1PERLRE(1) Perl Programmers Reference Guide PERLRE(1)
2
6 perlre - Perl regular expressions
7
9 This page describes the syntax of regular expressions in Perl.
10 If you haven't used regular expressions before, a quick-start introduc‐
12 tion is available in perlrequick, and a longer tutorial introduction is
13 available in perlretut.
14 For reference on how regular expressions are used in matching opera‐
16 tions, plus various examples of the same, see discussions of "m//",
17 "s///", "qr//" and "??" in "Regexp Quote-Like Operators" in perlop.
18 Matching operations can have various modifiers. Modifiers that relate
20 to the interpretation of the regular expression inside are listed
21 below. Modifiers that alter the way a regular expression is used by
22 Perl are detailed in "Regexp Quote-Like Operators" in perlop and "Gory
23 details of parsing quoted constructs" in perlop.
24 i Do case-insensitive pattern matching.
26 If "use locale" is in effect, the case map is taken from the cur‐
28 rent locale. See perllocale.
29 m Treat string as multiple lines. That is, change "^" and "$" from
31 matching the start or end of the string to matching the start or
32 end of any line anywhere within the string.
33 s Treat string as single line. That is, change "." to match any
35 character whatsoever, even a newline, which normally it would not
36 match.
37 The "/s" and "/m" modifiers both override the $* setting. That is,
39 no matter what $* contains, "/s" without "/m" will force "^" to
40 match only at the beginning of the string and "$" to match only at
41 the end (or just before a newline at the end) of the string.
42 Together, as /ms, they let the "." match any character whatsoever,
43 while still allowing "^" and "$" to match, respectively, just after
44 and just before newlines within the string.
45 x Extend your pattern's legibility by permitting whitespace and com‐
47 ments.
48 These are usually written as "the "/x" modifier", even though the
50 delimiter in question might not really be a slash. Any of these modi‐
51 fiers may also be embedded within the regular expression itself using
52 the "(?...)" construct. See below.
53 The "/x" modifier itself needs a little more explanation. It tells the
55 regular expression parser to ignore whitespace that is neither back‐
56 slashed nor within a character class. You can use this to break up
57 your regular expression into (slightly) more readable parts. The "#"
58 character is also treated as a metacharacter introducing a comment,
59 just as in ordinary Perl code. This also means that if you want real
60 whitespace or "#" characters in the pattern (outside a character class,
61 where they are unaffected by "/x"), that you'll either have to escape
62 them or encode them using octal or hex escapes. Taken together, these
63 features go a long way towards making Perl's regular expressions more
64 readable. Note that you have to be careful not to include the pattern
65 delimiter in the comment--perl has no way of knowing you did not intend
66 to close the pattern early. See the C-comment deletion code in perlop.
67 Regular Expressions
69 The patterns used in Perl pattern matching derive from supplied in the
71 Version 8 regex routines. (The routines are derived (distantly) from
72 Henry Spencer's freely redistributable reimplementation of the V8 rou‐
73 tines.) See "Version 8 Regular Expressions" for details.
74 In particular the following metacharacters have their standard
76 egrep-ish meanings:
77 \ Quote the next metacharacter
79 ^ Match the beginning of the line
80 . Match any character (except newline)
81 $ Match the end of the line (or before newline at the end)
82 ⎪ Alternation
83 () Grouping
84 [] Character class
85 By default, the "^" character is guaranteed to match only the beginning
87 of the string, the "$" character only the end (or before the newline at
88 the end), and Perl does certain optimizations with the assumption that
89 the string contains only one line. Embedded newlines will not be
90 matched by "^" or "$". You may, however, wish to treat a string as a
91 multi-line buffer, such that the "^" will match after any newline
92 within the string, and "$" will match before any newline. At the cost
93 of a little more overhead, you can do this by using the /m modifier on
94 the pattern match operator. (Older programs did this by setting $*,
95 but this practice is now deprecated.)
96 To simplify multi-line substitutions, the "." character never matches a
98 newline unless you use the "/s" modifier, which in effect tells Perl to
99 pretend the string is a single line--even if it isn't. The "/s" modi‐
100 fier also overrides the setting of $*, in case you have some (badly
101 behaved) older code that sets it in another module.
102 The following standard quantifiers are recognized:
104 * Match 0 or more times
106 + Match 1 or more times
107 ? Match 1 or 0 times
108 {n} Match exactly n times
109 {n,} Match at least n times
110 {n,m} Match at least n but not more than m times
111 (If a curly bracket occurs in any other context, it is treated as a
113 regular character. In particular, the lower bound is not optional.)
114 The "*" modifier is equivalent to "{0,}", the "+" modifier to "{1,}",
115 and the "?" modifier to "{0,1}". n and m are limited to integral val‐
116 ues less than a preset limit defined when perl is built. This is usu‐
117 ally 32766 on the most common platforms. The actual limit can be seen
118 in the error message generated by code such as this:
119 $_ **= $_ , / {$_} / for 2 .. 42;
121 By default, a quantified subpattern is "greedy", that is, it will match
123 as many times as possible (given a particular starting location) while
124 still allowing the rest of the pattern to match. If you want it to
125 match the minimum number of times possible, follow the quantifier with
126 a "?". Note that the meanings don't change, just the "greediness":
127 *? Match 0 or more times
129 +? Match 1 or more times
130 ?? Match 0 or 1 time
131 {n}? Match exactly n times
132 {n,}? Match at least n times
133 {n,m}? Match at least n but not more than m times
134 Because patterns are processed as double quoted strings, the following
136 also work:
137 \t tab (HT, TAB)
139 \n newline (LF, NL)
140 \r return (CR)
141 \f form feed (FF)
142 \a alarm (bell) (BEL)
143 \e escape (think troff) (ESC)
144 \033 octal char (think of a PDP-11)
145 \x1B hex char
146 \x{263a} wide hex char (Unicode SMILEY)
147 \c[ control char
148 \N{name} named char
149 \l lowercase next char (think vi)
150 \u uppercase next char (think vi)
151 \L lowercase till \E (think vi)
152 \U uppercase till \E (think vi)
153 \E end case modification (think vi)
154 \Q quote (disable) pattern metacharacters till \E
155 If "use locale" is in effect, the case map used by "\l", "\L", "\u" and
157 "\U" is taken from the current locale. See perllocale. For documenta‐
158 tion of "\N{name}", see charnames.
