Commit | Line | Data |
---|---|---|
a0d0e21e LW |
1 | =head1 NAME |
2 | ||
3 | perlre - Perl regular expressions | |
4 | ||
5 | =head1 DESCRIPTION | |
6 | ||
91e0c79e MJD |
7 | This page describes the syntax of regular expressions in Perl. |
8 | ||
cc46d5f2 | 9 | If you haven't used regular expressions before, a quick-start |
91e0c79e MJD |
10 | introduction is available in L<perlrequick>, and a longer tutorial |
11 | introduction is available in L<perlretut>. | |
12 | ||
13 | For reference on how regular expressions are used in matching | |
14 | operations, plus various examples of the same, see discussions of | |
15 | C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like | |
16 | Operators">. | |
cb1a09d0 | 17 | |
19799a22 | 18 | Matching operations can have various modifiers. Modifiers |
5a964f20 | 19 | that relate to the interpretation of the regular expression inside |
19799a22 GS |
20 | are listed below. Modifiers that alter the way a regular expression |
21 | is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and | |
1e66bd83 | 22 | L<perlop/"Gory details of parsing quoted constructs">. |
a0d0e21e | 23 | |
55497cff | 24 | =over 4 |
25 | ||
26 | =item i | |
27 | ||
28 | Do case-insensitive pattern matching. | |
29 | ||
a034a98d DD |
30 | If C<use locale> is in effect, the case map is taken from the current |
31 | locale. See L<perllocale>. | |
32 | ||
54310121 | 33 | =item m |
55497cff | 34 | |
35 | Treat string as multiple lines. That is, change "^" and "$" from matching | |
14218588 | 36 | the start or end of the string to matching the start or end of any |
7f761169 | 37 | line anywhere within the string. |
55497cff | 38 | |
54310121 | 39 | =item s |
55497cff | 40 | |
41 | Treat string as single line. That is, change "." to match any character | |
19799a22 | 42 | whatsoever, even a newline, which normally it would not match. |
55497cff | 43 | |
19799a22 GS |
44 | The C</s> and C</m> modifiers both override the C<$*> setting. That |
45 | is, no matter what C<$*> contains, C</s> without C</m> will force | |
46 | "^" to match only at the beginning of the string and "$" to match | |
47 | only at the end (or just before a newline at the end) of the string. | |
48 | Together, as /ms, they let the "." match any character whatsoever, | |
fb55449c | 49 | while still allowing "^" and "$" to match, respectively, just after |
19799a22 | 50 | and just before newlines within the string. |
7b8d334a | 51 | |
54310121 | 52 | =item x |
55497cff | 53 | |
54 | Extend your pattern's legibility by permitting whitespace and comments. | |
55 | ||
56 | =back | |
a0d0e21e LW |
57 | |
58 | These are usually written as "the C</x> modifier", even though the delimiter | |
14218588 | 59 | in question might not really be a slash. Any of these |
a0d0e21e | 60 | modifiers may also be embedded within the regular expression itself using |
14218588 | 61 | the C<(?...)> construct. See below. |
a0d0e21e | 62 | |
4633a7c4 | 63 | The C</x> modifier itself needs a little more explanation. It tells |
55497cff | 64 | the regular expression parser to ignore whitespace that is neither |
65 | backslashed nor within a character class. You can use this to break up | |
4633a7c4 | 66 | your regular expression into (slightly) more readable parts. The C<#> |
54310121 | 67 | character is also treated as a metacharacter introducing a comment, |
55497cff | 68 | just as in ordinary Perl code. This also means that if you want real |
14218588 | 69 | whitespace or C<#> characters in the pattern (outside a character |
5a964f20 | 70 | class, where they are unaffected by C</x>), that you'll either have to |
55497cff | 71 | escape them or encode them using octal or hex escapes. Taken together, |
72 | these features go a long way towards making Perl's regular expressions | |
0c815be9 HS |
73 | more readable. Note that you have to be careful not to include the |
74 | pattern delimiter in the comment--perl has no way of knowing you did | |
5a964f20 | 75 | not intend to close the pattern early. See the C-comment deletion code |
0c815be9 | 76 | in L<perlop>. |
a0d0e21e LW |
77 | |
78 | =head2 Regular Expressions | |
79 | ||
19799a22 | 80 | The patterns used in Perl pattern matching derive from supplied in |
14218588 | 81 | the Version 8 regex routines. (The routines are derived |
19799a22 GS |
82 | (distantly) from Henry Spencer's freely redistributable reimplementation |
83 | of the V8 routines.) See L<Version 8 Regular Expressions> for | |
84 | details. | |
a0d0e21e LW |
85 | |
86 | In particular the following metacharacters have their standard I<egrep>-ish | |
87 | meanings: | |
88 | ||
54310121 | 89 | \ Quote the next metacharacter |
a0d0e21e LW |
90 | ^ Match the beginning of the line |
91 | . Match any character (except newline) | |
c07a80fd | 92 | $ Match the end of the line (or before newline at the end) |
a0d0e21e LW |
93 | | Alternation |
94 | () Grouping | |
95 | [] Character class | |
96 | ||
14218588 GS |
97 | By default, the "^" character is guaranteed to match only the |
98 | beginning of the string, the "$" character only the end (or before the | |
99 | newline at the end), and Perl does certain optimizations with the | |
a0d0e21e LW |
100 | assumption that the string contains only one line. Embedded newlines |
101 | will not be matched by "^" or "$". You may, however, wish to treat a | |
4a6725af | 102 | string as a multi-line buffer, such that the "^" will match after any |
a0d0e21e LW |
103 | newline within the string, and "$" will match before any newline. At the |
104 | cost of a little more overhead, you can do this by using the /m modifier | |
105 | on the pattern match operator. (Older programs did this by setting C<$*>, | |
5f05dabc | 106 | but this practice is now deprecated.) |
a0d0e21e | 107 | |
14218588 | 108 | To simplify multi-line substitutions, the "." character never matches a |
55497cff | 109 | newline unless you use the C</s> modifier, which in effect tells Perl to pretend |
a0d0e21e LW |
110 | the string is a single line--even if it isn't. The C</s> modifier also |
111 | overrides the setting of C<$*>, in case you have some (badly behaved) older | |
112 | code that sets it in another module. | |
113 | ||
114 | The following standard quantifiers are recognized: | |
115 | ||
116 | * Match 0 or more times | |
117 | + Match 1 or more times | |
118 | ? Match 1 or 0 times | |
119 | {n} Match exactly n times | |
120 | {n,} Match at least n times | |
121 | {n,m} Match at least n but not more than m times | |
122 | ||
123 | (If a curly bracket occurs in any other context, it is treated | |
b975c076 JL |
124 | as a regular character. In particular, the lower bound |
125 | is not optional.) The "*" modifier is equivalent to C<{0,}>, the "+" | |
25f94b33 | 126 | modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited |
9c79236d GS |
127 | to integral values less than a preset limit defined when perl is built. |
128 | This is usually 32766 on the most common platforms. The actual limit can | |
129 | be seen in the error message generated by code such as this: | |
130 | ||
820475bd | 131 | $_ **= $_ , / {$_} / for 2 .. 42; |
a0d0e21e | 132 | |
54310121 | 133 | By default, a quantified subpattern is "greedy", that is, it will match as |
134 | many times as possible (given a particular starting location) while still | |
135 | allowing the rest of the pattern to match. If you want it to match the | |
136 | minimum number of times possible, follow the quantifier with a "?". Note | |
137 | that the meanings don't change, just the "greediness": | |
a0d0e21e LW |
138 | |
139 | *? Match 0 or more times | |
140 | +? Match 1 or more times | |
141 | ?? Match 0 or 1 time | |
142 | {n}? Match exactly n times | |
143 | {n,}? Match at least n times | |
144 | {n,m}? Match at least n but not more than m times | |
145 | ||
5f05dabc | 146 | Because patterns are processed as double quoted strings, the following |
a0d0e21e LW |
147 | also work: |
148 | ||
0f36ee90 | 149 | \t tab (HT, TAB) |
150 | \n newline (LF, NL) | |
151 | \r return (CR) | |
152 | \f form feed (FF) | |
153 | \a alarm (bell) (BEL) | |
154 | \e escape (think troff) (ESC) | |
cb1a09d0 AD |
155 | \033 octal char (think of a PDP-11) |
156 | \x1B hex char | |
a0ed51b3 | 157 | \x{263a} wide hex char (Unicode SMILEY) |
a0d0e21e | 158 | \c[ control char |
4a2d328f | 159 | \N{name} named char |
