1=head1 NAME 2 3perlre - Perl regular expressions 4 5=head1 DESCRIPTION 6 7This page describes the syntax of regular expressions in Perl. For a 8description of how to I<use> regular expressions in matching 9operations, plus various examples of the same, see discussions 10of C<m//>, C<s///>, C<qr//> and C<??> in L<perlop/"Regexp Quote-Like Operators">. 11 12Matching operations can have various modifiers. Modifiers 13that relate to the interpretation of the regular expression inside 14are listed below. Modifiers that alter the way a regular expression 15is used by Perl are detailed in L<perlop/"Regexp Quote-Like Operators"> and 16L<perlop/"Gory details of parsing quoted constructs">. 17 18=over 4 19 20=item i 21 22Do case-insensitive pattern matching. 23 24If C<use locale> is in effect, the case map is taken from the current 25locale. See L<perllocale>. 26 27=item m 28 29Treat string as multiple lines. That is, change "^" and "$" from matching 30the start or end of the string to matching the start or end of any 31line anywhere within the string. 32 33=item s 34 35Treat string as single line. That is, change "." to match any character 36whatsoever, even a newline, which normally it would not match. 37 38The C</s> and C</m> modifiers both override the C<$*> setting. That 39is, no matter what C<$*> contains, C</s> without C</m> will force 40"^" to match only at the beginning of the string and "$" to match 41only at the end (or just before a newline at the end) of the string. 42Together, as /ms, they let the "." match any character whatsoever, 43while still allowing "^" and "$" to match, respectively, just after 44and just before newlines within the string. 45 46=item x 47 48Extend your pattern's legibility by permitting whitespace and comments. 49 50=back 51 52These are usually written as "the C</x> modifier", even though the delimiter 53in question might not really be a slash. Any of these 54modifiers may also be embedded within the regular expression itself using 55the C<(?...)> construct. See below. 56 57The C</x> modifier itself needs a little more explanation. It tells 58the regular expression parser to ignore whitespace that is neither 59backslashed nor within a character class. You can use this to break up 60your regular expression into (slightly) more readable parts. The C<#> 61character is also treated as a metacharacter introducing a comment, 62just as in ordinary Perl code. This also means that if you want real 63whitespace or C<#> characters in the pattern (outside a character 64class, where they are unaffected by C</x>), that you'll either have to 65escape them or encode them using octal or hex escapes. Taken together, 66these features go a long way towards making Perl's regular expressions 67more readable. Note that you have to be careful not to include the 68pattern delimiter in the comment--perl has no way of knowing you did 69not intend to close the pattern early. See the C-comment deletion code 70in L<perlop>. 71 72=head2 Regular Expressions 73 74The patterns used in Perl pattern matching derive from supplied in 75the Version 8 regex routines. (The routines are derived 76(distantly) from Henry Spencer's freely redistributable reimplementation 77of the V8 routines.) See L<Version 8 Regular Expressions> for 78details. 79 80In particular the following metacharacters have their standard I<egrep>-ish 81meanings: 82 83 \ Quote the next metacharacter 84 ^ Match the beginning of the line 85 . Match any character (except newline) 86 $ Match the end of the line (or before newline at the end) 87 | Alternation 88 () Grouping 89 [] Character class 90 91By default, the "^" character is guaranteed to match only the 92beginning of the string, the "$" character only the end (or before the 93newline at the end), and Perl does certain optimizations with the 94assumption that the string contains only one line. Embedded newlines 95will not be matched by "^" or "$". You may, however, wish to treat a 96string as a multi-line buffer, such that the "^" will match after any 97newline within the string, and "$" will match before any newline. At the 98cost of a little more overhead, you can do this by using the /m modifier 99on the pattern match operator. (Older programs did this by setting C<$*>, 100but this practice is now deprecated.) 101 102To simplify multi-line substitutions, the "." character never matches a 103newline unless you use the C</s> modifier, which in effect tells Perl to pretend 104the string is a single line--even if it isn't. The C</s> modifier also 105overrides the setting of C<$*>, in case you have some (badly behaved) older 106code that sets it in another module. 107 108The following standard quantifiers are recognized: 109 110 * Match 0 or more times 111 + Match 1 or more times 112 ? Match 1 or 0 times 113 {n} Match exactly n times 114 {n,} Match at least n times 115 {n,m} Match at least n but not more than m times 116 117(If a curly bracket occurs in any other context, it is treated 118as a regular character.) The "*" modifier is equivalent to C<{0,}>, the "+" 119modifier to C<{1,}>, and the "?" modifier to C<{0,1}>. n and m are limited 120to integral values less than a preset limit defined when perl is built. 121This is usually 32766 on the most common platforms. The actual limit can 122be seen in the error message generated by code such as this: 123 124 $_ **= $_ , / {$_} / for 2 .. 42; 125 126By default, a quantified subpattern is "greedy", that is, it will match as 127many times as possible (given a particular starting location) while still 128allowing the rest of the pattern to match. If you want it to match the 129minimum number of times possible, follow the quantifier with a "?". Note 130that the meanings don't change, just the "greediness": 131 132 *? Match 0 or more times 133 +? Match 1 or more times 134 ?? Match 0 or 1 time 135 {n}? Match exactly n times 136 {n,}? Match at least n times 137 {n,m}? Match at least n but not more than m times 138 139Because patterns are processed as double quoted strings, the following 140also work: 141 142 \t tab (HT, TAB) 143 \n newline (LF, NL) 144 \r return (CR) 145 \f form feed (FF) 146 \a alarm (bell) (BEL) 147 \e escape (think troff) (ESC) 148 \033 octal char (think of a PDP-11) 149 \x1B hex char 150 \x{263a} wide hex char (Unicode SMILEY) 151 \c[ control char 152 \N{name} named char 153 \l lowercase next char (think vi) 154 \u uppercase next char (think vi) 155 \L lowercase till \E (think vi) 156 \U uppercase till \E (think vi) 157 \E end case modification (think vi) 158 \Q quote (disable) pattern metacharacters till \E 159 160If C<use locale> is in effect, the case map used by C<\l>, C<\L>, C<\u> 161and C<\U> is taken from the current locale. See L<perllocale>. For 162documentation of C<\N{name}>, see L<charnames>. 