159 You cannot include a literal "$" or "@" within a "\Q" sequence. An
161 unescaped "$" or "@" interpolates the corresponding variable, while
162 escaping will cause the literal string "\$" to be matched. You'll need
163 to write something like "m/\Quser\E\@\Qhost/".
164 In addition, Perl defines the following:
166 \w Match a "word" character (alphanumeric plus "_")
168 \W Match a non-"word" character
169 \s Match a whitespace character
170 \S Match a non-whitespace character
171 \d Match a digit character
172 \D Match a non-digit character
173 \pP Match P, named property. Use \p{Prop} for longer names.
174 \PP Match non-P
175 \X Match eXtended Unicode "combining character sequence",
176 equivalent to (?:\PM\pM*)
177 \C Match a single C char (octet) even under Unicode.
178 NOTE: breaks up characters into their UTF-8 bytes,
179 so you may end up with malformed pieces of UTF-8.
180 Unsupported in lookbehind.
181 A "\w" matches a single alphanumeric character (an alphabetic charac‐
183 ter, or a decimal digit) or "_", not a whole word. Use "\w+" to match
184 a string of Perl-identifier characters (which isn't the same as match‐
185 ing an English word). If "use locale" is in effect, the list of alpha‐
186 betic characters generated by "\w" is taken from the current locale.
187 See perllocale. You may use "\w", "\W", "\s", "\S", "\d", and "\D"
188 within character classes, but if you try to use them as endpoints of a
189 range, that's not a range, the "-" is understood literally. If Unicode
190 is in effect, "\s" matches also "\x{85}", "\x{2028}, and "\x{2029}",
191 see perlunicode for more details about "\pP", "\PP", and "\X", and per‐
192 luniintro about Unicode in general. You can define your own "\p" and
193 "\P" properties, see perlunicode.
194 The POSIX character class syntax
196 [:class:]
198 is also available. The available classes and their backslash equiva‐
200 lents (if available) are as follows:
201 alpha
203 alnum
204 ascii
205 blank [1]
206 cntrl
207 digit \d
208 graph
209 lower
210 print
211 punct
212 space \s [2]
213 upper
214 word \w [3]
215 xdigit
216 [1] A GNU extension equivalent to "[ \t]", "all horizontal whitespace".
218 [2] Not exactly equivalent to "\s" since the "[[:space:]]" includes
220 also the (very rare) "vertical tabulator", "\ck", chr(11).
221 [3] A Perl extension, see above.
223 For example use "[:upper:]" to match all the uppercase characters.
225 Note that the "[]" are part of the "[::]" construct, not part of the
226 whole character class. For example:
227 [01[:alpha:]%]
229 matches zero, one, any alphabetic character, and the percentage sign.
231 The following equivalences to Unicode \p{} constructs and equivalent
233 backslash character classes (if available), will hold:
234 [:...:] \p{...} backslash
236 alpha IsAlpha
238 alnum IsAlnum
239 ascii IsASCII
240 blank IsSpace
241 cntrl IsCntrl
242 digit IsDigit \d
243 graph IsGraph
244 lower IsLower
245 print IsPrint
246 punct IsPunct
247 space IsSpace
248 IsSpacePerl \s
249 upper IsUpper
250 word IsWord
251 xdigit IsXDigit
252 For example "[:lower:]" and "\p{IsLower}" are equivalent.
254 If the "utf8" pragma is not used but the "locale" pragma is, the
256 classes correlate with the usual isalpha(3) interface (except for
257 "word" and "blank").
258 The assumedly non-obviously named classes are:
260 cntrl
262 Any control character. Usually characters that don't produce out‐
263 put as such but instead control the terminal somehow: for example
264 newline and backspace are control characters. All characters with
265 ord() less than 32 are most often classified as control characters
266 (assuming ASCII, the ISO Latin character sets, and Unicode), as is
267 the character with the ord() value of 127 ("DEL").
268 graph
270 Any alphanumeric or punctuation (special) character.
271 print
273 Any alphanumeric or punctuation (special) character or the space
274 character.
275 punct
277 Any punctuation (special) character.
278 xdigit
280 Any hexadecimal digit. Though this may feel silly ([0-9A-Fa-f]
281 would work just fine) it is included for completeness.
282 You can negate the [::] character classes by prefixing the class name
284 with a '^'. This is a Perl extension. For example:
285 POSIX traditional Unicode
287 [:^digit:] \D \P{IsDigit}
289 [:^space:] \S \P{IsSpace}
290 [:^word:] \W \P{IsWord}
291 Perl respects the POSIX standard in that POSIX character classes are
293 only supported within a character class. The POSIX character classes
294 [.cc.] and [=cc=] are recognized but not supported and trying to use
295 them will cause an error.
296 Perl defines the following zero-width assertions:
298 \b Match a word boundary
300 \B Match a non-(word boundary)
301 \A Match only at beginning of string
302 \Z Match only at end of string, or before newline at the end
303 \z Match only at end of string
304 \G Match only at pos() (e.g. at the end-of-match position
305 of prior m//g)
306 A word boundary ("\b") is a spot between two characters that has a "\w"
308 on one side of it and a "\W" on the other side of it (in either order),
309 counting the imaginary characters off the beginning and end of the
310 string as matching a "\W". (Within character classes "\b" represents
311 backspace rather than a word boundary, just as it normally does in any
312 double-quoted string.) The "\A" and "\Z" are just like "^" and "$",
313 except that they won't match multiple times when the "/m" modifier is
314 used, while "^" and "$" will match at every internal line boundary. To
315 match the actual end of the string and not ignore an optional trailing
316 newline, use "\z".
317 The "\G" assertion can be used to chain global matches (using "m//g"),
319 as described in "Regexp Quote-Like Operators" in perlop. It is also
320 useful when writing "lex"-like scanners, when you have several patterns
321 that you want to match against consequent substrings of your string,
322 see the previous reference. The actual location where "\G" will match
323 can also be influenced by using "pos()" as an lvalue: see "pos" in
324 perlfunc. Currently "\G" is only fully supported when anchored to the
325 start of the pattern; while it is permitted to use it elsewhere, as in
326 "/(?<=\G..)./g", some such uses ("/.\G/g", for example) currently cause
327 problems, and it is recommended that you avoid such usage for now.
328 The bracketing construct "( ... )" creates capture buffers. To refer
330 to the digit'th buffer use \<digit> within the match. Outside the
331 match use "$" instead of "\". (The \<digit> notation works in certain
332 circumstances outside the match. See the warning below about \1 vs $1
333 for details.) Referring back to another part of the match is called a
334 backreference.