cb1a09d0 AD |
160 | \l lowercase next char (think vi) |
161 | \u uppercase next char (think vi) | |
162 | \L lowercase till \E (think vi) | |
163 | \U uppercase till \E (think vi) | |
164 | \E end case modification (think vi) | |
5a964f20 | 165 | \Q quote (disable) pattern metacharacters till \E |
a0d0e21e | 166 | |
a034a98d | 167 | If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> |
423cee85 | 168 | and C<\U> is taken from the current locale. See L<perllocale>. For |
4a2d328f | 169 | documentation of C<\N{name}>, see L<charnames>. |
a034a98d | 170 | |
1d2dff63 GS |
171 | You cannot include a literal C<$> or C<@> within a C<\Q> sequence. |
172 | An unescaped C<$> or C<@> interpolates the corresponding variable, | |
173 | while escaping will cause the literal string C<\$> to be matched. | |
174 | You'll need to write something like C<m/\Quser\E\@\Qhost/>. | |
175 | ||
a0d0e21e LW |
176 | In addition, Perl defines the following: |
177 | ||
178 | \w Match a "word" character (alphanumeric plus "_") | |
36bbe248 | 179 | \W Match a non-"word" character |
a0d0e21e LW |
180 | \s Match a whitespace character |
181 | \S Match a non-whitespace character | |
182 | \d Match a digit character | |
183 | \D Match a non-digit character | |
a0ed51b3 LW |
184 | \pP Match P, named property. Use \p{Prop} for longer names. |
185 | \PP Match non-P | |
f244e06d | 186 | \X Match eXtended Unicode "combining character sequence", |
07698885 | 187 | equivalent to (?:\PM\pM*) |
72ff2908 | 188 | \C Match a single C char (octet) even under Unicode. |
07698885 | 189 | NOTE: breaks up characters into their UTF-8 bytes, |
72ff2908 | 190 | so you may end up with malformed pieces of UTF-8. |
f14c76ed | 191 | Unsupported in lookbehind. |
a0d0e21e | 192 | |
08ce8fc6 JH |
193 | A C<\w> matches a single alphanumeric character (an alphabetic |
194 | character, or a decimal digit) or C<_>, not a whole word. Use C<\w+> | |
195 | to match a string of Perl-identifier characters (which isn't the same | |
196 | as matching an English word). If C<use locale> is in effect, the list | |
197 | of alphabetic characters generated by C<\w> is taken from the current | |
198 | locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, | |
1209ba90 | 199 | C<\d>, and C<\D> within character classes, but if you try to use them |
08ce8fc6 JH |
200 | as endpoints of a range, that's not a range, the "-" is understood |
201 | literally. If Unicode is in effect, C<\s> matches also "\x{85}", | |
202 | "\x{2028}, and "\x{2029}", see L<perlunicode> for more details about | |
491fd90a JH |
203 | C<\pP>, C<\PP>, and C<\X>, and L<perluniintro> about Unicode in general. |
204 | You can define your own C<\p> and C<\P> propreties, see L<perlunicode>. | |
a0d0e21e | 205 | |
b8c5462f JH |
206 | The POSIX character class syntax |
207 | ||
820475bd | 208 | [:class:] |
b8c5462f | 209 | |
26b44a0a JH |
210 | is also available. The available classes and their backslash |
211 | equivalents (if available) are as follows: | |
b8c5462f JH |
212 | |
213 | alpha | |
214 | alnum | |
215 | ascii | |
aaa51d5e | 216 | blank [1] |
b8c5462f JH |
217 | cntrl |
218 | digit \d | |
219 | graph | |
220 | lower | |
221 | ||
222 | punct | |
aaa51d5e | 223 | space \s [2] |
b8c5462f | 224 | upper |
aaa51d5e | 225 | word \w [3] |
b8c5462f JH |
226 | xdigit |
227 | ||
07698885 RGS |
228 | =over |
229 | ||
230 | =item [1] | |
231 | ||
232 | A GNU extension equivalent to C<[ \t]>, `all horizontal whitespace'. | |
233 | ||
234 | =item [2] | |
235 | ||
236 | Not exactly equivalent to C<\s> since the C<[[:space:]]> includes | |
08ce8fc6 | 237 | also the (very rare) `vertical tabulator', "\ck", chr(11). |
07698885 RGS |
238 | |
239 | =item [3] | |
240 | ||
08ce8fc6 | 241 | A Perl extension, see above. |
07698885 RGS |
242 | |
243 | =back | |
aaa51d5e | 244 | |
26b44a0a | 245 | For example use C<[:upper:]> to match all the uppercase characters. |
aaa51d5e JF |
246 | Note that the C<[]> are part of the C<[::]> construct, not part of the |
247 | whole character class. For example: | |
b8c5462f | 248 | |
820475bd | 249 | [01[:alpha:]%] |
b8c5462f | 250 | |
593df60c | 251 | matches zero, one, any alphabetic character, and the percentage sign. |
b8c5462f | 252 | |
72ff2908 JH |
253 | The following equivalences to Unicode \p{} constructs and equivalent |
254 | backslash character classes (if available), will hold: | |
255 | ||
256 | [:...:] \p{...} backslash | |
b8c5462f JH |
257 | |
258 | alpha IsAlpha | |
259 | alnum IsAlnum | |
260 | ascii IsASCII | |
aaa51d5e | 261 | blank IsSpace |
b8c5462f | 262 | cntrl IsCntrl |
3bec3564 | 263 | digit IsDigit \d |
b8c5462f JH |
264 | graph IsGraph |
265 | lower IsLower | |
266 | print IsPrint | |
267 | punct IsPunct | |
268 | space IsSpace | |
3bec3564 | 269 | IsSpacePerl \s |
b8c5462f JH |
270 | upper IsUpper |
271 | word IsWord | |
272 | xdigit IsXDigit | |
273 | ||
26b44a0a | 274 | For example C<[:lower:]> and C<\p{IsLower}> are equivalent. |
b8c5462f JH |
275 | |
276 | If the C<utf8> pragma is not used but the C<locale> pragma is, the | |
aaa51d5e JF |
277 | classes correlate with the usual isalpha(3) interface (except for |
278 | `word' and `blank'). | |
b8c5462f JH |
279 | |
280 | The assumedly non-obviously named classes are: | |
281 | ||
282 | =over 4 | |
283 | ||
284 | =item cntrl | |
285 | ||
820475bd GS |
286 | Any control character. Usually characters that don't produce output as |
287 | such but instead control the terminal somehow: for example newline and | |
288 | backspace are control characters. All characters with ord() less than | |
593df60c | 289 | 32 are most often classified as control characters (assuming ASCII, |
7be5a6cf JF |
290 | the ISO Latin character sets, and Unicode), as is the character with |
291 | the ord() value of 127 (C<DEL>). | |
b8c5462f JH |
292 | |
293 | =item graph | |
294 | ||
f1cbbd6e | 295 | Any alphanumeric or punctuation (special) character. |
b8c5462f JH |
296 | |
297 | =item print | |
298 | ||
f79b3095 | 299 | Any alphanumeric or punctuation (special) character or the space character. |
b8c5462f JH |
300 | |
301 | =item punct | |
302 | ||
f1cbbd6e | 303 | Any punctuation (special) character. |
b8c5462f JH |
304 | |
305 | =item xdigit | |
306 | ||
593df60c | 307 | Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would |
820475bd | 308 | work just fine) it is included for completeness. |
b8c5462f | 309 | |
b8c5462f JH |
310 | =back |
311 | ||
312 | You can negate the [::] character classes by prefixing the class name | |
313 | with a '^'. This is a Perl extension. For example: | |
314 | ||
72ff2908 | 315 | POSIX traditional Unicode |
93733859 JH |
316 | |
317 | [:^digit:] \D \P{IsDigit} | |
318 | [:^space:] \S \P{IsSpace} | |
319 | [:^word:] \W \P{IsWord} | |
b8c5462f | 320 | |
54c18d04 MK |
321 | Perl respects the POSIX standard in that POSIX character classes are |
322 | only supported within a character class. The POSIX character classes | |
323 | [.cc.] and [=cc=] are recognized but B<not> supported and trying to | |
324 | use them will cause an error. | |
b8c5462f | 325 | |
a0d0e21e LW |
326 | Perl defines the following zero-width assertions: |
327 | ||
328 | \b Match a word boundary | |
329 | \B Match a non-(word boundary) | |
b85d18e9 IZ |
330 | \A Match only at beginning of string |
331 | \Z Match only at end of string, or before newline at the end | |
332 | \z Match only at end of string | |
9da458fc IZ |
333 | \G Match only at pos() (e.g. at the end-of-match position |
334 | of prior m//g) | |
a0d0e21e | 335 | |
14218588 | 336 | A word boundary (C<\b>) is a spot between two characters |
19799a22 GS |
337 | that has a C<\w> on one side of it and a C<\W> on the other side |
338 | of it (in either order), counting the imaginary characters off the | |
339 | beginning and end of the string as matching a C<\W>. (Within | |
340 | character classes C<\b> represents backspace rather than a word | |
341 | boundary, just as it normally does in any double-quoted string.) | |
342 | The C<\A> and C<\Z> are just like "^" and "$", except that they | |
343 | won't match multiple times when the C</m> modifier is used, while | |
344 | "^" and "$" will match at every internal line boundary. To match | |
345 | the actual end of the string and not ignore an optional trailing | |
346 | newline, use C<\z>. | |
347 | ||