163 164You cannot include a literal C<$> or C<@> within a C<\Q> sequence. 165An unescaped C<$> or C<@> interpolates the corresponding variable, 166while escaping will cause the literal string C<\$> to be matched. 167You'll need to write something like C<m/\Quser\E\@\Qhost/>. 168 169In addition, Perl defines the following: 170 171 \w Match a "word" character (alphanumeric plus "_") 172 \W Match a non-"word" character 173 \s Match a whitespace character 174 \S Match a non-whitespace character 175 \d Match a digit character 176 \D Match a non-digit character 177 \pP Match P, named property. Use \p{Prop} for longer names. 178 \PP Match non-P 179 \X Match eXtended Unicode "combining character sequence", 180 equivalent to C<(?:\PM\pM*)> 181 \C Match a single C char (octet) even under utf8. 182 183A C<\w> matches a single alphanumeric character or C<_>, not a whole word. 184Use C<\w+> to match a string of Perl-identifier characters (which isn't 185the same as matching an English word). If C<use locale> is in effect, the 186list of alphabetic characters generated by C<\w> is taken from the 187current locale. See L<perllocale>. You may use C<\w>, C<\W>, C<\s>, C<\S>, 188C<\d>, and C<\D> within character classes, but if you try to use them 189as endpoints of a range, that's not a range, the "-" is understood literally. 190See L<utf8> for details about C<\pP>, C<\PP>, and C<\X>. 191 192The POSIX character class syntax 193 194 [:class:] 195 196is also available. The available classes and their backslash 197equivalents (if available) are as follows: 198 199 alpha 200 alnum 201 ascii 202 blank [1] 203 cntrl 204 digit \d 205 graph 206 lower 207 print 208 punct 209 space \s [2] 210 upper 211 word \w [3] 212 xdigit 213 214 [1] A GNU extension equivalent to C<[ \t]>, `all horizontal whitespace'. 215 [2] Not I<exactly equivalent> to C<\s> since the C<[[:space:]]> includes 216 also the (very rare) `vertical tabulator', "\ck", chr(11). 217 [3] A Perl extension. 218 219For example use C<[:upper:]> to match all the uppercase characters. 220Note that the C<[]> are part of the C<[::]> construct, not part of the 221whole character class. For example: 222 223 [01[:alpha:]%] 224 225matches zero, one, any alphabetic character, and the percentage sign. 226 227If the C<utf8> pragma is used, the following equivalences to Unicode 228\p{} constructs and equivalent backslash character classes (if available), 229will hold: 230 231 alpha IsAlpha 232 alnum IsAlnum 233 ascii IsASCII 234 blank IsSpace 235 cntrl IsCntrl 236 digit IsDigit \d 237 graph IsGraph 238 lower IsLower 239 print IsPrint 240 punct IsPunct 241 space IsSpace 242 IsSpacePerl \s 243 upper IsUpper 244 word IsWord 245 xdigit IsXDigit 246 247For example C<[:lower:]> and C<\p{IsLower}> are equivalent. 248 249If the C<utf8> pragma is not used but the C<locale> pragma is, the 250classes correlate with the usual isalpha(3) interface (except for 251`word' and `blank'). 252 253The assumedly non-obviously named classes are: 254 255=over 4 256 257=item cntrl 258 259Any control character. Usually characters that don't produce output as 260such but instead control the terminal somehow: for example newline and 261backspace are control characters. All characters with ord() less than 26232 are most often classified as control characters (assuming ASCII, 263the ISO Latin character sets, and Unicode). 264 265=item graph 266 267Any alphanumeric or punctuation (special) character. 268 269=item print 270 271Any alphanumeric or punctuation (special) character or space. 272 273=item punct 274 275Any punctuation (special) character. 276 277=item xdigit 278 279Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f] would 280work just fine) it is included for completeness. 281 282=back 283 284You can negate the [::] character classes by prefixing the class name 285with a '^'. This is a Perl extension. For example: 286 287 POSIX trad. Perl utf8 Perl 288 289 [:^digit:] \D \P{IsDigit} 290 [:^space:] \S \P{IsSpace} 291 [:^word:] \W \P{IsWord} 292 293The POSIX character classes [.cc.] and [=cc=] are recognized but 294B<not> supported and trying to use them will cause an error. 295 296Perl defines the following zero-width assertions: 297 298 \b Match a word boundary 299 \B Match a non-(word boundary) 300 \A Match only at beginning of string 301 \Z Match only at end of string, or before newline at the end 302 \z Match only at end of string 303 \G Match only at pos() (e.g. at the end-of-match position 304 of prior m//g) 305 306A word boundary (C<\b>) is a spot between two characters 307that has a C<\w> on one side of it and a C<\W> on the other side 308of it (in either order), counting the imaginary characters off the 309beginning and end of the string as matching a C<\W>. (Within 310character classes C<\b> represents backspace rather than a word 311boundary, just as it normally does in any double-quoted string.) 312The C<\A> and C<\Z> are just like "^" and "$", except that they 313won't match multiple times when the C</m> modifier is used, while 314"^" and "$" will match at every internal line boundary. To match 315the actual end of the string and not ignore an optional trailing 316newline, use C<\z>. 317 318The C<\G> assertion can be used to chain global matches (using 319C<m//g>), as described in L<perlop/"Regexp Quote-Like Operators">. 320It is also useful when writing C<lex>-like scanners, when you have 321several patterns that you want to match against consequent substrings 322of your string, see the previous reference. The actual location 323where C<\G> will match can also be influenced by using C<pos()> as 324an lvalue. See L<perlfunc/pos>. 325 326The bracketing construct C<( ... )> creates capture buffers. To 327refer to the digit'th buffer use \<digit> within the 328match. Outside the match use "$" instead of "\". (The 329\<digit> notation works in certain circumstances outside 330the match. See the warning below about \1 vs $1 for details.) 331Referring back to another part of the match is called a 332I<backreference>. 