3 perlfaq6 - Regexes ($Revision: 1.27 $, $Date: 1999/05/23 16:08:30 $)
7 This section is surprisingly small because the rest of the FAQ is
8 littered with answers involving regular expressions. For example,
9 decoding a URL and checking whether something is a number are handled
10 with regular expressions, but those answers are found elsewhere in
11 this document (in L<perlfaq9>: ``How do I decode or create those %-encodings
12 on the web'' and L<perfaq4>: ``How do I determine whether a scalar is
13 a number/whole/integer/float'', to be precise).
15 =head2 How can I hope to use regular expressions without creating illegible and unmaintainable code?
17 Three techniques can make regular expressions maintainable and
22 =item Comments Outside the Regex
24 Describe what you're doing and how you're doing it, using normal Perl
27 # turn the line into the first word, a colon, and the
28 # number of characters on the rest of the line
29 s/^(\w+)(.*)/ lc($1) . ":" . length($2) /meg;
31 =item Comments Inside the Regex
33 The C</x> modifier causes whitespace to be ignored in a regex pattern
34 (except in a character class), and also allows you to use normal
35 comments there, too. As you can imagine, whitespace and comments help
38 C</x> lets you turn this:
40 s{<(?:[^>'"]*|".*?"|'.*?')+>}{}gs;
44 s{ < # opening angle bracket
45 (?: # Non-backreffing grouping paren
46 [^>'"] * # 0 or more things that are neither > nor ' nor "
48 ".*?" # a section between double quotes (stingy match)
50 '.*?' # a section between single quotes (stingy match)
51 ) + # all occurring one or more times
52 > # closing angle bracket
53 }{}gsx; # replace with nothing, i.e. delete
55 It's still not quite so clear as prose, but it is very useful for
56 describing the meaning of each part of the pattern.
58 =item Different Delimiters
60 While we normally think of patterns as being delimited with C</>
61 characters, they can be delimited by almost any character. L<perlre>
62 describes this. For example, the C<s///> above uses braces as
63 delimiters. Selecting another delimiter can avoid quoting the
64 delimiter within the pattern:
66 s/\/usr\/local/\/usr\/share/g; # bad delimiter choice
67 s#/usr/local#/usr/share#g; # better
71 =head2 I'm having trouble matching over more than one line. What's wrong?
73 Either you don't have more than one line in the string you're looking at
74 (probably), or else you aren't using the correct modifier(s) on your
77 There are many ways to get multiline data into a string. If you want
78 it to happen automatically while reading input, you'll want to set $/
79 (probably to '' for paragraphs or C<undef> for the whole file) to
80 allow you to read more than one line at a time.
82 Read L<perlre> to help you decide which of C</s> and C</m> (or both)
83 you might want to use: C</s> allows dot to include newline, and C</m>
84 allows caret and dollar to match next to a newline, not just at the
85 end of the string. You do need to make sure that you've actually
86 got a multiline string in there.
88 For example, this program detects duplicate words, even when they span
89 line breaks (but not paragraph ones). For this example, we don't need
90 C</s> because we aren't using dot in a regular expression that we want
91 to cross line boundaries. Neither do we need C</m> because we aren't
92 wanting caret or dollar to match at any point inside the record next
93 to newlines. But it's imperative that $/ be set to something other
94 than the default, or else we won't actually ever have a multiline
97 $/ = ''; # read in more whole paragraph, not just one line
99 while ( /\b([\w'-]+)(\s+\1)+\b/gi ) { # word starts alpha
100 print "Duplicate $1 at paragraph $.\n";
104 Here's code that finds sentences that begin with "From " (which would
105 be mangled by many mailers):
107 $/ = ''; # read in more whole paragraph, not just one line
109 while ( /^From /gm ) { # /m makes ^ match next to \n
110 print "leading from in paragraph $.\n";
114 Here's code that finds everything between START and END in a paragraph:
116 undef $/; # read in whole file, not just one line or paragraph
118 while ( /START(.*?)END/sgm ) { # /s makes . cross line boundaries
123 =head2 How can I pull out lines between two patterns that are themselves on different lines?
