045 | s/\w??/<$&>/g; | |
3046 | ||
20fb949f | 3047 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
5d458dd8 | 3048 | match given by non-greedy C<??> is the zero-length match, and the I<second |
c84d73f1 IZ |
3049 | best> match is what is matched by C<\w>. Thus zero-length matches |
3050 | alternate with one-character-long matches. | |
3051 | ||
5d458dd8 | 3052 | Similarly, for repeated C<m/()/g> the second-best match is the match at the |
c84d73f1 IZ |
3053 | position one notch further in the string. |
3054 | ||
19799a22 | 3055 | The additional state of being I<matched with zero-length> is associated with |
7711f978 | 3056 | the matched string, and is reset by each assignment to C<pos()>. |
9da458fc IZ |
3057 | Zero-length matches at the end of the previous match are ignored |
3058 | during C<split>. | |
c84d73f1 | 3059 | |
0d017f4d | 3060 | =head2 Combining RE Pieces |
35a734be IZ |
3061 | |
3062 | Each of the elementary pieces of regular expressions which were described | |
3063 | before (such as C<ab> or C<\Z>) could match at most one substring | |
3064 | at the given position of the input string. However, in a typical regular | |
3065 | expression these elementary pieces are combined into more complicated | |
57fbbe96 KW |
3066 | patterns using combining operators C<ST>, C<S|T>, C<S*> I<etc>. |
3067 | (in these examples C<"S"> and C<"T"> are regular subexpressions). | |
35a734be IZ |
3068 | |
3069 | Such combinations can include alternatives, leading to a problem of choice: | |
3070 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
3071 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
5a0de581 | 3072 | actually matched is the concept of backtracking (see L</"Backtracking">). |
35a734be IZ |
3073 | However, this description is too low-level and makes you think |
3074 | in terms of a particular implementation. | |
3075 | ||
3076 | Another description starts with notions of "better"/"worse". All the | |
3077 | substrings which may be matched by the given regular expression can be | |
3078 | sorted from the "best" match to the "worst" match, and it is the "best" | |
3079 | match which is chosen. This substitutes the question of "what is chosen?" | |
3080 | by the question of "which matches are better, and which are worse?". | |
3081 | ||
3082 | Again, for elementary pieces there is no such question, since at most | |
3083 | one match at a given position is possible. This section describes the | |
3084 | notion of better/worse for combining operators. In the description | |
57fbbe96 | 3085 | below C<"S"> and C<"T"> are regular subexpressions. |
35a734be | 3086 | |
13a2d996 | 3087 | =over 4 |
35a734be IZ |
3088 | |
3089 | =item C<ST> | |
3090 | ||
57fbbe96 KW |
3091 | Consider two possible matches, C<AB> and C<A'B'>, C<"A"> and C<A'> are |
3092 | substrings which can be matched by C<"S">, C<"B"> and C<B'> are substrings | |
3093 | which can be matched by C<"T">. | |
35a734be | 3094 | |
57fbbe96 | 3095 | If C<"A"> is a better match for C<"S"> than C<A'>, C<AB> is a better |
35a734be IZ |
3096 | match than C<A'B'>. |
3097 | ||
57fbbe96 KW |
3098 | If C<"A"> and C<A'> coincide: C<AB> is a better match than C<AB'> if |
3099 | C<"B"> is a better match for C<"T"> than C<B'>. | |
35a734be IZ |
3100 | |
3101 | =item C<S|T> | |
3102 | ||
57fbbe96 | 3103 | When C<"S"> can match, it is a better match than when only C<"T"> can match. |
35a734be | 3104 | |
57fbbe96 KW |
3105 | Ordering of two matches for C<"S"> is the same as for C<"S">. Similar for |
3106 | two matches for C<"T">. | |
35a734be IZ |
3107 | |
3108 | =item C<S{REPEAT_COUNT}> | |
3109 | ||
3110 | Matches as C<SSS...S> (repeated as many times as necessary). | |
3111 | ||
3112 | =item C<S{min,max}> | |
3113 | ||
3114 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
3115 | ||
3116 | =item C<S{min,max}?> | |
3117 | ||
3118 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
3119 | ||
3120 | =item C<S?>, C<S*>, C<S+> | |
3121 | ||
3122 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
3123 | ||
3124 | =item C<S??>, C<S*?>, C<S+?> | |
3125 | ||
3126 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