333 334There is no limit to the number of captured substrings that you may 335use. However Perl also uses \10, \11, etc. as aliases for \010, 336\011, etc. (Recall that 0 means octal, so \011 is the character at 337number 9 in your coded character set; which would be the 10th character, 338a horizontal tab under ASCII.) Perl resolves this 339ambiguity by interpreting \10 as a backreference only if at least 10 340left parentheses have opened before it. Likewise \11 is a 341backreference only if at least 11 left parentheses have opened 342before it. And so on. \1 through \9 are always interpreted as 343backreferences. 344 345Examples: 346 347 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words 348 349 if (/(.)\1/) { # find first doubled char 350 print "'$1' is the first doubled character\n"; 351 } 352 353 if (/Time: (..):(..):(..)/) { # parse out values 354 $hours = $1; 355 $minutes = $2; 356 $seconds = $3; 357 } 358 359Several special variables also refer back to portions of the previous 360match. C<$+> returns whatever the last bracket match matched. 361C<$&> returns the entire matched string. (At one point C<$0> did 362also, but now it returns the name of the program.) C<$`> returns 363everything before the matched string. And C<$'> returns everything 364after the matched string. 365 366The numbered variables ($1, $2, $3, etc.) and the related punctuation 367set (C<$+>, C<$&>, C<$`>, and C<$'>) are all dynamically scoped 368until the end of the enclosing block or until the next successful 369match, whichever comes first. (See L<perlsyn/"Compound Statements">.) 370 371B<WARNING>: Once Perl sees that you need one of C<$&>, C<$`>, or 372C<$'> anywhere in the program, it has to provide them for every 373pattern match. This may substantially slow your program. Perl 374uses the same mechanism to produce $1, $2, etc, so you also pay a 375price for each pattern that contains capturing parentheses. (To 376avoid this cost while retaining the grouping behaviour, use the 377extended regular expression C<(?: ... )> instead.) But if you never 378use C<$&>, C<$`> or C<$'>, then patterns I<without> capturing 379parentheses will not be penalized. So avoid C<$&>, C<$'>, and C<$`> 380if you can, but if you can't (and some algorithms really appreciate 381them), once you've used them once, use them at will, because you've 382already paid the price. As of 5.005, C<$&> is not so costly as the 383other two. 384 385Backslashed metacharacters in Perl are alphanumeric, such as C<\b>, 386C<\w>, C<\n>. Unlike some other regular expression languages, there 387are no backslashed symbols that aren't alphanumeric. So anything 388that looks like \\, \(, \), \<, \>, \{, or \} is always 389interpreted as a literal character, not a metacharacter. This was 390once used in a common idiom to disable or quote the special meanings 391of regular expression metacharacters in a string that you want to 392use for a pattern. Simply quote all non-"word" characters: 393 394 $pattern =~ s/(\W)/\\$1/g; 395 396(If C<use locale> is set, then this depends on the current locale.) 397Today it is more common to use the quotemeta() function or the C<\Q> 398metaquoting escape sequence to disable all metacharacters' special 399meanings like this: 400 401 /$unquoted\Q$quoted\E$unquoted/ 402 403Beware that if you put literal backslashes (those not inside 404interpolated variables) between C<\Q> and C<\E>, double-quotish 405backslash interpolation may lead to confusing results. If you 406I<need> to use literal backslashes within C<\Q...\E>, 407consult L<perlop/"Gory details of parsing quoted constructs">. 408 409=head2 Extended Patterns 410 411Perl also defines a consistent extension syntax for features not 412found in standard tools like B<awk> and B<lex>. The syntax is a 413pair of parentheses with a question mark as the first thing within 414the parentheses. The character after the question mark indicates 415the extension. 416 417The stability of these extensions varies widely. Some have been 418part of the core language for many years. Others are experimental 419and may change without warning or be completely removed. Check 420the documentation on an individual feature to verify its current 421status. 422 423A question mark was chosen for this and for the minimal-matching 424construct because 1) question marks are rare in older regular 425expressions, and 2) whenever you see one, you should stop and 426"question" exactly what is going on. That's psychology... 427 428=over 10 429 430=item C<(?#text)> 431 432A comment. The text is ignored. If the C</x> modifier enables 433whitespace formatting, a simple C<#> will suffice. Note that Perl closes 434the comment as soon as it sees a C<)>, so there is no way to put a literal 435C<)> in the comment. 436 437=item C<(?imsx-imsx)> 438 439One or more embedded pattern-match modifiers. This is particularly 440useful for dynamic patterns, such as those read in from a configuration 441file, read in as an argument, are specified in a table somewhere, 442etc. Consider the case that some of which want to be case sensitive 443and some do not. The case insensitive ones need to include merely 444C<(?i)> at the front of the pattern. For example: 445 446 $pattern = "foobar"; 447 if ( /$pattern/i ) { } 448 449 # more flexible: 450 451 $pattern = "(?i)foobar"; 452 if ( /$pattern/ ) { } 453 454Letters after a C<-> turn those modifiers off. These modifiers are 455localized inside an enclosing group (if any). For example, 456 457 ( (?i) blah ) \s+ \1 458 459will match a repeated (I<including the case>!) word C<blah> in any 460case, assuming C<x> modifier, and no C<i> modifier outside this 461group. 462 463=item C<(?:pattern)> 464 465=item C<(?imsx-imsx:pattern)> 466 467This is for clustering, not capturing; it groups subexpressions like 468"()", but doesn't make backreferences as "()" does. So 469 470 @fields = split(/\b(?:a|b|c)\b/) 471 472is like 473 474 @fields = split(/\b(a|b|c)\b/) 475 476but doesn't spit out extra fields. It's also cheaper not to capture 477characters if you don't need to. 478 479Any letters between C<?> and C<:> act as flags modifiers as with 480C<(?imsx-imsx)>. For example, 481 482 /(?s-i:more.*than).*million/i 483 484is equivalent to the more verbose 485 486 /(?:(?s-i)more.*than).*million/i 487 488=item C<(?=pattern)> 489 490A zero-width positive look-ahead assertion. For example, C</\w+(?=\t)/> 491matches a word followed by a tab, without including the tab in C<$&>. 492 493=item C<(?!pattern)> 494 495A zero-width negative look-ahead assertion. For example C</foo(?!bar)/> 496matches any occurrence of "foo" that isn't followed by "bar". Note 497however that look-ahead and look-behind are NOT the same thing. You cannot 498use this for look-behind. 