335 There is no limit to the number of captured substrings that you may
337 use. However Perl also uses \10, \11, etc. as aliases for \010, \011,
338 etc. (Recall that 0 means octal, so \011 is the character at number 9
339 in your coded character set; which would be the 10th character, a hori‐
340 zontal tab under ASCII.) Perl resolves this ambiguity by interpreting
341 \10 as a backreference only if at least 10 left parentheses have opened
342 before it. Likewise \11 is a backreference only if at least 11 left
343 parentheses have opened before it. And so on. \1 through \9 are
344 always interpreted as backreferences.
345 Examples:
347 s/^([^ ]*) *([^ ]*)/$2 $1/; # swap first two words
349 if (/(.)\1/) { # find first doubled char
351 print "'$1' is the first doubled character\n";
352 }
353 if (/Time: (..):(..):(..)/) { # parse out values
355 $hours = $1;
356 $minutes = $2;
357 $seconds = $3;
358 }
359 Several special variables also refer back to portions of the previous
361 match. $+ returns whatever the last bracket match matched. $& returns
362 the entire matched string. (At one point $0 did also, but now it
363 returns the name of the program.) $` returns everything before the
364 matched string. $' returns everything after the matched string. And
365 $^N contains whatever was matched by the most-recently closed group
366 (submatch). $^N can be used in extended patterns (see below), for exam‐
367 ple to assign a submatch to a variable.
368 The numbered match variables ($1, $2, $3, etc.) and the related punctu‐
370 ation set ($+, $&, $`, $', and $^N) are all dynamically scoped until
371 the end of the enclosing block or until the next successful match,
372 whichever comes first. (See "Compound Statements" in perlsyn.)
373 NOTE: failed matches in Perl do not reset the match variables, which
375 makes it easier to write code that tests for a series of more specific
376 cases and remembers the best match.
377 WARNING: Once Perl sees that you need one of $&, $`, or $' anywhere in
379 the program, it has to provide them for every pattern match. This may
380 substantially slow your program. Perl uses the same mechanism to pro‐
381 duce $1, $2, etc, so you also pay a price for each pattern that con‐
382 tains capturing parentheses. (To avoid this cost while retaining the
383 grouping behaviour, use the extended regular expression "(?: ... )"
384 instead.) But if you never use $&, $` or $', then patterns without
385 capturing parentheses will not be penalized. So avoid $&, $', and $`
386 if you can, but if you can't (and some algorithms really appreciate
387 them), once you've used them once, use them at will, because you've
388 already paid the price. As of 5.005, $& is not so costly as the other
389 two.
390 Backslashed metacharacters in Perl are alphanumeric, such as "\b",
392 "\w", "\n". Unlike some other regular expression languages, there are
393 no backslashed symbols that aren't alphanumeric. So anything that
394 looks like \\, \(, \), \<, \>, \{, or \} is always interpreted as a
395 literal character, not a metacharacter. This was once used in a common
396 idiom to disable or quote the special meanings of regular expression
397 metacharacters in a string that you want to use for a pattern. Simply
398 quote all non-"word" characters:
399 $pattern =~ s/(\W)/\\$1/g;
401 (If "use locale" is set, then this depends on the current locale.)
403 Today it is more common to use the quotemeta() function or the "\Q"
404 metaquoting escape sequence to disable all metacharacters' special
405 meanings like this:
406 /$unquoted\Q$quoted\E$unquoted/
408 Beware that if you put literal backslashes (those not inside interpo‐
410 lated variables) between "\Q" and "\E", double-quotish backslash inter‐
411 polation may lead to confusing results. If you need to use literal
412 backslashes within "\Q...\E", consult "Gory details of parsing quoted
413 constructs" in perlop.
414 Extended Patterns
416 Perl also defines a consistent extension syntax for features not found
418 in standard tools like awk and lex. The syntax is a pair of parenthe‐
419 ses with a question mark as the first thing within the parentheses.
420 The character after the question mark indicates the extension.
421 The stability of these extensions varies widely. Some have been part
423 of the core language for many years. Others are experimental and may
424 change without warning or be completely removed. Check the documenta‐
425 tion on an individual feature to verify its current status.
426 A question mark was chosen for this and for the minimal-matching con‐
428 struct because 1) question marks are rare in older regular expressions,
429 and 2) whenever you see one, you should stop and "question" exactly
430 what is going on. That's psychology...
431 "(?#text)"
433 A comment. The text is ignored. If the "/x" modifier
434 enables whitespace formatting, a simple "#" will suffice.
435 Note that Perl closes the comment as soon as it sees a ")",
436 so there is no way to put a literal ")" in the comment.
437 "(?imsx-imsx)"
439 One or more embedded pattern-match modifiers, to be turned on
440 (or turned off, if preceded by "-") for the remainder of the
441 pattern or the remainder of the enclosing pattern group (if
442 any). This is particularly useful for dynamic patterns, such
443 as those read in from a configuration file, read in as an
444 argument, are specified in a table somewhere, etc. Consider
445 the case that some of which want to be case sensitive and
446 some do not. The case insensitive ones need to include
447 merely "(?i)" at the front of the pattern. For example:
448 $pattern = "foobar";
450 if ( /$pattern/i ) { }
451 # more flexible:
453 $pattern = "(?i)foobar";
455 if ( /$pattern/ ) { }
456 These modifiers are restored at the end of the enclosing
458 group. For example,
459 ( (?i) blah ) \s+ \1
461 will match a repeated (including the case!) word "blah" in
463 any case, assuming "x" modifier, and no "i" modifier outside
464 this group.
465 "(?:pattern)"
467 "(?imsx-imsx:pattern)"
468 This is for clustering, not capturing; it groups subexpres‐
469 sions like "()", but doesn't make backreferences as "()"
470 does. So
471 @fields = split(/\b(?:a⎪b⎪c)\b/)
473 is like
475 @fields = split(/\b(a⎪b⎪c)\b/)
477 but doesn't spit out extra fields. It's also cheaper not to
479 capture characters if you don't need to.
480 Any letters between "?" and ":" act as flags modifiers as
482 with "(?imsx-imsx)". For example,
483 /(?s-i:more.*than).*million/i
485 is equivalent to the more verbose
487 /(?:(?s-i)more.*than).*million/i
489 "(?=pattern)"
491 A zero-width positive look-ahead assertion. For example,
492 "/\w+(?=\t)/" matches a word followed by a tab, without
493 including the tab in $&.