3127 | ||
c47ff5f1 | 3128 | =item C<< (?>S) >> |
35a734be | 3129 | |
57fbbe96 | 3130 | Matches the best match for C<"S"> and only that. |
35a734be IZ |
3131 | |
3132 | =item C<(?=S)>, C<(?<=S)> | |
3133 | ||
57fbbe96 KW |
3134 | Only the best match for C<"S"> is considered. (This is important only if |
3135 | C<"S"> has capturing parentheses, and backreferences are used somewhere | |
35a734be IZ |
3136 | else in the whole regular expression.) |
3137 | ||
3138 | =item C<(?!S)>, C<(?<!S)> | |
3139 | ||
3140 | For this grouping operator there is no need to describe the ordering, since | |
57fbbe96 | 3141 | only whether or not C<"S"> can match is important. |
35a734be | 3142 | |
407fecf1 | 3143 | =item C<(??{ I<EXPR> })>, C<(?I<PARNO>)> |
35a734be IZ |
3144 | |
3145 | The ordering is the same as for the regular expression which is | |
407fecf1 | 3146 | the result of I<EXPR>, or the pattern contained by capture group I<PARNO>. |
35a734be | 3147 | |
407fecf1 | 3148 | =item C<(?(I<condition>)I<yes-pattern>|I<no-pattern>)> |
35a734be | 3149 | |
407fecf1 | 3150 | Recall that which of I<yes-pattern> or I<no-pattern> actually matches is |
35a734be IZ |
3151 | already determined. The ordering of the matches is the same as for the |
3152 | chosen subexpression. | |
3153 | ||
3154 | =back | |
3155 | ||
3156 | The above recipes describe the ordering of matches I<at a given position>. | |
3157 | One more rule is needed to understand how a match is determined for the | |
3158 | whole regular expression: a match at an earlier position is always better | |
3159 | than a match at a later position. | |
3160 | ||
0d017f4d | 3161 | =head2 Creating Custom RE Engines |
c84d73f1 | 3162 | |
0b928c2f FC |
3163 | As of Perl 5.10.0, one can create custom regular expression engines. This |
3164 | is not for the faint of heart, as they have to plug in at the C level. See | |
3165 | L<perlreapi> for more details. | |
3166 | ||
3167 | As an alternative, overloaded constants (see L<overload>) provide a simple | |
3168 | way to extend the functionality of the RE engine, by substituting one | |
3169 | pattern for another. | |
c84d73f1 IZ |
3170 | |
3171 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
0d017f4d | 3172 | matches at a boundary between whitespace characters and non-whitespace |
c84d73f1 IZ |
3173 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly |
3174 | at these positions, so we want to have each C<\Y|> in the place of the | |
3175 | more complicated version. We can create a module C<customre> to do | |
3176 | this: | |
3177 | ||
3178 | package customre; | |
3179 | use overload; | |
3180 | ||
3181 | sub import { | |
3182 | shift; | |
3183 | die "No argument to customre::import allowed" if @_; | |
3184 | overload::constant 'qr' => \&convert; | |
3185 | } | |
3186 | ||
3187 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
3188 | ||
580a9fe1 RGS |
3189 | # We must also take care of not escaping the legitimate \\Y| |
3190 | # sequence, hence the presence of '\\' in the conversion rules. | |
5d458dd8 | 3191 | my %rules = ( '\\' => '\\\\', |
f793d64a | 3192 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); |
c84d73f1 IZ |
3193 | sub convert { |
3194 | my $re = shift; | |
5d458dd8 | 3195 | $re =~ s{ |
c84d73f1 IZ |
3196 | \\ ( \\ | Y . ) |
3197 | } | |
5d458dd8 | 3198 | { $rules{$1} or invalid($re,$1) }sgex; |
c84d73f1 IZ |
3199 | return $re; |
3200 | } | |
3201 | ||
3202 | Now C<use customre> enables the new escape in constant regular | |
57fbbe96 | 3203 | expressions, I<i.e.>, those without any runtime variable interpolations. |
c84d73f1 IZ |
3204 | As documented in L<overload>, this conversion will work only over |
3205 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
3206 | part of this regular expression needs to be converted explicitly | |
7711f978 | 3207 | (but only if the special meaning of C<\Y|> should be enabled inside C<$re>): |
c84d73f1 IZ |
3208 | |
3209 | use customre; | |
3210 | $re = <>; | |
3211 | chomp $re; | |
3212 | $re = customre::convert $re; | |
3213 | /\Y|$re\Y|/; | |
3214 | ||
83f32aba RS |
3215 | =head2 Embedded Code Execution Frequency |