499 500If you are looking for a "bar" that isn't preceded by a "foo", C</(?!foo)bar/> 501will not do what you want. That's because the C<(?!foo)> is just saying that 502the next thing cannot be "foo"--and it's not, it's a "bar", so "foobar" will 503match. You would have to do something like C</(?!foo)...bar/> for that. We 504say "like" because there's the case of your "bar" not having three characters 505before it. You could cover that this way: C</(?:(?!foo)...|^.{0,2})bar/>. 506Sometimes it's still easier just to say: 507 508 if (/bar/ && $` !~ /foo$/) 509 510For look-behind see below. 511 512=item C<(?<=pattern)> 513 514A zero-width positive look-behind assertion. For example, C</(?<=\t)\w+/> 515matches a word that follows a tab, without including the tab in C<$&>. 516Works only for fixed-width look-behind. 517 518=item C<(?<!pattern)> 519 520A zero-width negative look-behind assertion. For example C</(?<!bar)foo/> 521matches any occurrence of "foo" that does not follow "bar". Works 522only for fixed-width look-behind. 523 524=item C<(?{ code })> 525 526B<WARNING>: This extended regular expression feature is considered 527highly experimental, and may be changed or deleted without notice. 528 529This zero-width assertion evaluate any embedded Perl code. It 530always succeeds, and its C<code> is not interpolated. Currently, 531the rules to determine where the C<code> ends are somewhat convoluted. 532 533The C<code> is properly scoped in the following sense: If the assertion 534is backtracked (compare L<"Backtracking">), all changes introduced after 535C<local>ization are undone, so that 536 537 $_ = 'a' x 8; 538 m< 539 (?{ $cnt = 0 }) # Initialize $cnt. 540 ( 541 a 542 (?{ 543 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe. 544 }) 545 )* 546 aaaa 547 (?{ $res = $cnt }) # On success copy to non-localized 548 # location. 549 >x; 550 551will set C<$res = 4>. Note that after the match, $cnt returns to the globally 552introduced value, because the scopes that restrict C<local> operators 553are unwound. 554 555This assertion may be used as a C<(?(condition)yes-pattern|no-pattern)> 556switch. If I<not> used in this way, the result of evaluation of 557C<code> is put into the special variable C<$^R>. This happens 558immediately, so C<$^R> can be used from other C<(?{ code })> assertions 559inside the same regular expression. 560 561The assignment to C<$^R> above is properly localized, so the old 562value of C<$^R> is restored if the assertion is backtracked; compare 563L<"Backtracking">. 564 565For reasons of security, this construct is forbidden if the regular 566expression involves run-time interpolation of variables, unless the 567perilous C<use re 'eval'> pragma has been used (see L<re>), or the 568variables contain results of C<qr//> operator (see 569L<perlop/"qr/STRING/imosx">). 570 571This restriction is because of the wide-spread and remarkably convenient 572custom of using run-time determined strings as patterns. For example: 573 574 $re = <>; 575 chomp $re; 576 $string =~ /$re/; 577 578Before Perl knew how to execute interpolated code within a pattern, 579this operation was completely safe from a security point of view, 580although it could raise an exception from an illegal pattern. If 581you turn on the C<use re 'eval'>, though, it is no longer secure, 582so you should only do so if you are also using taint checking. 583Better yet, use the carefully constrained evaluation within a Safe 584module. See L<perlsec> for details about both these mechanisms. 585 586=item C<(??{ code })> 587 588B<WARNING>: This extended regular expression feature is considered 589highly experimental, and may be changed or deleted without notice. 590A simplified version of the syntax may be introduced for commonly 591used idioms. 592 593This is a "postponed" regular subexpression. The C<code> is evaluated 594at run time, at the moment this subexpression may match. The result 595of evaluation is considered as a regular expression and matched as 596if it were inserted instead of this construct. 597 598The C<code> is not interpolated. As before, the rules to determine 599where the C<code> ends are currently somewhat convoluted. 600 601The following pattern matches a parenthesized group: 602 603 $re = qr{ 604 \( 605 (?: 606 (?> [^()]+ ) # Non-parens without backtracking 607 | 608 (??{ $re }) # Group with matching parens 609 )* 610 \) 611 }x; 612 613=item C<< (?>pattern) >> 614 615B<WARNING>: This extended regular expression feature is considered 616highly experimental, and may be changed or deleted without notice. 617 618An "independent" subexpression, one which matches the substring 619that a I<standalone> C<pattern> would match if anchored at the given 620position, and it matches I<nothing other than this substring>. This 621construct is useful for optimizations of what would otherwise be 622"eternal" matches, because it will not backtrack (see L<"Backtracking">). 623It may also be useful in places where the "grab all you can, and do not 624give anything back" semantic is desirable. 625 626For example: C<< ^(?>a*)ab >> will never match, since C<< (?>a*) >> 627(anchored at the beginning of string, as above) will match I<all> 628characters C<a> at the beginning of string, leaving no C<a> for 629C<ab> to match. In contrast, C<a*ab> will match the same as C<a+b>, 630since the match of the subgroup C<a*> is influenced by the following 631group C<ab> (see L<"Backtracking">). In particular, C<a*> inside 632C<a*ab> will match fewer characters than a standalone C<a*>, since 633this makes the tail match. 634 635An effect similar to C<< (?>pattern) >> may be achieved by writing 636C<(?=(pattern))\1>. This matches the same substring as a standalone 637C<a+>, and the following C<\1> eats the matched string; it therefore 638makes a zero-length assertion into an analogue of C<< (?>...) >>. 639(The difference between these two constructs is that the second one 640uses a capturing group, thus shifting ordinals of backreferences 641in the rest of a regular expression.) 642 643Consider this pattern: 644 645 m{ \( 646 ( 647 [^()]+ # x+ 648 | 649 \( [^()]* \) 650 )+ 651 \) 652 }x 653 654That will efficiently match a nonempty group with matching parentheses 655two levels deep or less. However, if there is no such group, it 656will take virtually forever on a long string. That's because there 657are so many different ways to split a long string into several 658substrings. This is what C<(.