494 "(?!pattern)"
496 A zero-width negative look-ahead assertion. For example
497 "/foo(?!bar)/" matches any occurrence of "foo" that isn't
498 followed by "bar". Note however that look-ahead and look-
499 behind are NOT the same thing. You cannot use this for
500 look-behind.
501 If you are looking for a "bar" that isn't preceded by a
503 "foo", "/(?!foo)bar/" will not do what you want. That's
504 because the "(?!foo)" is just saying that the next thing can‐
505 not be "foo"--and it's not, it's a "bar", so "foobar" will
506 match. You would have to do something like "/(?!foo)...bar/"
507 for that. We say "like" because there's the case of your
508 "bar" not having three characters before it. You could cover
509 that this way: "/(?:(?!foo)...⎪^.{0,2})bar/". Sometimes it's
510 still easier just to say:
511 if (/bar/ && $` !~ /foo$/)
513 For look-behind see below.
515 "(?<=pattern)"
517 A zero-width positive look-behind assertion. For example,
518 "/(?<=\t)\w+/" matches a word that follows a tab, without
519 including the tab in $&. Works only for fixed-width
520 look-behind.
521 "(?<!pattern)"
523 A zero-width negative look-behind assertion. For example
524 "/(?<!bar)foo/" matches any occurrence of "foo" that does not
525 follow "bar". Works only for fixed-width look-behind.
526 "(?{ code })"
528 WARNING: This extended regular expression feature is consid‐
529 ered highly experimental, and may be changed or deleted with‐
530 out notice.
531 This zero-width assertion evaluates any embedded Perl code.
533 It always succeeds, and its "code" is not interpolated. Cur‐
534 rently, the rules to determine where the "code" ends are
535 somewhat convoluted.
536 This feature can be used together with the special variable
538 $^N to capture the results of submatches in variables without
539 having to keep track of the number of nested parentheses. For
540 example:
541 $_ = "The brown fox jumps over the lazy dog";
543 /the (\S+)(?{ $color = $^N }) (\S+)(?{ $animal = $^N })/i;
544 print "color = $color, animal = $animal\n";
545 Inside the "(?{...})" block, $_ refers to the string the reg‐
547 ular expression is matching against. You can also use "pos()"
548 to know what is the current position of matching within this
549 string.
550 The "code" is properly scoped in the following sense: If the
552 assertion is backtracked (compare "Backtracking"), all
553 changes introduced after "local"ization are undone, so that
554 $_ = 'a' x 8;
556 m<
557 (?{ $cnt = 0 }) # Initialize $cnt.
558 (
559 a
560 (?{
561 local $cnt = $cnt + 1; # Update $cnt, backtracking-safe.
562 })
563 )*
564 aaaa
565 (?{ $res = $cnt }) # On success copy to non-localized
566 # location.
567 >x;
568 will set "$res = 4". Note that after the match, $cnt returns
570 to the globally introduced value, because the scopes that
571 restrict "local" operators are unwound.
572 This assertion may be used as a "(?(condition)yes-pat‐
574 tern⎪no-pattern)" switch. If not used in this way, the
575 result of evaluation of "code" is put into the special vari‐
576 able $^R. This happens immediately, so $^R can be used from
577 other "(?{ code })" assertions inside the same regular
578 expression.
579 The assignment to $^R above is properly localized, so the old
581 value of $^R is restored if the assertion is backtracked;
582 compare "Backtracking".
583 For reasons of security, this construct is forbidden if the
585 regular expression involves run-time interpolation of vari‐
586 ables, unless the perilous "use re 'eval'" pragma has been
587 used (see re), or the variables contain results of "qr//"
588 operator (see "qr/STRING/imosx" in perlop).
589 This restriction is because of the wide-spread and remarkably
591 convenient custom of using run-time determined strings as
592 patterns. For example:
593 $re = <>;
595 chomp $re;
596 $string =~ /$re/;
597 Before Perl knew how to execute interpolated code within a
599 pattern, this operation was completely safe from a security
600 point of view, although it could raise an exception from an
601 illegal pattern. If you turn on the "use re 'eval'", though,
602 it is no longer secure, so you should only do so if you are
603 also using taint checking. Better yet, use the carefully
604 constrained evaluation within a Safe compartment. See
605 perlsec for details about both these mechanisms.
606 "(??{ code })"
608 WARNING: This extended regular expression feature is consid‐
609 ered highly experimental, and may be changed or deleted with‐
610 out notice. A simplified version of the syntax may be intro‐
611 duced for commonly used idioms.
612 This is a "postponed" regular subexpression. The "code" is
614 evaluated at run time, at the moment this subexpression may
615 match. The result of evaluation is considered as a regular
616 expression and matched as if it were inserted instead of this
617 construct.
618 The "code" is not interpolated. As before, the rules to
620 determine where the "code" ends are currently somewhat convo‐
621 luted.
622 The following pattern matches a parenthesized group:
624 $re = qr{
626 \(
627 (?:
628 (?> [^()]+ ) # Non-parens without backtracking
629 ⎪
630 (??{ $re }) # Group with matching parens
631 )*
632 \)
633 }x;
634 "(?>pattern)"
636 WARNING: This extended regular expression feature is consid‐
637 ered highly experimental, and may be changed or deleted with‐
638 out notice.
639 An "independent" subexpression, one which matches the sub‐
641 string that a standalone "pattern" would match if anchored at
642 the given position, and it matches nothing other than this
643 substring. This construct is useful for optimizations of
644 what would otherwise be "eternal" matches, because it will
645 not backtrack (see "Backtracking"). It may also be useful in
646 places where the "grab all you can, and do not give anything
647 back" semantic is desirable.
648 For example: "^(?>a*)ab" will never match, since "(?>a*)"
650 (anchored at the beginning of string, as above) will match
651 all characters "a" at the beginning of string, leaving no "a"
652 for "ab" to match. In contrast, "a*ab" will match the same
653 as "a+b", since the match of the subgroup "a*" is influenced
654 by the following group "ab" (see "Backtracking"). In partic‐
655 ular, "a*" inside "a*ab" will match fewer characters than a
656 standalone "a*", since this makes the tail match.
657 An effect similar to "(?>pattern)" may be achieved by writing
659 "(?=(pattern))\1". This matches the same substring as a
660 standalone "a+", and the following "\1" eats the matched
661 string; it therefore makes a zero-length assertion into an
662 analogue of "(?>...)". (The difference between these two
663 constructs is that the second one uses a capturing group,
664 thus shifting ordinals of backreferences in the rest of a
665 regular expression.)