348 | The C<\G> assertion can be used to chain global matches (using | |
349 | C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. | |
350 | It is also useful when writing C<lex>-like scanners, when you have | |
351 | several patterns that you want to match against consequent substrings | |
352 | of your string, see the previous reference. The actual location | |
353 | where C<\G> will match can also be influenced by using C<pos()> as | |
25cf8c22 HS |
354 | an lvalue: see L<perlfunc/pos>. Currently C<\G> is only fully |
355 | supported when anchored to the start of the pattern; while it | |
356 | is permitted to use it elsewhere, as in C</(?<=\G..)./g>, some | |
357 | such uses (C</.\G/g>, for example) currently cause problems, and | |
358 | it is recommended that you avoid such usage for now. | |
c47ff5f1 | 359 | |
14218588 | 360 | The bracketing construct C<( ... )> creates capture buffers. To |
c47ff5f1 | 361 | refer to the digit'th buffer use \<digit> within the |
14218588 | 362 | match. Outside the match use "$" instead of "\". (The |
c47ff5f1 | 363 | \<digit> notation works in certain circumstances outside |
14218588 GS |
364 | the match. See the warning below about \1 vs $1 for details.) |
365 | Referring back to another part of the match is called a | |
366 | I<backreference>. | |
367 | ||
368 | There is no limit to the number of captured substrings that you may | |
369 | use. However Perl also uses \10, \11, etc. as aliases for \010, | |
fb55449c JH |
370 | \011, etc. (Recall that 0 means octal, so \011 is the character at |
371 | number 9 in your coded character set; which would be the 10th character, | |
372 | a horizontal tab under ASCII.) Perl resolves this | |
373 | ambiguity by interpreting \10 as a backreference only if at least 10 | |
374 | left parentheses have opened before it. Likewise \11 is a | |
375 | backreference only if at least 11 left parentheses have opened | |
376 | before it. And so on. \1 through \9 are always interpreted as | |
377 | backreferences. | |
14218588 GS |
378 | |
379 | Examples: | |
a0d0e21e LW |
380 | |
381 | s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words | |
382 | ||
14218588 GS |
383 | if (/(.)\1/) { # find first doubled char |
384 | print "'$1' is the first doubled character\n"; | |
385 | } | |
c47ff5f1 | 386 | |
14218588 | 387 | if (/Time: (..):(..):(..)/) { # parse out values |
a0d0e21e LW |
388 | $hours = $1; |
389 | $minutes = $2; | |
390 | $seconds = $3; | |
391 | } | |
c47ff5f1 | 392 | |
14218588 GS |
393 | Several special variables also refer back to portions of the previous |
394 | match. C<$+> returns whatever the last bracket match matched. | |
395 | C<$&> returns the entire matched string. (At one point C<$0> did | |
396 | also, but now it returns the name of the program.) C<$`> returns | |
77ea4f6d JV |
397 | everything before the matched string. C<$'> returns everything |
398 | after the matched string. And C<$^N> contains whatever was matched by | |
399 | the most-recently closed group (submatch). C<$^N> can be used in | |
400 | extended patterns (see below), for example to assign a submatch to a | |
401 | variable. | |
14218588 GS |
402 | |
403 | The numbered variables ($1, $2, $3, etc.) and the related punctuation | |
77ea4f6d | 404 | set (C<$+>, C<$&>, C<$`>, C<$'>, and C<$^N>) are all dynamically scoped |
14218588 GS |
405 | until the end of the enclosing block or until the next successful |
406 | match, whichever comes first. (See L<perlsyn/"Compound Statements">.) | |
407 | ||
408 | B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or | |
409 | C<$'> anywhere in the program, it has to provide them for every | |
410 | pattern match. This may substantially slow your program. Perl | |
411 | uses the same mechanism to produce $1, $2, etc, so you also pay a | |
412 | price for each pattern that contains capturing parentheses. (To | |
413 | avoid this cost while retaining the grouping behaviour, use the | |
414 | extended regular expression C<(?: ... )> instead.) But if you never | |
415 | use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing | |
416 | parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> | |
417 | if you can, but if you can't (and some algorithms really appreciate | |
418 | them), once you've used them once, use them at will, because you've | |
419 | already paid the price. As of 5.005, C<$&> is not so costly as the | |
420 | other two. | |
68dc0745 | 421 | |
19799a22 GS |
422 | Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, |
423 | C<\w>, C<\n>. Unlike some other regular expression languages, there | |
424 | are no backslashed symbols that aren't alphanumeric. So anything | |
c47ff5f1 | 425 | that looks like \\, \(, \), \<, \>, \{, or \} is always |
19799a22 GS |
426 | interpreted as a literal character, not a metacharacter. This was |
427 | once used in a common idiom to disable or quote the special meanings | |
428 | of regular expression metacharacters in a string that you want to | |
36bbe248 | 429 | use for a pattern. Simply quote all non-"word" characters: |
a0d0e21e LW |
430 | |
431 | $pattern =~ s/(\W)/\\$1/g; | |
432 | ||
f1cbbd6e | 433 | (If C<use locale> is set, then this depends on the current locale.) |
14218588 GS |
434 | Today it is more common to use the quotemeta() function or the C<\Q> |
435 | metaquoting escape sequence to disable all metacharacters' special | |
436 | meanings like this: | |
a0d0e21e LW |
437 | |
438 | /$unquoted\Q$quoted\E$unquoted/ | |
439 | ||
9da458fc IZ |
440 | Beware that if you put literal backslashes (those not inside |
441 | interpolated variables) between C<\Q> and C<\E>, double-quotish | |
442 | backslash interpolation may lead to confusing results. If you | |
443 | I<need> to use literal backslashes within C<\Q...\E>, | |
444 | consult L<perlop/"Gory details of parsing quoted constructs">. | |
445 | ||
19799a22 GS |
446 | =head2 Extended Patterns |
447 | ||
14218588 GS |
448 | Perl also defines a consistent extension syntax for features not |
449 | found in standard tools like B<awk> and B<lex>. The syntax is a | |
450 | pair of parentheses with a question mark as the first thing within | |
451 | the parentheses. The character after the question mark indicates | |
452 | the extension. | |
19799a22 | 453 | |
14218588 GS |
454 | The stability of these extensions varies widely. Some have been |
455 | part of the core language for many years. Others are experimental | |
456 | and may change without warning or be completely removed. Check | |
457 | the documentation on an individual feature to verify its current | |
458 | status. | |
19799a22 | 459 | |
14218588 GS |
460 | A question mark was chosen for this and for the minimal-matching |
461 | construct because 1) question marks are rare in older regular | |
462 | expressions, and 2) whenever you see one, you should stop and | |
463 | "question" exactly what is going on. That's psychology... | |
a0d0e21e LW |
464 | |
465 | =over 10 | |
466 | ||
cc6b7395 | 467 | =item C<(?#text)> |
a0d0e21e | 468 | |
14218588 | 469 | A comment. The text is ignored. If the C</x> modifier enables |
19799a22 | 470 | whitespace formatting, a simple C<#> will suffice. Note that Perl closes |
259138e3 GS |
471 | the comment as soon as it sees a C<)>, so there is no way to put a literal |
472 | C<)> in the comment. | |
a0d0e21e | 473 | |
19799a22 GS |
474 | =item C<(?imsx-imsx)> |
475 | ||
0b6d1084 JH |
476 | One or more embedded pattern-match modifiers, to be turned on (or |
477 | turned off, if preceded by C<->) for the remainder of the pattern or | |
478 | the remainder of the enclosing pattern group (if any). This is | |
479 | particularly useful for dynamic patterns, such as those read in from a | |
480 | configuration file, read in as an argument, are specified in a table | |
481 | somewhere, etc. Consider the case that some of which want to be case | |
482 | sensitive and some do not. The case insensitive ones need to include | |
483 | merely C<(?i)> at the front of the pattern. For example: | |
19799a22 GS |
484 | |
485 | $pattern = "foobar"; | |
486 | if ( /$pattern/i ) { } | |
487 | ||
488 | # more flexible: | |
489 | ||
490 | $pattern = "(?i)foobar"; | |
491 | if ( /$pattern/ ) { } | |
492 | ||
0b6d1084 | 493 | These modifiers are restored at the end of the enclosing group. For example, |
19799a22 GS |
494 | |
495 | ( (?i) blah ) \s+ \1 | |
496 | ||