+)+> is doing, and C<(.+)+> is similar 659to a subpattern of the above pattern. Consider how the pattern 660above detects no-match on C<((()aaaaaaaaaaaaaaaaaa> in several 661seconds, but that each extra letter doubles this time. This 662exponential performance will make it appear that your program has 663hung. However, a tiny change to this pattern 664 665 m{ \( 666 ( 667 (?> [^()]+ ) # change x+ above to (?> x+ ) 668 | 669 \( [^()]* \) 670 )+ 671 \) 672 }x 673 674which uses C<< (?>...) >> matches exactly when the one above does (verifying 675this yourself would be a productive exercise), but finishes in a fourth 676the time when used on a similar string with 1000000 C<a>s. Be aware, 677however, that this pattern currently triggers a warning message under 678the C<use warnings> pragma or B<-w> switch saying it 679C<"matches the null string many times">): 680 681On simple groups, such as the pattern C<< (?> [^()]+ ) >>, a comparable 682effect may be achieved by negative look-ahead, as in C<[^()]+ (?! [^()] )>. 683This was only 4 times slower on a string with 1000000 C<a>s. 684 685The "grab all you can, and do not give anything back" semantic is desirable 686in many situations where on the first sight a simple C<()*> looks like 687the correct solution. Suppose we parse text with comments being delimited 688by C<#> followed by some optional (horizontal) whitespace. Contrary to 689its appearance, C<#[ \t]*> I<is not> the correct subexpression to match 690the comment delimiter, because it may "give up" some whitespace if 691the remainder of the pattern can be made to match that way. The correct 692answer is either one of these: 693 694 (?>#[ \t]*) 695 #[ \t]*(?![ \t]) 696 697For example, to grab non-empty comments into $1, one should use either 698one of these: 699 700 / (?> \# [ \t]* ) ( .+ ) /x; 701 / \# [ \t]* ( [^ \t] .* ) /x; 702 703Which one you pick depends on which of these expressions better reflects 704the above specification of comments. 705 706=item C<(?(condition)yes-pattern|no-pattern)> 707 708=item C<(?(condition)yes-pattern)> 709 710B<WARNING>: This extended regular expression feature is considered 711highly experimental, and may be changed or deleted without notice. 712 713Conditional expression. C<(condition)> should be either an integer in 714parentheses (which is valid if the corresponding pair of parentheses 715matched), or look-ahead/look-behind/evaluate zero-width assertion. 716 717For example: 718 719 m{ ( \( )? 720 [^()]+ 721 (?(1) \) ) 722 }x 723 724matches a chunk of non-parentheses, possibly included in parentheses 725themselves. 726 727=back 728 729=head2 Backtracking 730 731NOTE: This section presents an abstract approximation of regular 732expression behavior. For a more rigorous (and complicated) view of 733the rules involved in selecting a match among possible alternatives, 734see L<Combining pieces together>. 735 736A fundamental feature of regular expression matching involves the 737notion called I<backtracking>, which is currently used (when needed) 738by all regular expression quantifiers, namely C<*>, C<*?>, C<+>, 739C<+?>, C<{n,m}>, and C<{n,m}?>. Backtracking is often optimized 740internally, but the general principle outlined here is valid. 741 742For a regular expression to match, the I<entire> regular expression must 743match, not just part of it. So if the beginning of a pattern containing a 744quantifier succeeds in a way that causes later parts in the pattern to 745fail, the matching engine backs up and recalculates the beginning 746part--that's why it's called backtracking. 747 748Here is an example of backtracking: Let's say you want to find the 749word following "foo" in the string "Food is on the foo table.": 750 751 $_ = "Food is on the foo table."; 752 if ( /\b(foo)\s+(\w+)/i ) { 753 print "$2 follows $1.\n"; 754 } 755 756When the match runs, the first part of the regular expression (C<\b(foo)>) 757finds a possible match right at the beginning of the string, and loads up 758$1 with "Foo". However, as soon as the matching engine sees that there's 759no whitespace following the "Foo" that it had saved in $1, it realizes its 760mistake and starts over again one character after where it had the 761tentative match. This time it goes all the way until the next occurrence 762of "foo". The complete regular expression matches this time, and you get 763the expected output of "table follows foo." 764 765Sometimes minimal matching can help a lot. Imagine you'd like to match 766everything between "foo" and "bar". Initially, you write something 767like this: 768 769 $_ = "The food is under the bar in the barn."; 770 if ( /foo(.*)bar/ ) { 771 print "got <$1>\n"; 772 } 773 774Which perhaps unexpectedly yields: 775 776 got <d is under the bar in the > 777 778That's because C<.*> was greedy, so you get everything between the 779I<first> "foo" and the I<last> "bar". Here it's more effective 780to use minimal matching to make sure you get the text between a "foo" 781and the first "bar" thereafter. 782 783 if ( /foo(.*?)bar/ ) { print "got <$1>\n" } 784 got <d is under the > 785 786Here's another example: let's say you'd like to match a number at the end 787of a string, and you also want to keep the preceding part the match. 788So you write this: 789 790 $_ = "I have 2 numbers: 53147"; 791 if ( /(.*)(\d*)/ ) { # Wrong! 792 print "Beginning is <$1>, number is <$2>.\n"; 793 } 794 795That won't work at all, because C<.*> was greedy and gobbled up the 796whole string. As C<\d*> can match on an empty string the complete 797regular expression matched successfully. 798 799 Beginning is <I have 2 numbers: 53147>, number is <>. 800 801Here are some variants, most of which don't work: 802 803 $_ = "I have 2 numbers: 53147"; 804 @pats = qw{ 805 (.*)(\d*) 806 (.*)(\d+) 807 (.*?)(\d*) 808 (.*?)(\d+) 809 (.*)(\d+)$ 810 (.*?)(\d+)$ 811 (.*)\b(\d+)$ 812 (.*\D)(\d+)$ 813 }; 814 815 for $pat (@pats) { 816 printf "%-12s ", $pat; 817 if ( /$pat/ ) { 818 print "<$1> <$2>\n"; 819 } else { 820 print "FAIL\n"; 821 } 822 } 823 824That will print out: 825 826 (.*)(\d*) <I have 2 numbers: 53147> <> 827 (.*)(\d+) <I have 2 numbers: 5314> <7> 828 (.*?)(\d*) <> <> 829 (.*?)(\d+) <I have > <2> 830 (.*)(\d+)$ <I have 2 numbers: 5314> <7> 831 (.*?)(\d+)$ <I have 2 numbers: > <53147> 832 (.*)\b(\d+)$ <I have 2 numbers: > <53147> 833 (.