3216 | ||
57fbbe96 | 3217 | The exact rules for how often C<(??{})> and C<(?{})> are executed in a pattern |
83f32aba RS |
3218 | are unspecified. In the case of a successful match you can assume that |
3219 | they DWIM and will be executed in left to right order the appropriate | |
3220 | number of times in the accepting path of the pattern as would any other | |
3221 | meta-pattern. How non-accepting pathways and match failures affect the | |
3222 | number of times a pattern is executed is specifically unspecified and | |
3223 | may vary depending on what optimizations can be applied to the pattern | |
3224 | and is likely to change from version to version. | |
3225 | ||
3226 | For instance in | |
3227 | ||
3228 | "aaabcdeeeee"=~/a(?{print "a"})b(?{print "b"})cde/; | |
3229 | ||
3230 | the exact number of times "a" or "b" are printed out is unspecified for | |
3231 | failure, but you may assume they will be printed at least once during | |
3232 | a successful match, additionally you may assume that if "b" is printed, | |
3233 | it will be preceded by at least one "a". | |
3234 | ||
3235 | In the case of branching constructs like the following: | |
3236 | ||
3237 | /a(b|(?{ print "a" }))c(?{ print "c" })/; | |
3238 | ||
3239 | you can assume that the input "ac" will output "ac", and that "abc" | |
3240 | will output only "c". | |
3241 | ||
3242 | When embedded code is quantified, successful matches will call the | |
3243 | code once for each matched iteration of the quantifier. For | |
3244 | example: | |
3245 | ||
3246 | "good" =~ /g(?:o(?{print "o"}))*d/; | |
3247 | ||
3248 | will output "o" twice. | |
3249 | ||
0b928c2f | 3250 | =head2 PCRE/Python Support |
1f1031fe | 3251 | |
0b928c2f | 3252 | As of Perl 5.10.0, Perl supports several Python/PCRE-specific extensions |
1f1031fe | 3253 | to the regex syntax. While Perl programmers are encouraged to use the |
0b928c2f | 3254 | Perl-specific syntax, the following are also accepted: |
1f1031fe YO |
3255 | |
3256 | =over 4 | |
3257 | ||
407fecf1 | 3258 | =item C<< (?PE<lt>I<NAME>E<gt>I<pattern>) >> |
1f1031fe | 3259 | |
407fecf1 | 3260 | Define a named capture group. Equivalent to C<< (?<I<NAME>>I<pattern>) >>. |
1f1031fe | 3261 | |
407fecf1 | 3262 | =item C<< (?P=I<NAME>) >> |
1f1031fe | 3263 | |
407fecf1 | 3264 | Backreference to a named capture group. Equivalent to C<< \g{I<NAME>} >>. |
1f1031fe | 3265 | |
407fecf1 | 3266 | =item C<< (?P>I<NAME>) >> |
1f1031fe | 3267 | |
407fecf1 | 3268 | Subroutine call to a named capture group. Equivalent to C<< (?&I<NAME>) >>. |
1f1031fe | 3269 | |
ee9b8eae | 3270 | =back |
1f1031fe | 3271 | |
19799a22 GS |
3272 | =head1 BUGS |
3273 | ||
ed7efc79 | 3274 | There are a number of issues with regard to case-insensitive matching |
57fbbe96 | 3275 | in Unicode rules. See C<"i"> under L</Modifiers> above. |
ed7efc79 | 3276 | |
9da458fc IZ |
3277 | This document varies from difficult to understand to completely |
3278 | and utterly opaque. The wandering prose riddled with jargon is | |
3279 | hard to fathom in several places. | |
3280 | ||
3281 | This document needs a rewrite that separates the tutorial content | |
3282 | from the reference content. | |
19799a22 GS |
3283 | |
3284 | =head1 SEE ALSO | |
9fa51da4 | 3285 | |
c57a3350 KW |
3286 | The syntax of patterns used in Perl pattern matching evolved from those |
3287 | supplied in the Bell Labs Research Unix 8th Edition (Version 8) regex | |
3288 | routines. (The code is actually derived (distantly) from Henry | |
3289 | Spencer's freely redistributable reimplementation of those V8 routines.) | |
3290 | ||
91e0c79e MJD |
3291 | L<perlrequick>. |
3292 | ||
3293 | L<perlretut>. | |
3294 | ||
9b599b2a GS |
3295 | L<perlop/"Regexp Quote-Like Operators">. |
3296 | ||
1e66bd83 PP |
3297 | L<perlop/"Gory details of parsing quoted constructs">. |
3298 | ||
14218588 GS |
3299 | L<perlfaq6>. |
3300 | ||
9b599b2a GS |
3301 | L<perlfunc/pos>. |
3302 | ||
3303 | L<perllocale>. | |
3304 | ||
fb55449c JH |
3305 | L<perlebcdic>. |
3306 | ||
14218588 GS |
3307 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
3308 | by O'Reilly and Associates. |