666 Consider this pattern:
668 m{ \(
670 (
671 [^()]+ # x+
672 ⎪
673 \( [^()]* \)
674 )+
675 \)
676 }x
677 That will efficiently match a nonempty group with matching
679 parentheses two levels deep or less. However, if there is no
680 such group, it will take virtually forever on a long string.
681 That's because there are so many different ways to split a
682 long string into several substrings. This is what "(.+)+" is
683 doing, and "(.+)+" is similar to a subpattern of the above
684 pattern. Consider how the pattern above detects no-match on
685 "((()aaaaaaaaaaaaaaaaaa" in several seconds, but that each
686 extra letter doubles this time. This exponential performance
687 will make it appear that your program has hung. However, a
688 tiny change to this pattern
689 m{ \(
691 (
692 (?> [^()]+ ) # change x+ above to (?> x+ )
693 ⎪
694 \( [^()]* \)
695 )+
696 \)
697 }x
698 which uses "(?>...)" matches exactly when the one above does
700 (verifying this yourself would be a productive exercise), but
701 finishes in a fourth the time when used on a similar string
702 with 1000000 "a"s. Be aware, however, that this pattern cur‐
703 rently triggers a warning message under the "use warnings"
704 pragma or -w switch saying it "matches null string many times
705 in regex".
706 On simple groups, such as the pattern "(?> [^()]+ )", a com‐
708 parable effect may be achieved by negative look-ahead, as in
709 "[^()]+ (?! [^()] )". This was only 4 times slower on a
710 string with 1000000 "a"s.
711 The "grab all you can, and do not give anything back" seman‐
713 tic is desirable in many situations where on the first sight
714 a simple "()*" looks like the correct solution. Suppose we
715 parse text with comments being delimited by "#" followed by
716 some optional (horizontal) whitespace. Contrary to its
717 appearance, "#[ \t]*" is not the correct subexpression to
718 match the comment delimiter, because it may "give up" some
719 whitespace if the remainder of the pattern can be made to
720 match that way. The correct answer is either one of these:
721 (?>#[ \t]*)
723 #[ \t]*(?![ \t])
724 For example, to grab non-empty comments into $1, one should
726 use either one of these:
727 / (?> \# [ \t]* ) ( .+ ) /x;
729 / \# [ \t]* ( [^ \t] .* ) /x;
730 Which one you pick depends on which of these expressions bet‐
732 ter reflects the above specification of comments.
733 "(?(condition)yes-pattern⎪no-pattern)"
735 "(?(condition)yes-pattern)"
736 WARNING: This extended regular expression feature is consid‐
737 ered highly experimental, and may be changed or deleted with‐
738 out notice.
739 Conditional expression. "(condition)" should be either an
741 integer in parentheses (which is valid if the corresponding
742 pair of parentheses matched), or look-ahead/look-behind/eval‐
743 uate zero-width assertion.
744 For example:
746 m{ ( \( )?
748 [^()]+
749 (?(1) \) )
750 }x
751 matches a chunk of non-parentheses, possibly included in
753 parentheses themselves.
754 Backtracking
756 NOTE: This section presents an abstract approximation of regular
758 expression behavior. For a more rigorous (and complicated) view of the
759 rules involved in selecting a match among possible alternatives, see
760 "Combining pieces together".
761 A fundamental feature of regular expression matching involves the
763 notion called backtracking, which is currently used (when needed) by
764 all regular expression quantifiers, namely "*", "*?", "+", "+?",
765 "{n,m}", and "{n,m}?". Backtracking is often optimized internally, but
766 the general principle outlined here is valid.
767 For a regular expression to match, the entire regular expression must
769 match, not just part of it. So if the beginning of a pattern contain‐
770 ing a quantifier succeeds in a way that causes later parts in the pat‐
771 tern to fail, the matching engine backs up and recalculates the begin‐
772 ning part--that's why it's called backtracking.
773 Here is an example of backtracking: Let's say you want to find the
775 word following "foo" in the string "Food is on the foo table.":
776 $_ = "Food is on the foo table.";
778 if ( /\b(foo)\s+(\w+)/i ) {
779 print "$2 follows $1.\n";
780 }
781 When the match runs, the first part of the regular expression
783 ("\b(foo)") finds a possible match right at the beginning of the
784 string, and loads up $1 with "Foo". However, as soon as the matching
785 engine sees that there's no whitespace following the "Foo" that it had
786 saved in $1, it realizes its mistake and starts over again one charac‐
787 ter after where it had the tentative match. This time it goes all the
788 way until the next occurrence of "foo". The complete regular expression
789 matches this time, and you get the expected output of "table follows
790 foo."
791 Sometimes minimal matching can help a lot. Imagine you'd like to match
793 everything between "foo" and "bar". Initially, you write something
794 like this:
795 $_ = "The food is under the bar in the barn.";
797 if ( /foo(.*)bar/ ) {
798 print "got <$1>\n";
799 }
800 Which perhaps unexpectedly yields:
802 got <d is under the bar in the >
804 That's because ".*" was greedy, so you get everything between the first
806 "foo" and the last "bar". Here it's more effective to use minimal
807 matching to make sure you get the text between a "foo" and the first
808 "bar" thereafter.
809 if ( /foo(.*?)bar/ ) { print "got <$1>\n" }
811 got <d is under the >
812 Here's another example: let's say you'd like to match a number at the
814 end of a string, and you also want to keep the preceding part of the
815 match. So you write this:
816 $_ = "I have 2 numbers: 53147";
818 if ( /(.*)(\d*)/ ) { # Wrong!
819 print "Beginning is <$1>, number is <$2>.\n";
820 }
821 That won't work at all, because ".*" was greedy and gobbled up the
823 whole string. As "\d*" can match on an empty string the complete regu‐
824 lar expression matched successfully.