*\D)(\d+)$ <I have 2 numbers: > <53147> 834 835As you see, this can be a bit tricky. It's important to realize that a 836regular expression is merely a set of assertions that gives a definition 837of success. There may be 0, 1, or several different ways that the 838definition might succeed against a particular string. And if there are 839multiple ways it might succeed, you need to understand backtracking to 840know which variety of success you will achieve. 841 842When using look-ahead assertions and negations, this can all get even 843tricker. Imagine you'd like to find a sequence of non-digits not 844followed by "123". You might try to write that as 845 846 $_ = "ABC123"; 847 if ( /^\D*(?!123)/ ) { # Wrong! 848 print "Yup, no 123 in $_\n"; 849 } 850 851But that isn't going to match; at least, not the way you're hoping. It 852claims that there is no 123 in the string. Here's a clearer picture of 853why it that pattern matches, contrary to popular expectations: 854 855 $x = 'ABC123' ; 856 $y = 'ABC445' ; 857 858 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/ ; 859 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/ ; 860 861 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/ ; 862 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/ ; 863 864This prints 865 866 2: got ABC 867 3: got AB 868 4: got ABC 869 870You might have expected test 3 to fail because it seems to a more 871general purpose version of test 1. The important difference between 872them is that test 3 contains a quantifier (C<\D*>) and so can use 873backtracking, whereas test 1 will not. What's happening is 874that you've asked "Is it true that at the start of $x, following 0 or more 875non-digits, you have something that's not 123?" If the pattern matcher had 876let C<\D*> expand to "ABC", this would have caused the whole pattern to 877fail. 878 879The search engine will initially match C<\D*> with "ABC". Then it will 880try to match C<(?!123> with "123", which fails. But because 881a quantifier (C<\D*>) has been used in the regular expression, the 882search engine can backtrack and retry the match differently 883in the hope of matching the complete regular expression. 884 885The pattern really, I<really> wants to succeed, so it uses the 886standard pattern back-off-and-retry and lets C<\D*> expand to just "AB" this 887time. Now there's indeed something following "AB" that is not 888"123". It's "C123", which suffices. 889 890We can deal with this by using both an assertion and a negation. 891We'll say that the first part in $1 must be followed both by a digit 892and by something that's not "123". Remember that the look-aheads 893are zero-width expressions--they only look, but don't consume any 894of the string in their match. So rewriting this way produces what 895you'd expect; that is, case 5 will fail, but case 6 succeeds: 896 897 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/ ; 898 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/ ; 899 900 6: got ABC 901 902In other words, the two zero-width assertions next to each other work as though 903they're ANDed together, just as you'd use any built-in assertions: C</^$/> 904matches only if you're at the beginning of the line AND the end of the 905line simultaneously. The deeper underlying truth is that juxtaposition in 906regular expressions always means AND, except when you write an explicit OR 907using the vertical bar. C</ab/> means match "a" AND (then) match "b", 908although the attempted matches are made at different positions because "a" 909is not a zero-width assertion, but a one-width assertion. 910 911B<WARNING>: particularly complicated regular expressions can take 912exponential time to solve because of the immense number of possible 913ways they can use backtracking to try match. For example, without 914internal optimizations done by the regular expression engine, this will 915take a painfully long time to run: 916 917 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/ 918 919And if you used C<*>'s in the internal groups instead of limiting them 920to 0 through 5 matches, then it would take forever--or until you ran 921out of stack space. Moreover, these internal optimizations are not 922always applicable. For example, if you put C<{0,5}> instead of C<*> 923on the external group, no current optimization is applicable, and the 924match takes a long time to finish. 925 926A powerful tool for optimizing such beasts is what is known as an 927"independent group", 928which does not backtrack (see L<C<< (?>pattern) >>>). Note also that 929zero-length look-ahead/look-behind assertions will not backtrack to make 930the tail match, since they are in "logical" context: only 931whether they match is considered relevant. For an example 932where side-effects of look-ahead I<might> have influenced the 933following match, see L<C<< (?>pattern) >>>. 934 935=head2 Version 8 Regular Expressions 936 937In case you're not familiar with the "regular" Version 8 regex 938routines, here are the pattern-matching rules not described above. 939 940Any single character matches itself, unless it is a I<metacharacter> 941with a special meaning described here or above. You can cause 942characters that normally function as metacharacters to be interpreted 943literally by prefixing them with a "\" (e.g., "\." matches a ".", not any 944character; "\\" matches a "\"). A series of characters matches that 945series of characters in the target string, so the pattern C<blurfl> 946would match "blurfl" in the target string. 947 948You can specify a character class, by enclosing a list of characters 949in C<[]>, which will match any one character from the list. If the 950first character after the "[" is "^", the class matches any character not 951in the list. Within a list, the "-" character specifies a 952range, so that C<a-z> represents all characters between "a" and "z", 953inclusive. If you want either "-" or "]" itself to be a member of a 954class, put it at the start of the list (possibly after a "^"), or 955escape it with a backslash. "-" is also taken literally when it is 956at the end of the list, just before the closing "]". (The 957following all specify the same class of three characters: C<[-az]>, 958C<[az-]>, and C<[a\-z]>. All are different from C<[a-z]>, which 959specifies a class containing twenty-six characters, even on EBCDIC 960based coded character sets.) Also, if you try to use the character 961classes C<\w>, C<\W>, C<\s>, C<\S>, C<\d>, or C<\D> as endpoints of 962a range, that's not a range, the "-" is understood literally. 