497 | will match a repeated (I<including the case>!) word C<blah> in any | |
14218588 | 498 | case, assuming C<x> modifier, and no C<i> modifier outside this |
19799a22 GS |
499 | group. |
500 | ||
5a964f20 | 501 | =item C<(?:pattern)> |
a0d0e21e | 502 | |
ca9dfc88 IZ |
503 | =item C<(?imsx-imsx:pattern)> |
504 | ||
5a964f20 TC |
505 | This is for clustering, not capturing; it groups subexpressions like |
506 | "()", but doesn't make backreferences as "()" does. So | |
a0d0e21e | 507 | |
5a964f20 | 508 | @fields = split(/\b(?:a|b|c)\b/) |
a0d0e21e LW |
509 | |
510 | is like | |
511 | ||
5a964f20 | 512 | @fields = split(/\b(a|b|c)\b/) |
a0d0e21e | 513 | |
19799a22 GS |
514 | but doesn't spit out extra fields. It's also cheaper not to capture |
515 | characters if you don't need to. | |
a0d0e21e | 516 | |
19799a22 GS |
517 | Any letters between C<?> and C<:> act as flags modifiers as with |
518 | C<(?imsx-imsx)>. For example, | |
ca9dfc88 IZ |
519 | |
520 | /(?s-i:more.*than).*million/i | |
521 | ||
14218588 | 522 | is equivalent to the more verbose |
ca9dfc88 IZ |
523 | |
524 | /(?:(?s-i)more.*than).*million/i | |
525 | ||
5a964f20 | 526 | =item C<(?=pattern)> |
a0d0e21e | 527 | |
19799a22 | 528 | A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> |
a0d0e21e LW |
529 | matches a word followed by a tab, without including the tab in C<$&>. |
530 | ||
5a964f20 | 531 | =item C<(?!pattern)> |
a0d0e21e | 532 | |
19799a22 | 533 | A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> |
a0d0e21e | 534 | matches any occurrence of "foo" that isn't followed by "bar". Note |
19799a22 GS |
535 | however that look-ahead and look-behind are NOT the same thing. You cannot |
536 | use this for look-behind. | |
7b8d334a | 537 | |
5a964f20 | 538 | If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> |
7b8d334a GS |
539 | will not do what you want. That's because the C<(?!foo)> is just saying that |
540 | the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will | |
541 | match. You would have to do something like C</(?!foo)...bar/> for that. We | |
542 | say "like" because there's the case of your "bar" not having three characters | |
543 | before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. | |
544 | Sometimes it's still easier just to say: | |
a0d0e21e | 545 | |
a3cb178b | 546 | if (/bar/ && $` !~ /foo$/) |
a0d0e21e | 547 | |
19799a22 | 548 | For look-behind see below. |
c277df42 | 549 | |
c47ff5f1 | 550 | =item C<(?<=pattern)> |
c277df42 | 551 | |
c47ff5f1 | 552 | A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> |
19799a22 GS |
553 | matches a word that follows a tab, without including the tab in C<$&>. |
554 | Works only for fixed-width look-behind. | |
c277df42 | 555 | |
5a964f20 | 556 | =item C<(?<!pattern)> |
c277df42 | 557 | |
19799a22 GS |
558 | A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> |
559 | matches any occurrence of "foo" that does not follow "bar". Works | |
560 | only for fixed-width look-behind. | |
c277df42 | 561 | |
cc6b7395 | 562 | =item C<(?{ code })> |
c277df42 | 563 | |
19799a22 GS |
564 | B<WARNING>: This extended regular expression feature is considered |
565 | highly experimental, and may be changed or deleted without notice. | |
c277df42 | 566 | |
cc46d5f2 | 567 | This zero-width assertion evaluates any embedded Perl code. It |
19799a22 GS |
568 | always succeeds, and its C<code> is not interpolated. Currently, |
569 | the rules to determine where the C<code> ends are somewhat convoluted. | |
570 | ||
77ea4f6d JV |
571 | This feature can be used together with the special variable C<$^N> to |
572 | capture the results of submatches in variables without having to keep | |
573 | track of the number of nested parentheses. For example: | |
574 | ||
575 | $_ = "The brown fox jumps over the lazy dog"; | |
576 | /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i; | |
577 | print "color = $color, animal = $animal\n"; | |
578 | ||
19799a22 GS |
579 | The C<code> is properly scoped in the following sense: If the assertion |
580 | is backtracked (compare L<"Backtracking">), all changes introduced after | |
581 | C<local>ization are undone, so that | |
b9ac3b5b GS |
582 | |
583 | $_ = 'a' x 8; | |
584 | m< | |
585 | (?{ $cnt = 0 }) # Initialize $cnt. | |
586 | ( | |
587 | a | |
588 | (?{ | |
589 | local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. | |
590 | }) | |
591 | )* | |
592 | aaaa | |
593 | (?{ $res = $cnt }) # On success copy to non-localized | |
594 | # location. | |
595 | >x; | |
596 | ||
19799a22 | 597 | will set C<$res = 4>. Note that after the match, $cnt returns to the globally |
14218588 | 598 | introduced value, because the scopes that restrict C<local> operators |
b9ac3b5b GS |
599 | are unwound. |
600 | ||
19799a22 GS |
601 | This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> |
602 | switch. If I<not> used in this way, the result of evaluation of | |
603 | C<code> is put into the special variable C<$^R>. This happens | |
604 | immediately, so C<$^R> can be used from other C<(?{ code })> assertions | |
605 | inside the same regular expression. | |
b9ac3b5b | 606 | |
19799a22 GS |
607 | The assignment to C<$^R> above is properly localized, so the old |
608 | value of C<$^R> is restored if the assertion is backtracked; compare | |
609 | L<"Backtracking">. | |
b9ac3b5b | 610 | |
19799a22 GS |
611 | For reasons of security, this construct is forbidden if the regular |
612 | expression involves run-time interpolation of variables, unless the | |
613 | perilous C<use re 'eval'> pragma has been used (see L<re>), or the | |
614 | variables contain results of C<qr//> operator (see | |
615 | L<perlop/"qr/STRING/imosx">). | |
871b0233 | 616 | |
14218588 | 617 | This restriction is because of the wide-spread and remarkably convenient |
19799a22 | 618 | custom of using run-time determined strings as patterns. For example: |
871b0233 IZ |
619 | |
620 | $re = <>; | |
621 | chomp $re; | |
622 | $string =~ /$re/; | |
623 | ||
14218588 GS |
624 | Before Perl knew how to execute interpolated code within a pattern, |
625 | this operation was completely safe from a security point of view, | |
626 | although it could raise an exception from an illegal pattern. If | |
627 | you turn on the C<use re 'eval'>, though, it is no longer secure, | |
628 | so you should only do so if you are also using taint checking. | |
629 | Better yet, use the carefully constrained evaluation within a Safe | |
cc46d5f2 | 630 | compartment. See L<perlsec> for details about both these mechanisms. |
871b0233 | 631 | |
14455d6c | 632 | =item C<(??{ code })> |
0f5d15d6 | 633 | |
19799a22 GS |
634 | B<WARNING>: This extended regular expression feature is considered |
635 | highly experimental, and may be changed or deleted without notice. | |
9da458fc IZ |
636 | A simplified version of the syntax may be introduced for commonly |
637 | used idioms. | |
0f5d15d6 | 638 | |
19799a22 GS |
639 | This is a "postponed" regular subexpression. The C<code> is evaluated |
640 | at run time, at the moment this subexpression may match. The result | |
641 | of evaluation is considered as a regular expression and matched as | |
642 | if it were inserted instead of this construct. | |
0f5d15d6 | 643 | |
428594d9 | 644 | The C<code> is not interpolated. As before, the rules to determine |
19799a22 GS |
645 | where the C<code> ends are currently somewhat convoluted. |
646 | ||
647 | The following pattern matches a parenthesized group: | |
0f5d15d6 IZ |
648 | |
649 | $re = qr{ | |
650 | \( | |
651 | (?: | |
652 | (?> [^()]+ ) # Non-parens without backtracking | |
653 | | | |
14455d6c | 654 | (??{ $re }) # Group with matching parens |
0f5d15d6 IZ |
655 | )* |
656 | \) | |
657 | }x; | |
658 | ||
c47ff5f1 | 659 | =item C<< (?>pattern) >> |
5a964f20 | 660 | |
19799a22 GS |
661 | B<WARNING>: This extended regular expression feature is considered |
662 | highly experimental, and may be changed or deleted without notice. | |
663 | ||
664 | An "independent" subexpression, one which matches the substring | |
665 | that a I<standalone> C<pattern> would match if anchored at the given | |
9da458fc | 666 | position, and it matches I<nothing other than this substring>. This |
19799a22 GS |
667 | construct is useful for optimizations of what would otherwise be |
668 | "eternal" matches, because it will not backtrack (see L<"Backtracking">). | |