825 Beginning is <I have 2 numbers: 53147>, number is <>.
827 Here are some variants, most of which don't work:
829 $_ = "I have 2 numbers: 53147";
831 @pats = qw{
832 (.*)(\d*)
833 (.*)(\d+)
834 (.*?)(\d*)
835 (.*?)(\d+)
836 (.*)(\d+)$
837 (.*?)(\d+)$
838 (.*)\b(\d+)$
839 (.*\D)(\d+)$
840 };
841 for $pat (@pats) {
843 printf "%-12s ", $pat;
844 if ( /$pat/ ) {
845 print "<$1> <$2>\n";
846 } else {
847 print "FAIL\n";
848 }
849 }
850 That will print out:
852 (.*)(\d*) <I have 2 numbers: 53147> <>
854 (.*)(\d+) <I have 2 numbers: 5314> <7>
855 (.*?)(\d*) <> <>
856 (.*?)(\d+) <I have > <2>
857 (.*)(\d+)$ <I have 2 numbers: 5314> <7>
858 (.*?)(\d+)$ <I have 2 numbers: > <53147>
859 (.*)\b(\d+)$ <I have 2 numbers: > <53147>
860 (.*\D)(\d+)$ <I have 2 numbers: > <53147>
861 As you see, this can be a bit tricky. It's important to realize that a
863 regular expression is merely a set of assertions that gives a defini‐
864 tion of success. There may be 0, 1, or several different ways that the
865 definition might succeed against a particular string. And if there are
866 multiple ways it might succeed, you need to understand backtracking to
867 know which variety of success you will achieve.
868 When using look-ahead assertions and negations, this can all get even
870 trickier. Imagine you'd like to find a sequence of non-digits not fol‐
871 lowed by "123". You might try to write that as
872 $_ = "ABC123";
874 if ( /^\D*(?!123)/ ) { # Wrong!
875 print "Yup, no 123 in $_\n";
876 }
877 But that isn't going to match; at least, not the way you're hoping. It
879 claims that there is no 123 in the string. Here's a clearer picture of
880 why that pattern matches, contrary to popular expectations:
881 $x = 'ABC123';
883 $y = 'ABC445';
884 print "1: got $1\n" if $x =~ /^(ABC)(?!123)/;
886 print "2: got $1\n" if $y =~ /^(ABC)(?!123)/;
887 print "3: got $1\n" if $x =~ /^(\D*)(?!123)/;
889 print "4: got $1\n" if $y =~ /^(\D*)(?!123)/;
890 This prints
892 2: got ABC
894 3: got AB
895 4: got ABC
896 You might have expected test 3 to fail because it seems to a more gen‐
898 eral purpose version of test 1. The important difference between them
899 is that test 3 contains a quantifier ("\D*") and so can use backtrack‐
900 ing, whereas test 1 will not. What's happening is that you've asked
901 "Is it true that at the start of $x, following 0 or more non-digits,
902 you have something that's not 123?" If the pattern matcher had let
903 "\D*" expand to "ABC", this would have caused the whole pattern to
904 fail.
905 The search engine will initially match "\D*" with "ABC". Then it will
907 try to match "(?!123" with "123", which fails. But because a quanti‐
908 fier ("\D*") has been used in the regular expression, the search engine
909 can backtrack and retry the match differently in the hope of matching
910 the complete regular expression.
911 The pattern really, really wants to succeed, so it uses the standard
913 pattern back-off-and-retry and lets "\D*" expand to just "AB" this
914 time. Now there's indeed something following "AB" that is not "123".
915 It's "C123", which suffices.
916 We can deal with this by using both an assertion and a negation. We'll
918 say that the first part in $1 must be followed both by a digit and by
919 something that's not "123". Remember that the look-aheads are zero-
920 width expressions--they only look, but don't consume any of the string
921 in their match. So rewriting this way produces what you'd expect; that
922 is, case 5 will fail, but case 6 succeeds:
923 print "5: got $1\n" if $x =~ /^(\D*)(?=\d)(?!123)/;
925 print "6: got $1\n" if $y =~ /^(\D*)(?=\d)(?!123)/;
926 6: got ABC
928 In other words, the two zero-width assertions next to each other work
930 as though they're ANDed together, just as you'd use any built-in asser‐
931 tions: "/^$/" matches only if you're at the beginning of the line AND
932 the end of the line simultaneously. The deeper underlying truth is
933 that juxtaposition in regular expressions always means AND, except when
934 you write an explicit OR using the vertical bar. "/ab/" means match
935 "a" AND (then) match "b", although the attempted matches are made at
936 different positions because "a" is not a zero-width assertion, but a
937 one-width assertion.
938 WARNING: particularly complicated regular expressions can take exponen‐
940 tial time to solve because of the immense number of possible ways they
941 can use backtracking to try match. For example, without internal opti‐
942 mizations done by the regular expression engine, this will take a
943 painfully long time to run:
944 'aaaaaaaaaaaa' =~ /((a{0,5}){0,5})*[c]/
946 And if you used "*"'s in the internal groups instead of limiting them
948 to 0 through 5 matches, then it would take forever--or until you ran
949 out of stack space. Moreover, these internal optimizations are not
950 always applicable. For example, if you put "{0,5}" instead of "*" on
951 the external group, no current optimization is applicable, and the
952 match takes a long time to finish.
953 A powerful tool for optimizing such beasts is what is known as an
955 "independent group", which does not backtrack (see ""(?>pattern)"").
956 Note also that zero-length look-ahead/look-behind assertions will not
957 backtrack to make the tail match, since they are in "logical" context:
958 only whether they match is considered relevant. For an example where
959 side-effects of look-ahead might have influenced the following match,
960 see ""(?>pattern)"".
961 Version 8 Regular Expressions
963 In case you're not familiar with the "regular" Version 8 regex rou‐
965 tines, here are the pattern-matching rules not described above.
966 Any single character matches itself, unless it is a metacharacter with
968 a special meaning described here or above. You can cause characters
969 that normally function as metacharacters to be interpreted literally by
970 prefixing them with a "\" (e.g., "\." matches a ".", not any character;
971 "\\" matches a "\"). A series of characters matches that series of
972 characters in the target string, so the pattern "blurfl" would match
973 "blurfl" in the target string.
974 You can specify a character class, by enclosing a list of characters in
976 "[]", which will match any one character from the list. If the first
977 character after the "[" is "^", the class matches any character not in
978 the list. Within a list, the "-" character specifies a range, so that
979 "a-z" represents all characters between "a" and "z", inclusive. If you
980 want either "-" or "]" itself to be a member of a class, put it at the
981 start of the list (possibly after a "^"), or escape it with a back‐
982 slash. "-" is also taken literally when it is at the end of the list,
983 just before the closing "]". (The following all specify the same class
984 of three characters: "[-az]", "[az-]", and "[a\-z]". All are different
985 from "[a-z]", which specifies a class containing twenty-six characters,
986 even on EBCDIC based coded character sets.) Also, if you try to use
987 the character classes "\w", "\W", "\s", "\S", "\d", or "\D" as end‐
988 points of a range, that's not a range, the "-" is understood literally.