963 964Note also that the whole range idea is rather unportable between 965character sets--and even within character sets they may cause results 966you probably didn't expect. A sound principle is to use only ranges 967that begin from and end at either alphabets of equal case ([a-e], 968[A-E]), or digits ([0-9]). Anything else is unsafe. If in doubt, 969spell out the character sets in full. 970 971Characters may be specified using a metacharacter syntax much like that 972used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return, 973"\f" a form feed, etc. More generally, \I<nnn>, where I<nnn> is a string 974of octal digits, matches the character whose coded character set value 975is I<nnn>. Similarly, \xI<nn>, where I<nn> are hexadecimal digits, 976matches the character whose numeric value is I<nn>. The expression \cI<x> 977matches the character control-I<x>. Finally, the "." metacharacter 978matches any character except "\n" (unless you use C</s>). 979 980You can specify a series of alternatives for a pattern using "|" to 981separate them, so that C<fee|fie|foe> will match any of "fee", "fie", 982or "foe" in the target string (as would C<f(e|i|o)e>). The 983first alternative includes everything from the last pattern delimiter 984("(", "[", or the beginning of the pattern) up to the first "|", and 985the last alternative contains everything from the last "|" to the next 986pattern delimiter. That's why it's common practice to include 987alternatives in parentheses: to minimize confusion about where they 988start and end. 989 990Alternatives are tried from left to right, so the first 991alternative found for which the entire expression matches, is the one that 992is chosen. This means that alternatives are not necessarily greedy. For 993example: when matching C<foo|foot> against "barefoot", only the "foo" 994part will match, as that is the first alternative tried, and it successfully 995matches the target string. (This might not seem important, but it is 996important when you are capturing matched text using parentheses.) 997 998Also remember that "|" is interpreted as a literal within square brackets, 999so if you write C<[fee|fie|foe]> you're really only matching C<[feio|]>. 1000 1001Within a pattern, you may designate subpatterns for later reference 1002by enclosing them in parentheses, and you may refer back to the 1003I<n>th subpattern later in the pattern using the metacharacter 1004\I<n>. Subpatterns are numbered based on the left to right order 1005of their opening parenthesis. A backreference matches whatever 1006actually matched the subpattern in the string being examined, not 1007the rules for that subpattern. Therefore, C<(0|0x)\d*\s\1\d*> will 1008match "0x1234 0x4321", but not "0x1234 01234", because subpattern 10091 matched "0x", even though the rule C<0|0x> could potentially match 1010the leading 0 in the second number. 1011 1012=head2 Warning on \1 vs $1 1013 1014Some people get too used to writing things like: 1015 1016 $pattern =~ s/(\W)/\\\1/g; 1017 1018This is grandfathered for the RHS of a substitute to avoid shocking the 1019B<sed> addicts, but it's a dirty habit to get into. That's because in 1020PerlThink, the righthand side of a C<s///> is a double-quoted string. C<\1> in 1021the usual double-quoted string means a control-A. The customary Unix 1022meaning of C<\1> is kludged in for C<s///>. However, if you get into the habit 1023of doing that, you get yourself into trouble if you then add an C</e> 1024modifier. 1025 1026 s/(\d+)/ \1 + 1 /eg; # causes warning under -w 1027 1028Or if you try to do 1029 1030 s/(\d+)/\1000/; 1031 1032You can't disambiguate that by saying C<\{1}000>, whereas you can fix it with 1033C<${1}000>. The operation of interpolation should not be confused 1034with the operation of matching a backreference. Certainly they mean two 1035different things on the I<left> side of the C<s///>. 1036 1037=head2 Repeated patterns matching zero-length substring 1038 1039B<WARNING>: Difficult material (and prose) ahead. This section needs a rewrite. 1040 1041Regular expressions provide a terse and powerful programming language. As 1042with most other power tools, power comes together with the ability 1043to wreak havoc. 1044 1045A common abuse of this power stems from the ability to make infinite 1046loops using regular expressions, with something as innocuous as: 1047 1048 'foo' =~ m{ ( o? )* }x; 1049 1050The C<o?> can match at the beginning of C<'foo'>, and since the position 1051in the string is not moved by the match, C<o?> would match again and again 1052because of the C<*> modifier. Another common way to create a similar cycle 1053is with the looping modifier C<//g>: 1054 1055 @matches = ( 'foo' =~ m{ o? }xg ); 1056 1057or 1058 1059 print "match: <$&>\n" while 'foo' =~ m{ o? }xg; 1060 1061or the loop implied by split(). 1062 1063However, long experience has shown that many programming tasks may 1064be significantly simplified by using repeated subexpressions that 1065may match zero-length substrings. Here's a simple example being: 1066 1067 @chars = split //, $string; # // is not magic in split 1068 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// / 1069 1070Thus Perl allows such constructs, by I<forcefully breaking 1071the infinite loop>. The rules for this are different for lower-level 1072loops given by the greedy modifiers C<*+{}>, and for higher-level 1073ones like the C</g> modifier or split() operator. 1074 1075The lower-level loops are I<interrupted> (that is, the loop is 1076broken) when Perl detects that a repeated expression matched a 1077zero-length substring. Thus 1078 1079 m{ (?: NON_ZERO_LENGTH | ZERO_LENGTH )* }x; 1080 1081is made equivalent to 1082 1083 m{ (?: NON_ZERO_LENGTH )* 1084 | 1085 (?: ZERO_LENGTH )? 1086 }x; 1087 1088The higher level-loops preserve an additional state between iterations: 1089whether the last match was zero-length. To break the loop, the following 1090match after a zero-length match is prohibited to have a length of zero. 1091This prohibition interacts with backtracking (see L<"Backtracking">), 1092and so the I<second best> match is chosen if the I<best> match is of 1093zero length. 1094 1095For example: 1096 1097 $_ = 'bar'; 1098 s/\w??/<$&>/g; 1099 1100results in C<< <><b><><a><><r><> >>. At each position of the string the best 1101match given by non-greedy C<??> is the zero-length match, and the I<second 1102best> match is what is matched by C<\w>. Thus zero-length matches 1103alternate with one-character-long matches. 