9da458fc IZ |
669 | It may also be useful in places where the "grab all you can, and do not |
670 | give anything back" semantic is desirable. | |
19799a22 | 671 | |
c47ff5f1 | 672 | For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> |
19799a22 GS |
673 | (anchored at the beginning of string, as above) will match I<all> |
674 | characters C<a> at the beginning of string, leaving no C<a> for | |
675 | C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, | |
676 | since the match of the subgroup C<a*> is influenced by the following | |
677 | group C<ab> (see L<"Backtracking">). In particular, C<a*> inside | |
678 | C<a*ab> will match fewer characters than a standalone C<a*>, since | |
679 | this makes the tail match. | |
680 | ||
c47ff5f1 | 681 | An effect similar to C<< (?>pattern) >> may be achieved by writing |
19799a22 GS |
682 | C<(?=(pattern))\1>. This matches the same substring as a standalone |
683 | C<a+>, and the following C<\1> eats the matched string; it therefore | |
c47ff5f1 | 684 | makes a zero-length assertion into an analogue of C<< (?>...) >>. |
19799a22 GS |
685 | (The difference between these two constructs is that the second one |
686 | uses a capturing group, thus shifting ordinals of backreferences | |
687 | in the rest of a regular expression.) | |
688 | ||
689 | Consider this pattern: | |
c277df42 | 690 | |
871b0233 IZ |
691 | m{ \( |
692 | ( | |
9da458fc | 693 | [^()]+ # x+ |
871b0233 IZ |
694 | | |
695 | \( [^()]* \) | |
696 | )+ | |
697 | \) | |
698 | }x | |
5a964f20 | 699 | |
19799a22 GS |
700 | That will efficiently match a nonempty group with matching parentheses |
701 | two levels deep or less. However, if there is no such group, it | |
702 | will take virtually forever on a long string. That's because there | |
703 | are so many different ways to split a long string into several | |
704 | substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar | |
705 | to a subpattern of the above pattern. Consider how the pattern | |
706 | above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several | |
707 | seconds, but that each extra letter doubles this time. This | |
708 | exponential performance will make it appear that your program has | |
14218588 | 709 | hung. However, a tiny change to this pattern |
5a964f20 | 710 | |
871b0233 IZ |
711 | m{ \( |
712 | ( | |
9da458fc | 713 | (?> [^()]+ ) # change x+ above to (?> x+ ) |
871b0233 IZ |
714 | | |
715 | \( [^()]* \) | |
716 | )+ | |
717 | \) | |
718 | }x | |
c277df42 | 719 | |
c47ff5f1 | 720 | which uses C<< (?>...) >> matches exactly when the one above does (verifying |
5a964f20 TC |
721 | this yourself would be a productive exercise), but finishes in a fourth |
722 | the time when used on a similar string with 1000000 C<a>s. Be aware, | |
723 | however, that this pattern currently triggers a warning message under | |
9f1b1f2d | 724 | the C<use warnings> pragma or B<-w> switch saying it |
6bab786b | 725 | C<"matches null string many times in regex">. |
c277df42 | 726 | |
c47ff5f1 | 727 | On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable |
19799a22 | 728 | effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. |
c277df42 IZ |
729 | This was only 4 times slower on a string with 1000000 C<a>s. |
730 | ||
9da458fc IZ |
731 | The "grab all you can, and do not give anything back" semantic is desirable |
732 | in many situations where on the first sight a simple C<()*> looks like | |
733 | the correct solution. Suppose we parse text with comments being delimited | |
734 | by C<#> followed by some optional (horizontal) whitespace. Contrary to | |
4375e838 | 735 | its appearance, C<#[ \t]*> I<is not> the correct subexpression to match |
9da458fc IZ |
736 | the comment delimiter, because it may "give up" some whitespace if |
737 | the remainder of the pattern can be made to match that way. The correct | |
738 | answer is either one of these: | |
739 | ||
740 | (?>#[ \t]*) | |
741 | #[ \t]*(?![ \t]) | |
742 | ||
743 | For example, to grab non-empty comments into $1, one should use either | |
744 | one of these: | |
745 | ||
746 | / (?> \# [ \t]* ) ( .+ ) /x; | |
747 | / \# [ \t]* ( [^ \t] .* ) /x; | |
748 | ||
749 | Which one you pick depends on which of these expressions better reflects | |
750 | the above specification of comments. | |
751 | ||
5a964f20 | 752 | =item C<(?(condition)yes-pattern|no-pattern)> |
c277df42 | 753 | |
5a964f20 | 754 | =item C<(?(condition)yes-pattern)> |
c277df42 | 755 | |
19799a22 GS |
756 | B<WARNING>: This extended regular expression feature is considered |
757 | highly experimental, and may be changed or deleted without notice. | |
758 | ||
c277df42 IZ |
759 | Conditional expression. C<(condition)> should be either an integer in |
760 | parentheses (which is valid if the corresponding pair of parentheses | |
19799a22 | 761 | matched), or look-ahead/look-behind/evaluate zero-width assertion. |
c277df42 | 762 | |
19799a22 | 763 | For example: |
c277df42 | 764 | |
5a964f20 | 765 | m{ ( \( )? |
871b0233 | 766 | [^()]+ |
5a964f20 | 767 | (?(1) \) ) |
871b0233 | 768 | }x |
c277df42 IZ |
769 | |
770 | matches a chunk of non-parentheses, possibly included in parentheses | |
771 | themselves. | |
a0d0e21e | 772 | |
a0d0e21e LW |
773 | =back |
774 | ||
c07a80fd | 775 | =head2 Backtracking |
776 | ||
35a734be IZ |
777 | NOTE: This section presents an abstract approximation of regular |
778 | expression behavior. For a more rigorous (and complicated) view of | |
779 | the rules involved in selecting a match among possible alternatives, | |
780 | see L<Combining pieces together>. | |
781 | ||
c277df42 | 782 | A fundamental feature of regular expression matching involves the |
5a964f20 | 783 | notion called I<backtracking>, which is currently used (when needed) |
c277df42 | 784 | by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, |
9da458fc IZ |
785 | C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized |
786 | internally, but the general principle outlined here is valid. | |
c07a80fd | 787 | |
788 | For a regular expression to match, the I<entire> regular expression must | |
789 | match, not just part of it. So if the beginning of a pattern containing a | |
790 | quantifier succeeds in a way that causes later parts in the pattern to | |
791 | fail, the matching engine backs up and recalculates the beginning | |
792 | part--that's why it's called backtracking. | |
793 | ||
794 | Here is an example of backtracking: Let's say you want to find the | |
795 | word following "foo" in the string "Food is on the foo table.": | |
796 | ||
797 | $_ = "Food is on the foo table."; | |
798 | if ( /\b(foo)\s+(\w+)/i ) { | |
799 | print "$2 follows $1.\n"; | |
800 | } | |
801 | ||
802 | When the match runs, the first part of the regular expression (C<\b(foo)>) | |
803 | finds a possible match right at the beginning of the string, and loads up | |
804 | $1 with "Foo". However, as soon as the matching engine sees that there's | |
805 | no whitespace following the "Foo" that it had saved in $1, it realizes its | |
68dc0745 | 806 | mistake and starts over again one character after where it had the |
c07a80fd | 807 | tentative match. This time it goes all the way until the next occurrence |
808 | of "foo". The complete regular expression matches this time, and you get | |
809 | the expected output of "table follows foo." | |
810 | ||
811 | Sometimes minimal matching can help a lot. Imagine you'd like to match | |
812 | everything between "foo" and "bar". Initially, you write something | |
813 | like this: | |
814 | ||
815 | $_ = "The food is under the bar in the barn."; | |
816 | if ( /foo(.*)bar/ ) { | |
817 | print "got <$1>\n"; | |
818 | } | |
819 | ||
820 | Which perhaps unexpectedly yields: | |
821 | ||
822 | got <d is under the bar in the > | |
823 | ||
824 | That's because C<.*> was greedy, so you get everything between the | |
14218588 | 825 | I<first> "foo" and the I<last> "bar". Here it's more effective |
c07a80fd | 826 | to use minimal matching to make sure you get the text between a "foo" |
827 | and the first "bar" thereafter. | |
828 | ||
829 | if ( /foo(.*?)bar/ ) { print "got <$1>\n" } | |
830 | got <d is under the > | |
831 | ||
832 | Here's another example: let's say you'd like to match a number at the end | |
b6e13d97 | 833 | of a string, and you also want to keep the preceding part of the match. |