125 You can use Perl's somewhat exotic C<..> operator (documented in
128 perl -ne 'print if /START/ .. /END/' file1 file2 ...
130 If you wanted text and not lines, you would use
132 perl -0777 -ne 'print "$1\n" while /START(.*?)END/gs' file1 file2 ...
134 But if you want nested occurrences of C<START> through C<END>, you'll
135 run up against the problem described in the question in this section
136 on matching balanced text.
138 Here's another example of using C<..>:
141 $in_header = 1 .. /^$/;
142 $in_body = /^$/ .. eof();
143 # now choose between them
145 reset if eof(); # fix $.
148 =head2 I put a regular expression into $/ but it didn't work. What's wrong?
150 $/ must be a string, not a regular expression. Awk has to be better
153 Actually, you could do this if you don't mind reading the whole file
157 @records = split /your_pattern/, <FH>;
159 The Net::Telnet module (available from CPAN) has the capability to
160 wait for a pattern in the input stream, or timeout if it doesn't
161 appear within a certain time.
163 ## Create a file with three lines.
165 print FH "The first line\nThe second line\nThe third line\n";
168 ## Get a read/write filehandle to it.
169 $fh = new FileHandle "+<file";
171 ## Attach it to a "stream" object.
173 $file = new Net::Telnet (-fhopen => $fh);
175 ## Search for the second line and print out the third.
176 $file->waitfor('/second line\n/');
177 print $file->getline;
179 =head2 How do I substitute case insensitively on the LHS while preserving case on the RHS?
181 Here's a lovely Perlish solution by Larry Rosler. It exploits
182 properties of bitwise xor on ASCII strings.
184 $_= "this is a TEsT case";
190 { uc $new | (uc $1 ^ $1) .
191 (uc(substr $1, -1) ^ substr $1, -1) x
192 (length($new) - length $1)
197 And here it is as a subroutine, modelled after the above:
199 sub preserve_case($$) {
200 my ($old, $new) = @_;
201 my $mask = uc $old ^ $old;
204 substr($mask, -1) x (length($new) - length($old))
207 $a = "this is a TEsT case";
208 $a =~ s/(test)/preserve_case($1, "success")/egi;
213 this is a SUcCESS case
215 As an alternative, to keep the case of the replacement word if it is
216 longer than the original, you can use this code, by Jeff Pinyan:
219 my ($from, $to) = @_;
220 my ($lf, $lt) = map length, @_;
222 if ($lt < $lf) { $from = substr $from, 0, $lt }
223 else { $from .= substr $to, $lf }
225 return uc $to | ($from ^ uc $from);
228 This changes the sentence to "this is a SUcCess case."
230 Just to show that C programmers can write C in any programming language,
231 if you prefer a more C-like solution, the following script makes the
232 substitution have the same case, letter by letter, as the original.
233 (It also happens to run about 240% slower than the Perlish solution runs.)
234 If the substitution has more characters than the string being substituted,
235 the case of the last character is used for the rest of the substitution.
237 # Original by Nathan Torkington, massaged by Jeffrey Friedl
239 sub preserve_case($$)
241 my ($old, $new) = @_;
242 my ($state) = 0; # 0 = no change; 1 = lc; 2 = uc
243 my ($i, $oldlen, $newlen, $c) = (0, length($old), length($new));
244 my ($len) = $oldlen < $newlen ? $oldlen : $newlen;
246 for ($i = 0; $i < $len; $i++) {
247 if ($c = substr($old, $i, 1), $c =~ /[\W\d_]/) {
249 } elsif (lc $c eq $c) {
250 substr($new, $i, 1) = lc(substr($new, $i, 1));
253 substr($new, $i, 1) = uc(substr($new, $i, 1));
257 # finish up with any remaining new (for when new is longer than old)
258 if ($newlen > $oldlen) {
260 substr($new, $oldlen) = lc(substr($new, $oldlen));
261 } elsif ($state == 2) {
262 substr($new, $oldlen) = uc(substr($new, $oldlen));
268 =head2 How can I make C<\w> match national character sets?
272 =head2 How can I match a locale-smart version of C</[a-zA-Z]/>?