1104 1105Similarly, for repeated C<m/()/g> the second-best match is the match at the 1106position one notch further in the string. 1107 1108The additional state of being I<matched with zero-length> is associated with 1109the matched string, and is reset by each assignment to pos(). 1110Zero-length matches at the end of the previous match are ignored 1111during C<split>. 1112 1113=head2 Combining pieces together 1114 1115Each of the elementary pieces of regular expressions which were described 1116before (such as C<ab> or C<\Z>) could match at most one substring 1117at the given position of the input string. However, in a typical regular 1118expression these elementary pieces are combined into more complicated 1119patterns using combining operators C<ST>, C<S|T>, C<S*> etc 1120(in these examples C<S> and C<T> are regular subexpressions). 1121 1122Such combinations can include alternatives, leading to a problem of choice: 1123if we match a regular expression C<a|ab> against C<"abc">, will it match 1124substring C<"a"> or C<"ab">? One way to describe which substring is 1125actually matched is the concept of backtracking (see L<"Backtracking">). 1126However, this description is too low-level and makes you think 1127in terms of a particular implementation. 1128 1129Another description starts with notions of "better"/"worse". All the 1130substrings which may be matched by the given regular expression can be 1131sorted from the "best" match to the "worst" match, and it is the "best" 1132match which is chosen. This substitutes the question of "what is chosen?" 1133by the question of "which matches are better, and which are worse?". 1134 1135Again, for elementary pieces there is no such question, since at most 1136one match at a given position is possible. This section describes the 1137notion of better/worse for combining operators. In the description 1138below C<S> and C<T> are regular subexpressions. 1139 1140=over 4 1141 1142=item C<ST> 1143 1144Consider two possible matches, C<AB> and C<A'B'>, C<A> and C<A'> are 1145substrings which can be matched by C<S>, C<B> and C<B'> are substrings 1146which can be matched by C<T>. 1147 1148If C<A> is better match for C<S> than C<A'>, C<AB> is a better 1149match than C<A'B'>. 1150 1151If C<A> and C<A'> coincide: C<AB> is a better match than C<AB'> if 1152C<B> is better match for C<T> than C<B'>. 1153 1154=item C<S|T> 1155 1156When C<S> can match, it is a better match than when only C<T> can match. 1157 1158Ordering of two matches for C<S> is the same as for C<S>. Similar for 1159two matches for C<T>. 1160 1161=item C<S{REPEAT_COUNT}> 1162 1163Matches as C<SSS...S> (repeated as many times as necessary). 1164 1165=item C<S{min,max}> 1166 1167Matches as C<S{max}|S{max-1}|...|S{min+1}|S{min}>. 1168 1169=item C<S{min,max}?> 1170 1171Matches as C<S{min}|S{min+1}|...|S{max-1}|S{max}>. 1172 1173=item C<S?>, C<S*>, C<S+> 1174 1175Same as C<S{0,1}>, C<S{0,BIG_NUMBER}>, C<S{1,BIG_NUMBER}> respectively. 1176 1177=item C<S??>, C<S*?>, C<S+?> 1178 1179Same as C<S{0,1}?>, C<S{0,BIG_NUMBER}?>, C<S{1,BIG_NUMBER}?> respectively. 1180 1181=item C<< (?>S) >> 1182 1183Matches the best match for C<S> and only that. 1184 1185=item C<(?=S)>, C<(?<=S)> 1186 1187Only the best match for C<S> is considered. (This is important only if 1188C<S> has capturing parentheses, and backreferences are used somewhere 1189else in the whole regular expression.) 1190 1191=item C<(?!S)>, C<(?<!S)> 1192 1193For this grouping operator there is no need to describe the ordering, since 1194only whether or not C<S> can match is important. 1195 1196=item C<(??{ EXPR })> 1197 1198The ordering is the same as for the regular expression which is 1199the result of EXPR. 1200 1201=item C<(?(condition)yes-pattern|no-pattern)> 1202 1203Recall that which of C<yes-pattern> or C<no-pattern> actually matches is 1204already determined. The ordering of the matches is the same as for the 1205chosen subexpression. 1206 1207=back 1208 1209The above recipes describe the ordering of matches I<at a given position>. 1210One more rule is needed to understand how a match is determined for the 1211whole regular expression: a match at an earlier position is always better 1212than a match at a later position. 1213 1214=head2 Creating custom RE engines 1215 1216Overloaded constants (see L<overload>) provide a simple way to extend 1217the functionality of the RE engine. 1218 1219Suppose that we want to enable a new RE escape-sequence C<\Y|> which 1220matches at boundary between white-space characters and non-whitespace 1221characters. Note that C<(?=\S)(?<!\S)|(?!\S)(?<=\S)> matches exactly 1222at these positions, so we want to have each C<\Y|> in the place of the 1223more complicated version. We can create a module C<customre> to do 1224this: 1225 1226 package customre; 1227 use overload; 1228 1229 sub import { 1230 shift; 1231 die "No argument to customre::import allowed" if @_; 1232 overload::constant 'qr' => \&convert; 1233 } 1234 1235 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"} 1236 1237 my %rules = ( '\\' => '\\', 1238 'Y|' => qr/(?=\S)(?<!\S)|(?!\S)(?<=\S)/ ); 1239 sub convert { 1240 my $re = shift; 1241 $re =~ s{ 1242 \\ ( \\ | Y . ) 1243 } 1244 { $rules{$1} or invalid($re,$1) }sgex; 1245 return $re; 1246 } 1247 1248Now C<use customre> enables the new escape in constant regular 1249expressions, i.e., those without any runtime variable interpolations. 1250As documented in L<overload>, this conversion will work only over 1251literal parts of regular expressions. For C<\Y|$re\Y|> the variable 1252part of this regular expression needs to be converted explicitly 1253(but only if the special meaning of C<\Y|> should be enabled inside $re): 1254 1255 use customre; 1256 $re = <>; 1257 chomp $re; 1258 $re = customre::convert $re; 1259 /\Y|$re\Y|/; 1260 1261=head1 BUGS 1262 1263This document varies from difficult to understand to completely 1264and utterly opaque. The wandering prose riddled with jargon is 1265hard to fathom in several places. 1266 1267This document needs a rewrite that separates the tutorial content 1268from the reference content. 1269 1270=head1 SEE ALSO 1271 1272L<perlop/"Regexp Quote-Like Operators">. 1273 1274L<perlop/"Gory details of parsing quoted constructs">. 1275 1276L<perlfaq6>. 1277 1278L<perlfunc/pos>. 1279 1280L<perllocale>. 1281 1282L<perlebcdic>. 1283 1284I<Mastering Regular Expressions> by Jeffrey Friedl, published 1285by O'Reilly and Associates. 1286