989 Note also that the whole range idea is rather unportable between char‐
991 acter sets--and even within character sets they may cause results you
992 probably didn't expect. A sound principle is to use only ranges that
993 begin from and end at either alphabets of equal case ([a-e], [A-E]), or
994 digits ([0-9]). Anything else is unsafe. If in doubt, spell out the
995 character sets in full.
996 Characters may be specified using a metacharacter syntax much like that
998 used in C: "\n" matches a newline, "\t" a tab, "\r" a carriage return,
999 "\f" a form feed, etc. More generally, \nnn, where nnn is a string of
1000 octal digits, matches the character whose coded character set value is
1001 nnn. Similarly, \xnn, where nn are hexadecimal digits, matches the
1002 character whose numeric value is nn. The expression \cx matches the
1003 character control-x. Finally, the "." metacharacter matches any char‐
1004 acter except "\n" (unless you use "/s").
1005 You can specify a series of alternatives for a pattern using "⎪" to
1007 separate them, so that "fee⎪fie⎪foe" will match any of "fee", "fie", or
1008 "foe" in the target string (as would "f(e⎪i⎪o)e"). The first alterna‐
1009 tive includes everything from the last pattern delimiter ("(", "[", or
1010 the beginning of the pattern) up to the first "⎪", and the last alter‐
1011 native contains everything from the last "⎪" to the next pattern delim‐
1012 iter. That's why it's common practice to include alternatives in
1013 parentheses: to minimize confusion about where they start and end.
1014 Alternatives are tried from left to right, so the first alternative
1016 found for which the entire expression matches, is the one that is cho‐
1017 sen. This means that alternatives are not necessarily greedy. For exam‐
1018 ple: when matching "foo⎪foot" against "barefoot", only the "foo" part
1019 will match, as that is the first alternative tried, and it successfully
1020 matches the target string. (This might not seem important, but it is
1021 important when you are capturing matched text using parentheses.)
1022 Also remember that "⎪" is interpreted as a literal within square brack‐
1024 ets, so if you write "[fee⎪fie⎪foe]" you're really only matching
1025 "[feio⎪]".
1026 Within a pattern, you may designate subpatterns for later reference by
1028 enclosing them in parentheses, and you may refer back to the nth sub‐
1029 pattern later in the pattern using the metacharacter \n. Subpatterns
1030 are numbered based on the left to right order of their opening paren‐
1031 thesis. A backreference matches whatever actually matched the subpat‐
1032 tern in the string being examined, not the rules for that subpattern.
1033 Therefore, "(0⎪0x)\d*\s\1\d*" will match "0x1234 0x4321", but not
1034 "0x1234 01234", because subpattern 1 matched "0x", even though the rule
1035 "0⎪0x" could potentially match the leading 0 in the second number.
1036 Warning on \1 vs $1
1038 Some people get too used to writing things like:
1040 $pattern =~ s/(\W)/\\\1/g;
1042 This is grandfathered for the RHS of a substitute to avoid shocking the
1044 sed addicts, but it's a dirty habit to get into. That's because in
1045 PerlThink, the righthand side of an "s///" is a double-quoted string.
1046 "\1" in the usual double-quoted string means a control-A. The custom‐
1047 ary Unix meaning of "\1" is kludged in for "s///". However, if you get
1048 into the habit of doing that, you get yourself into trouble if you then
1049 add an "/e" modifier.
1050 s/(\d+)/ \1 + 1 /eg; # causes warning under -w
1052 Or if you try to do
1054 s/(\d+)/\1000/;
1056 You can't disambiguate that by saying "\{1}000", whereas you can fix it
1058 with "${1}000". The operation of interpolation should not be confused
1059 with the operation of matching a backreference. Certainly they mean
1060 two different things on the left side of the "s///".
1061 Repeated patterns matching zero-length substring
1063 WARNING: Difficult material (and prose) ahead. This section needs a
1065 rewrite.
1066 Regular expressions provide a terse and powerful programming language.
1068 As with most other power tools, power comes together with the ability
1069 to wreak havoc.
1070 A common abuse of this power stems from the ability to make infinite
1072 loops using regular expressions, with something as innocuous as:
1073 'foo' =~ m{ ( o? )* }x;
1075 The "o?" can match at the beginning of 'foo', and since the position in
1077 the string is not moved by the match, "o?" would match again and again
1078 because of the "*" modifier. Another common way to create a similar
1079 cycle is with the looping modifier "//g":
1080 @matches = ( 'foo' =~ m{ o? }xg );
1082 or
1084 print "match: <$&>\n" while 'foo' =~ m{ o? }xg;
1086 or the loop implied by split().
1088 However, long experience has shown that many programming tasks may be
1090 significantly simplified by using repeated subexpressions that may
1091 match zero-length substrings. Here's a simple example being:
1092 @chars = split //, $string; # // is not magic in split
1094 ($whitewashed = $string) =~ s/()/ /g; # parens avoid magic s// /
1095 Thus Perl allows such constructs, by forcefully breaking the infinite
1097 loop. The rules for this are different for lower-level loops given by
1098 the greedy modifiers "*+{}", and for higher-level ones like the "/g"
1099 modifier or split() operator.
1100 The lower-level loops are interrupted (that is, the loop is broken)
1102 when Perl detects that a repeated expression matched a zero-length sub‐
1103 string. Thus
1104 m{ (?: NON_ZERO_LENGTH ⎪ ZERO_LENGTH )* }x;
1106 is made equivalent to
1108 m{ (?: NON_ZERO_LENGTH )*
1110 ⎪
1111 (?: ZERO_LENGTH )?
1112 }x;
1113 The higher level-loops preserve an additional state between iterations:
1115 whether the last match was zero-length. To break the loop, the follow‐
1116 ing match after a zero-length match is prohibited to have a length of
1117 zero. This prohibition interacts with backtracking (see "Backtrack‐
1118 ing"), and so the second best match is chosen if the best match is of
1119 zero length.
1120 For example:
1122 $_ = 'bar';
1124 s/\w??/<$&>/g;
1125 results in "<><b><><a><><r><>". At each position of the string the
1127 best match given by non-greedy "??" is the zero-length match, and the
1128 second best match is what is matched by "\w". Thus zero-length matches
1129 alternate with one-character-long matches.
1130 Similarly, for repeated "m/()/g" the second-best match is the match at
1132 the position one notch further in the string.