c07a80fd | 834 | So you write this: |
835 | ||
836 | $_ = "I have 2 numbers: 53147"; | |
837 | if ( /(.*)(\d*)/ ) { # Wrong! | |
838 | print "Beginning is <$1>, number is <$2>.\n"; | |
839 | } | |
840 | ||
841 | That won't work at all, because C<.*> was greedy and gobbled up the | |
842 | whole string. As C<\d*> can match on an empty string the complete | |
843 | regular expression matched successfully. | |
844 | ||
8e1088bc | 845 | Beginning is <I have 2 numbers: 53147>, number is <>. |
c07a80fd | 846 | |
847 | Here are some variants, most of which don't work: | |
848 | ||
849 | $_ = "I have 2 numbers: 53147"; | |
850 | @pats = qw{ | |
851 | (.*)(\d*) | |
852 | (.*)(\d+) | |
853 | (.*?)(\d*) | |
854 | (.*?)(\d+) | |
855 | (.*)(\d+)$ | |
856 | (.*?)(\d+)$ | |
857 | (.*)\b(\d+)$ | |
858 | (.*\D)(\d+)$ | |
859 | }; | |
860 | ||
861 | for $pat (@pats) { | |
862 | printf "%-12s ", $pat; | |
863 | if ( /$pat/ ) { | |
864 | print "<$1> <$2>\n"; | |
865 | } else { | |
866 | print "FAIL\n"; | |
867 | } | |
868 | } | |
869 | ||
870 | That will print out: | |
871 | ||
872 | (.*)(\d*) <I have 2 numbers: 53147> <> | |
873 | (.*)(\d+) <I have 2 numbers: 5314> <7> | |
874 | (.*?)(\d*) <> <> | |
875 | (.*?)(\d+) <I have > <2> | |
876 | (.*)(\d+)$ <I have 2 numbers: 5314> <7> | |
877 | (.*?)(\d+)$ <I have 2 numbers: > <53147> | |
878 | (.*)\b(\d+)$ <I have 2 numbers: > <53147> | |
879 | (.*\D)(\d+)$ <I have 2 numbers: > <53147> | |
880 | ||
881 | As you see, this can be a bit tricky. It's important to realize that a | |
882 | regular expression is merely a set of assertions that gives a definition | |
883 | of success. There may be 0, 1, or several different ways that the | |
884 | definition might succeed against a particular string. And if there are | |
5a964f20 TC |
885 | multiple ways it might succeed, you need to understand backtracking to |
886 | know which variety of success you will achieve. | |
c07a80fd | 887 | |
19799a22 | 888 | When using look-ahead assertions and negations, this can all get even |
8b19b778 | 889 | trickier. Imagine you'd like to find a sequence of non-digits not |
c07a80fd | 890 | followed by "123". You might try to write that as |
891 | ||
871b0233 IZ |
892 | $_ = "ABC123"; |
893 | if ( /^\D*(?!123)/ ) { # Wrong! | |
894 | print "Yup, no 123 in $_\n"; | |
895 | } | |
c07a80fd | 896 | |
897 | But that isn't going to match; at least, not the way you're hoping. It | |
898 | claims that there is no 123 in the string. Here's a clearer picture of | |
9b9391b2 | 899 | why that pattern matches, contrary to popular expectations: |
c07a80fd | 900 | |
901 | $x = 'ABC123' ; | |
902 | $y = 'ABC445' ; | |
903 | ||
904 | print "1: got $1\n" if $x =~ /^(ABC)(?!123)/ ; | |
905 | print "2: got $1\n" if $y =~ /^(ABC)(?!123)/ ; | |
906 | ||
907 | print "3: got $1\n" if $x =~ /^(\D*)(?!123)/ ; | |
908 | print "4: got $1\n" if $y =~ /^(\D*)(?!123)/ ; | |
909 | ||
910 | This prints | |
911 | ||
912 | 2: got ABC | |
913 | 3: got AB | |
914 | 4: got ABC | |
915 | ||
5f05dabc | 916 | You might have expected test 3 to fail because it seems to a more |
c07a80fd | 917 | general purpose version of test 1. The important difference between |
918 | them is that test 3 contains a quantifier (C<\D*>) and so can use | |
919 | backtracking, whereas test 1 will not. What's happening is | |
920 | that you've asked "Is it true that at the start of $x, following 0 or more | |
5f05dabc | 921 | non-digits, you have something that's not 123?" If the pattern matcher had |
c07a80fd | 922 | let C<\D*> expand to "ABC", this would have caused the whole pattern to |
54310121 | 923 | fail. |
14218588 | 924 | |
c07a80fd | 925 | The search engine will initially match C<\D*> with "ABC". Then it will |
14218588 | 926 | try to match C<(?!123> with "123", which fails. But because |
c07a80fd | 927 | a quantifier (C<\D*>) has been used in the regular expression, the |
928 | search engine can backtrack and retry the match differently | |
54310121 | 929 | in the hope of matching the complete regular expression. |
c07a80fd | 930 | |
5a964f20 TC |
931 | The pattern really, I<really> wants to succeed, so it uses the |
932 | standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this | |
c07a80fd | 933 | time. Now there's indeed something following "AB" that is not |
14218588 | 934 | "123". It's "C123", which suffices. |
c07a80fd | 935 | |
14218588 GS |
936 | We can deal with this by using both an assertion and a negation. |
937 | We'll say that the first part in $1 must be followed both by a digit | |
938 | and by something that's not "123". Remember that the look-aheads | |
939 | are zero-width expressions--they only look, but don't consume any | |
940 | of the string in their match. So rewriting this way produces what | |
c07a80fd | 941 | you'd expect; that is, case 5 will fail, but case 6 succeeds: |
942 | ||
943 | print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/ ; | |
944 | print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/ ; | |
945 | ||
946 | 6: got ABC | |
947 | ||
5a964f20 | 948 | In other words, the two zero-width assertions next to each other work as though |
19799a22 | 949 | they're ANDed together, just as you'd use any built-in assertions: C</^$/> |
c07a80fd | 950 | matches only if you're at the beginning of the line AND the end of the |
951 | line simultaneously. The deeper underlying truth is that juxtaposition in | |
952 | regular expressions always means AND, except when you write an explicit OR | |
953 | using the vertical bar. C</ab/> means match "a" AND (then) match "b", | |
954 | although the attempted matches are made at different positions because "a" | |
955 | is not a zero-width assertion, but a one-width assertion. | |
956 | ||
19799a22 | 957 | B<WARNING>: particularly complicated regular expressions can take |
14218588 | 958 | exponential time to solve because of the immense number of possible |
9da458fc IZ |
959 | ways they can use backtracking to try match. For example, without |
960 | internal optimizations done by the regular expression engine, this will | |
961 | take a painfully long time to run: | |
c07a80fd | 962 | |
e1901655 IZ |
963 | 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ |
964 | ||
965 | And if you used C<*>'s in the internal groups instead of limiting them | |
966 | to 0 through 5 matches, then it would take forever--or until you ran | |
967 | out of stack space. Moreover, these internal optimizations are not | |
968 | always applicable. For example, if you put C<{0,5}> instead of C<*> | |
969 | on the external group, no current optimization is applicable, and the | |
970 | match takes a long time to finish. | |
c07a80fd | 971 | |
9da458fc IZ |
972 | A powerful tool for optimizing such beasts is what is known as an |
973 | "independent group", | |
c47ff5f1 | 974 | which does not backtrack (see L<C<< (?>pattern) >>>). Note also that |
9da458fc | 975 | zero-length look-ahead/look-behind assertions will not backtrack to make |
14218588 GS |
976 | the tail match, since they are in "logical" context: only |
977 | whether they match is considered relevant. For an example | |
9da458fc | 978 | where side-effects of look-ahead I<might> have influenced the |
c47ff5f1 | 979 | following match, see L<C<< (?>pattern) >>>. |
c277df42 | 980 | |
a0d0e21e LW |
981 | =head2 Version 8 Regular Expressions |
982 | ||
5a964f20 | 983 | In case you're not familiar with the "regular" Version 8 regex |
a0d0e21e LW |
984 | routines, here are the pattern-matching rules not described above. |
985 | ||
54310121 | 986 | Any single character matches itself, unless it is a I<metacharacter> |
a0d0e21e | 987 | with a special meaning described here or above. You can cause |
5a964f20 | 988 | characters that normally function as metacharacters to be interpreted |
5f05dabc | 989 | literally by prefixing them with a "\" (e.g., "\." matches a ".", not any |
a0d0e21e LW |
990 | character; "\\" matches a "\"). A series of characters matches that |
991 | series of characters in the target string, so the pattern C<blurfl> | |
992 | would match "blurfl" in the target string. | |
993 | ||
994 | You can specify a character class, by enclosing a list of characters | |
5a964f20 | 995 | in C<[]>, which will match any one character from the list. If the |
a0d0e21e | 996 | first character after the "[" is "^", the class matches any character not |
14218588 | 997 | in the list. Within a list, the "-" character specifies a |
5a964f20 | 998 | range, so that C<a-z> represents all characters between "a" and "z", |
8a4f6ac2 GS |
999 | inclusive. If you want either "-" or "]" itself to be a member of a |
1000 | class, put it at the start of the list (possibly after a "^"), or | |
1001 | escape it with a backslash. "-" is also taken literally when it is | |
1002 | at the end of the list, just before the closing "]". (The | |
84850974 DD |
1003 | following all specify the same class of three characters: C<[-az]>, |
1004 | C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which | |
fb55449c JH |
1005 | specifies a class containing twenty-six characters, even on EBCDIC |
1006 | based coded character sets.) Also, if you try to use the character | |
1007 | classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of | |
1008 | a range, that's not a range, the "-" is understood literally. | |
a0d0e21e | 1009 | |
8ada0baa JH |
1010 | Note also that the whole range idea is rather unportable between |
1011 | character sets--and even within character sets they may cause results | |
1012 | you probably didn't expect. A sound principle is to use only ranges | |
1013 | that begin from and end at either alphabets of equal case ([a-e], | |
1014 | [A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, | |
1015 | spell out the character sets in full. | |
1016 | ||
54310121 | 1017 | Characters may be specified using a metacharacter syntax much like that |
a0d0e21e LW |
1018 | used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, |
1019 | "\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string | |
fb55449c JH |
1020 | of octal digits, matches the character whose coded character set value |
1021 | is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, | |
1022 | matches the character whose numeric value is I<nn>. The expression \cI<x> | |
1023 | matches the character control-I<x>. Finally, the "." metacharacter | |
1024 | matches any character except "\n" (unless you use C</s>). | |
a0d0e21e LW |
1025 | |
1026 | You can specify a series of alternatives for a pattern using "|" to | |
1027 | separate them, so that C<fee|fie|foe> will match any of "fee", "fie", | |
5a964f20 | 1028 | or "foe" in the target string (as would C<f(e|i|o)e>). The |
a0d0e21e LW |
1029 | first alternative includes everything from the last pattern delimiter |
1030 | ("(", "[", or the beginning of the pattern) up to the first "|", and | |
1031 | the last alternative contains everything from the last "|" to the next | |
14218588 GS |
1032 | pattern delimiter. That's why it's common practice to include |
1033 | alternatives in parentheses: to minimize confusion about where they | |
a3cb178b GS |
1034 | start and end. |
1035 | ||
5a964f20 | 1036 | Alternatives are tried from left to right, so the first |
a3cb178b GS |
1037 | alternative found for which the entire expression matches, is the one that |
1038 | is chosen. This means that alternatives are not necessarily greedy. For | |
628afcb5 | 1039 | example: when matching C<foo|foot> against "barefoot", only the "foo" |
a3cb178b GS |
1040 | part will match, as that is the first alternative tried, and it successfully |
1041 | matches the target string. (This might not seem important, but it is | |
1042 | important when you are capturing matched text using parentheses.) | |
1043 | ||
5a964f20 | 1044 | Also remember that "|" is interpreted as a literal within square brackets, |
a3cb178b | 1045 | so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. |
a0d0e21e | 1046 | |
14218588 GS |
1047 | Within a pattern, you may designate subpatterns for later reference |
1048 | by enclosing them in parentheses, and you may refer back to the | |
1049 | I<n>th subpattern later in the pattern using the metacharacter | |
1050 | \I<n>. Subpatterns are numbered based on the left to right order | |
1051 | of their opening parenthesis. A backreference matches whatever | |
1052 | actually matched the subpattern in the string being examined, not | |
1053 | the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will | |
1054 | match "0x1234 0x4321", but not "0x1234 01234", because subpattern | |
1055 | 1 matched "0x", even though the rule C<0|0x> could potentially match | |
1056 | the leading 0 in the second number. | |
cb1a09d0 | 1057 | |
19799a22 | 1058 | =head2 Warning on \1 vs $1 |
cb1a09d0 | 1059 | |
5a964f20 | 1060 | Some people get too used to writing things like: |
cb1a09d0 AD |
1061 | |
1062 | $pattern =~ s/(\W)/\\\1/g; | |
1063 | ||
1064 | This is grandfathered for the RHS of a substitute to avoid shocking the | |
1065 | B<sed> addicts, but it's a dirty habit to get into. That's because in | |
d1be9408 | 1066 | PerlThink, the righthand side of an C<s///> is a double-quoted string. C<\1> in |
cb1a09d0 AD |
1067 | the usual double-quoted string means a control-A. The customary Unix |
1068 | meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit | |
1069 | of doing that, you get yourself into trouble if you then add an C</e> | |
1070 | modifier. | |
1071 | ||
5a964f20 | 1072 | s/(\d+)/ \1 + 1 /eg; # causes warning under -w |
cb1a09d0 AD |
1073 | |
1074 | Or if you try to do | |
1075 | ||
1076 | s/(\d+)/\1000/; | |
1077 | ||
1078 | You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with | |
14218588 | 1079 | C<${1}000>. The operation of interpolation should not be confused |
cb1a09d0 AD |
1080 | with the operation of matching a backreference. Certainly they mean two |
1081 | different things on the I<left> side of the C<s///>. | |
9fa51da4 | 1082 | |
c84d73f1 IZ |
1083 | =head2 Repeated patterns matching zero-length substring |
1084 | ||
19799a22 | 1085 | B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. |
c84d73f1 IZ |
1086 | |
1087 | Regular expressions provide a terse and powerful programming language. As | |
1088 | with most other power tools, power comes together with the ability | |
1089 | to wreak havoc. | |
1090 | ||
1091 | A common abuse of this power stems from the ability to make infinite | |
628afcb5 | 1092 | loops using regular expressions, with something as innocuous as: |
c84d73f1 IZ |
1093 | |
1094 | 'foo' =~ m{ ( o? )* }x; | |
1095 | ||
1096 | The C<o?> can match at the beginning of C<'foo'>, and since the position | |
1097 | in the string is not moved by the match, C<o?> would match again and again | |
14218588 | 1098 | because of the C<*> modifier. Another common way to create a similar cycle |
c84d73f1 IZ |
1099 | is with the looping modifier C<//g>: |
1100 | ||
1101 | @matches = ( 'foo' =~ m{ o? }xg ); | |
1102 | ||
1103 | or | |
1104 | ||
1105 | print "match: <$&>\n" while 'foo' =~ m{ o? }xg; | |
1106 | ||
1107 | or the loop implied by split(). | |
1108 | ||
1109 | However, long experience has shown that many programming tasks may | |
14218588 GS |
1110 | be significantly simplified by using repeated subexpressions that |
1111 | may match zero-length substrings. Here's a simple example being: | |
c84d73f1 IZ |
1112 | |
1113 | @chars = split //, $string; # // is not magic in split | |
1114 | ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / | |
1115 | ||
9da458fc | 1116 | Thus Perl allows such constructs, by I<forcefully breaking |
c84d73f1 IZ |
1117 | the infinite loop>. The rules for this are different for lower-level |
1118 | loops given by the greedy modifiers C<*+{}>, and for higher-level | |
1119 | ones like the C</g> modifier or split() operator. | |
1120 | ||
19799a22 GS |
1121 | The lower-level loops are I<interrupted> (that is, the loop is |
1122 | broken) when Perl detects that a repeated expression matched a | |
1123 | zero-length substring. Thus | |
c84d73f1 IZ |
1124 | |
1125 | m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; | |
1126 | ||
1127 | is made equivalent to | |
1128 | ||
1129 | m{ (?: NON_ZERO_LENGTH )* | |
1130 | | | |
1131 | (?: ZERO_LENGTH )? | |
1132 | }x; | |
1133 | ||
1134 | The higher level-loops preserve an additional state between iterations: | |
1135 | whether the last match was zero-length. To break the loop, the following | |
1136 | match after a zero-length match is prohibited to have a length of zero. | |
1137 | This prohibition interacts with backtracking (see L<"Backtracking">), | |
1138 | and so the I<second best> match is chosen if the I<best> match is of | |