1133 The additional state of being matched with zero-length is associated
1135 with the matched string, and is reset by each assignment to pos().
1136 Zero-length matches at the end of the previous match are ignored during
1137 "split".
1138 Combining pieces together
1140 Each of the elementary pieces of regular expressions which were
1142 described before (such as "ab" or "\Z") could match at most one sub‐
1143 string at the given position of the input string. However, in a typi‐
1144 cal regular expression these elementary pieces are combined into more
1145 complicated patterns using combining operators "ST", "S⎪T", "S*" etc
1146 (in these examples "S" and "T" are regular subexpressions).
1147 Such combinations can include alternatives, leading to a problem of
1149 choice: if we match a regular expression "a⎪ab" against "abc", will it
1150 match substring "a" or "ab"? One way to describe which substring is
1151 actually matched is the concept of backtracking (see "Backtracking").
1152 However, this description is too low-level and makes you think in terms
1153 of a particular implementation.
1154 Another description starts with notions of "better"/"worse". All the
1156 substrings which may be matched by the given regular expression can be
1157 sorted from the "best" match to the "worst" match, and it is the "best"
1158 match which is chosen. This substitutes the question of "what is cho‐
1159 sen?" by the question of "which matches are better, and which are
1160 worse?".
1161 Again, for elementary pieces there is no such question, since at most
1163 one match at a given position is possible. This section describes the
1164 notion of better/worse for combining operators. In the description
1165 below "S" and "T" are regular subexpressions.
1166 "ST"
1168 Consider two possible matches, "AB" and "A'B'", "A" and "A'" are
1169 substrings which can be matched by "S", "B" and "B'" are substrings
1170 which can be matched by "T".
1171 If "A" is better match for "S" than "A'", "AB" is a better match
1173 than "A'B'".
1174 If "A" and "A'" coincide: "AB" is a better match than "AB'" if "B"
1176 is better match for "T" than "B'".
1177 "S⎪T"
1179 When "S" can match, it is a better match than when only "T" can
1180 match.
1181 Ordering of two matches for "S" is the same as for "S". Similar
1183 for two matches for "T".
1184 "S{REPEAT_COUNT}"
1186 Matches as "SSS...S" (repeated as many times as necessary).
1187 "S{min,max}"
1189 Matches as "S{max}⎪S{max-1}⎪...⎪S{min+1}⎪S{min}".
1190 "S{min,max}?"
1192 Matches as "S{min}⎪S{min+1}⎪...⎪S{max-1}⎪S{max}".
1193 "S?", "S*", "S+"
1195 Same as "S{0,1}", "S{0,BIG_NUMBER}", "S{1,BIG_NUMBER}" respec‐
1196 tively.
1197 "S??", "S*?", "S+?"
1199 Same as "S{0,1}?", "S{0,BIG_NUMBER}?", "S{1,BIG_NUMBER}?" respec‐
1200 tively.
1201 "(?>S)"
1203 Matches the best match for "S" and only that.
1204 "(?=S)", "(?<=S)"
1206 Only the best match for "S" is considered. (This is important only
1207 if "S" has capturing parentheses, and backreferences are used some‐
1208 where else in the whole regular expression.)
1209 "(?!S)", "(?<!S)"
1211 For this grouping operator there is no need to describe the order‐
1212 ing, since only whether or not "S" can match is important.
1213 "(??{ EXPR })"
1215 The ordering is the same as for the regular expression which is the
1216 result of EXPR.
1217 "(?(condition)yes-pattern⎪no-pattern)"
1219 Recall that which of "yes-pattern" or "no-pattern" actually matches
1220 is already determined. The ordering of the matches is the same as
1221 for the chosen subexpression.
1222 The above recipes describe the ordering of matches at a given position.
1224 One more rule is needed to understand how a match is determined for the
1225 whole regular expression: a match at an earlier position is always bet‐
1226 ter than a match at a later position.
1227 Creating custom RE engines
1229 Overloaded constants (see overload) provide a simple way to extend the
1231 functionality of the RE engine.
1232 Suppose that we want to enable a new RE escape-sequence "\Y⎪" which
1234 matches at boundary between whitespace characters and non-whitespace
1235 characters. Note that "(?=\S)(?<!\S)⎪(?!\S)(?<=\S)" matches exactly at
1236 these positions, so we want to have each "\Y⎪" in the place of the more
1237 complicated version. We can create a module "customre" to do this:
1238 package customre;
1240 use overload;
1241 sub import {
1243 shift;
1244 die "No argument to customre::import allowed" if @_;
1245 overload::constant 'qr' => \&convert;
1246 }
1247 sub invalid { die "/$_[0]/: invalid escape '\\$_[1]'"}
1249 # We must also take care of not escaping the legitimate \\Y⎪
1251 # sequence, hence the presence of '\\' in the conversion rules.
1252 my %rules = ( '\\' => '\\\\',
1253 'Y⎪' => qr/(?=\S)(?<!\S)⎪(?!\S)(?<=\S)/ );
1254 sub convert {
1255 my $re = shift;
1256 $re =~ s{
1257 \\ ( \\ ⎪ Y . )
1258 }
1259 { $rules{$1} or invalid($re,$1) }sgex;
1260 return $re;
1261 }
1262 Now "use customre" enables the new escape in constant regular expres‐
1264 sions, i.e., those without any runtime variable interpolations. As
1265 documented in overload, this conversion will work only over literal
1266 parts of regular expressions. For "\Y⎪$re\Y⎪" the variable part of
1267 this regular expression needs to be converted explicitly (but only if
1268 the special meaning of "\Y⎪" should be enabled inside $re):
1269 use customre;
1271 $re = <>;
1272 chomp $re;
1273 $re = customre::convert $re;
1274 /\Y⎪$re\Y⎪/;
1275
1277 This document varies from difficult to understand to completely and
1278 utterly opaque. The wandering prose riddled with jargon is hard to
1279 fathom in several places.
1280 This document needs a rewrite that separates the tutorial content from
1282 the reference content.
1283
1285 perlrequick.
1286 perlretut.
1288 "Regexp Quote-Like Operators" in perlop.
1290 "Gory details of parsing quoted constructs" in perlop.
1292 perlfaq6.
1294 "pos" in perlfunc.
1296 perllocale.
1298 perlebcdic.
1300 Mastering Regular Expressions by Jeffrey Friedl, published by O'Reilly
1302 and Associates.
1303perl v5.8.8 2006-01-07 PERLRE(1)