1139 | zero length. | |
1140 | ||
19799a22 | 1141 | For example: |
c84d73f1 IZ |
1142 | |
1143 | $_ = 'bar'; | |
1144 | s/\w??/<$&>/g; | |
1145 | ||
20fb949f | 1146 | results in C<< <><b><><a><><r><> >>. At each position of the string the best |
c84d73f1 IZ |
1147 | match given by non-greedy C<??> is the zero-length match, and the I<second |
1148 | best> match is what is matched by C<\w>. Thus zero-length matches | |
1149 | alternate with one-character-long matches. | |
1150 | ||
1151 | Similarly, for repeated C<m/()/g> the second-best match is the match at the | |
1152 | position one notch further in the string. | |
1153 | ||
19799a22 | 1154 | The additional state of being I<matched with zero-length> is associated with |
c84d73f1 | 1155 | the matched string, and is reset by each assignment to pos(). |
9da458fc IZ |
1156 | Zero-length matches at the end of the previous match are ignored |
1157 | during C<split>. | |
c84d73f1 | 1158 | |
35a734be IZ |
1159 | =head2 Combining pieces together |
1160 | ||
1161 | Each of the elementary pieces of regular expressions which were described | |
1162 | before (such as C<ab> or C<\Z>) could match at most one substring | |
1163 | at the given position of the input string. However, in a typical regular | |
1164 | expression these elementary pieces are combined into more complicated | |
1165 | patterns using combining operators C<ST>, C<S|T>, C<S*> etc | |
1166 | (in these examples C<S> and C<T> are regular subexpressions). | |
1167 | ||
1168 | Such combinations can include alternatives, leading to a problem of choice: | |
1169 | if we match a regular expression C<a|ab> against C<"abc">, will it match | |
1170 | substring C<"a"> or C<"ab">? One way to describe which substring is | |
1171 | actually matched is the concept of backtracking (see L<"Backtracking">). | |
1172 | However, this description is too low-level and makes you think | |
1173 | in terms of a particular implementation. | |
1174 | ||
1175 | Another description starts with notions of "better"/"worse". All the | |
1176 | substrings which may be matched by the given regular expression can be | |
1177 | sorted from the "best" match to the "worst" match, and it is the "best" | |
1178 | match which is chosen. This substitutes the question of "what is chosen?" | |
1179 | by the question of "which matches are better, and which are worse?". | |
1180 | ||
1181 | Again, for elementary pieces there is no such question, since at most | |
1182 | one match at a given position is possible. This section describes the | |
1183 | notion of better/worse for combining operators. In the description | |
1184 | below C<S> and C<T> are regular subexpressions. | |
1185 | ||
13a2d996 | 1186 | =over 4 |
35a734be IZ |
1187 | |
1188 | =item C<ST> | |
1189 | ||
1190 | Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are | |
1191 | substrings which can be matched by C<S>, C<B> and C<B'> are substrings | |
1192 | which can be matched by C<T>. | |
1193 | ||
1194 | If C<A> is better match for C<S> than C<A'>, C<AB> is a better | |
1195 | match than C<A'B'>. | |
1196 | ||
1197 | If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if | |
1198 | C<B> is better match for C<T> than C<B'>. | |
1199 | ||
1200 | =item C<S|T> | |
1201 | ||
1202 | When C<S> can match, it is a better match than when only C<T> can match. | |
1203 | ||
1204 | Ordering of two matches for C<S> is the same as for C<S>. Similar for | |
1205 | two matches for C<T>. | |
1206 | ||
1207 | =item C<S{REPEAT_COUNT}> | |
1208 | ||
1209 | Matches as C<SSS...S> (repeated as many times as necessary). | |
1210 | ||
1211 | =item C<S{min,max}> | |
1212 | ||
1213 | Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. | |
1214 | ||
1215 | =item C<S{min,max}?> | |
1216 | ||
1217 | Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. | |
1218 | ||
1219 | =item C<S?>, C<S*>, C<S+> | |
1220 | ||
1221 | Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. | |
1222 | ||
1223 | =item C<S??>, C<S*?>, C<S+?> | |
1224 | ||
1225 | Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. | |
1226 | ||
c47ff5f1 | 1227 | =item C<< (?>S) >> |
35a734be IZ |
1228 | |
1229 | Matches the best match for C<S> and only that. | |
1230 | ||
1231 | =item C<(?=S)>, C<(?<=S)> | |
1232 | ||
1233 | Only the best match for C<S> is considered. (This is important only if | |
1234 | C<S> has capturing parentheses, and backreferences are used somewhere | |
1235 | else in the whole regular expression.) | |
1236 | ||
1237 | =item C<(?!S)>, C<(?<!S)> | |
1238 | ||
1239 | For this grouping operator there is no need to describe the ordering, since | |
1240 | only whether or not C<S> can match is important. | |
1241 | ||
14455d6c | 1242 | =item C<(??{ EXPR })> |
35a734be IZ |
1243 | |
1244 | The ordering is the same as for the regular expression which is | |
1245 | the result of EXPR. | |
1246 | ||
1247 | =item C<(?(condition)yes-pattern|no-pattern)> | |
1248 | ||
1249 | Recall that which of C<yes-pattern> or C<no-pattern> actually matches is | |
1250 | already determined. The ordering of the matches is the same as for the | |
1251 | chosen subexpression. | |
1252 | ||
1253 | =back | |
1254 | ||
1255 | The above recipes describe the ordering of matches I<at a given position>. | |
1256 | One more rule is needed to understand how a match is determined for the | |
1257 | whole regular expression: a match at an earlier position is always better | |
1258 | than a match at a later position. | |
1259 | ||
c84d73f1 IZ |
1260 | =head2 Creating custom RE engines |
1261 | ||
1262 | Overloaded constants (see L<overload>) provide a simple way to extend | |
1263 | the functionality of the RE engine. | |
1264 | ||
1265 | Suppose that we want to enable a new RE escape-sequence C<\Y|> which | |
1266 | matches at boundary between white-space characters and non-whitespace | |
1267 | characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly | |
1268 | at these positions, so we want to have each C<\Y|> in the place of the | |
1269 | more complicated version. We can create a module C<customre> to do | |
1270 | this: | |
1271 | ||
1272 | package customre; | |
1273 | use overload; | |
1274 | ||
1275 | sub import { | |
1276 | shift; | |
1277 | die "No argument to customre::import allowed" if @_; | |
1278 | overload::constant 'qr' => \&convert; | |
1279 | } | |
1280 | ||
1281 | sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} | |
1282 | ||
1283 | my %rules = ( '\\' => '\\', | |
1284 | 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); | |
1285 | sub convert { | |
1286 | my $re = shift; | |
1287 | $re =~ s{ | |
1288 | \\ ( \\ | Y . ) | |
1289 | } | |
1290 | { $rules{$1} or invalid($re,$1) }sgex; | |
1291 | return $re; | |
1292 | } | |
1293 | ||
1294 | Now C<use customre> enables the new escape in constant regular | |
1295 | expressions, i.e., those without any runtime variable interpolations. | |
1296 | As documented in L<overload>, this conversion will work only over | |
1297 | literal parts of regular expressions. For C<\Y|$re\Y|> the variable | |
1298 | part of this regular expression needs to be converted explicitly | |
1299 | (but only if the special meaning of C<\Y|> should be enabled inside $re): | |
1300 | ||
1301 | use customre; | |
1302 | $re = <>; | |
1303 | chomp $re; | |
1304 | $re = customre::convert $re; | |
1305 | /\Y|$re\Y|/; | |
1306 | ||
19799a22 GS |
1307 | =head1 BUGS |
1308 | ||
9da458fc IZ |
1309 | This document varies from difficult to understand to completely |
1310 | and utterly opaque. The wandering prose riddled with jargon is | |
1311 | hard to fathom in several places. | |
1312 | ||
1313 | This document needs a rewrite that separates the tutorial content | |
1314 | from the reference content. | |
19799a22 GS |
1315 | |
1316 | =head1 SEE ALSO | |
9fa51da4 | 1317 | |
91e0c79e MJD |
1318 | L<perlrequick>. |
1319 | ||
1320 | L<perlretut>. | |
1321 | ||
9b599b2a GS |
1322 | L<perlop/"Regexp Quote-Like Operators">. |
1323 | ||
1e66bd83 PP |
1324 | L<perlop/"Gory details of parsing quoted constructs">. |
1325 | ||
14218588 GS |
1326 | L<perlfaq6>. |
1327 | ||
9b599b2a GS |
1328 | L<perlfunc/pos>. |
1329 | ||
1330 | L<perllocale>. | |
1331 | ||
fb55449c JH |
1332 | L<perlebcdic>. |
1333 | ||
14218588 GS |
1334 | I<Mastering Regular Expressions> by Jeffrey Friedl, published |
1335 | by O'Reilly and Associates. |