274 One alphabetic character would be C</[^\W\d_]/>, no matter what locale
275 you're in. Non-alphabetics would be C</[\W\d_]/> (assuming you don't
276 consider an underscore a letter).
278 =head2 How can I quote a variable to use in a regex?
280 The Perl parser will expand $variable and @variable references in
281 regular expressions unless the delimiter is a single quote. Remember,
282 too, that the right-hand side of a C<s///> substitution is considered
283 a double-quoted string (see L<perlop> for more details). Remember
284 also that any regex special characters will be acted on unless you
285 precede the substitution with \Q. Here's an example:
289 $rhs = "sleep, no more";
291 $string =~ s/\Q$lhs/$rhs/;
292 # $string is now "to sleep no more"
294 Without the \Q, the regex would also spuriously match "di".
296 =head2 What is C</o> really for?
298 Using a variable in a regular expression match forces a re-evaluation
299 (and perhaps recompilation) each time the regular expression is
300 encountered. The C</o> modifier locks in the regex the first time
301 it's used. This always happens in a constant regular expression, and
302 in fact, the pattern was compiled into the internal format at the same
303 time your entire program was.
305 Use of C</o> is irrelevant unless variable interpolation is used in
306 the pattern, and if so, the regex engine will neither know nor care
307 whether the variables change after the pattern is evaluated the I<very
310 C</o> is often used to gain an extra measure of efficiency by not
311 performing subsequent evaluations when you know it won't matter
312 (because you know the variables won't change), or more rarely, when
313 you don't want the regex to notice if they do.
315 For example, here's a "paragrep" program:
317 $/ = ''; # paragraph mode
323 =head2 How do I use a regular expression to strip C style comments from a file?
325 While this actually can be done, it's much harder than you'd think.
326 For example, this one-liner
328 perl -0777 -pe 's{/\*.*?\*/}{}gs' foo.c
330 will work in many but not all cases. You see, it's too simple-minded for
331 certain kinds of C programs, in particular, those with what appear to be
332 comments in quoted strings. For that, you'd need something like this,
333 created by Jeffrey Friedl and later modified by Fred Curtis.
337 s#/\*[^*]*\*+([^/*][^*]*\*+)*/|("(\\.|[^"\\])*"|'(\\.|[^'\\])*'|.[^/"'\\]*)#$2#gs
340 This could, of course, be more legibly written with the C</x> modifier, adding
341 whitespace and comments. Here it is expanded, courtesy of Fred Curtis.
344 /\* ## Start of /* ... */ comment
345 [^*]*\*+ ## Non-* followed by 1-or-more *'s
348 )* ## 0-or-more things which don't start with /
349 ## but do end with '*'
350 / ## End of /* ... */ comment
352 | ## OR various things which aren't comments:
355 " ## Start of " ... " string
361 " ## End of " ... " string
365 ' ## Start of ' ... ' string
371 ' ## End of ' ... ' string
375 . ## Anything other char
376 [^/"'\\]* ## Chars which doesn't start a comment, string or escape
380 A slight modification also removes C++ comments:
382 s#/\*[^*]*\*+([^/*][^*]*\*+)*/|//[^\n]*|("(\\.|[^"\\])*"|'(\\.|[^'\\])*'|.[^/"'\\]*)#$2#gs;
384 =head2 Can I use Perl regular expressions to match balanced text?
386 Although Perl regular expressions are more powerful than "mathematical"
387 regular expressions because they feature conveniences like backreferences
388 (C<\1> and its ilk), they still aren't powerful enough--with
389 the possible exception of bizarre and experimental features in the
390 development-track releases of Perl. You still need to use non-regex
391 techniques to parse balanced text, such as the text enclosed between
392 matching parentheses or braces, for example.
394 An elaborate subroutine (for 7-bit ASCII only) to pull out balanced
395 and possibly nested single chars, like C<`> and C<'>, C<{> and C<}>,
396 or C<(> and C<)> can be found in
397 http://www.perl.com/CPAN/authors/id/TOMC/scripts/pull_quotes.gz .
399 The C::Scan module from CPAN contains such subs for internal use,
400 but they are undocumented.
402 =head2 What does it mean that regexes are greedy? How can I get around it?
404 Most people mean that greedy regexes match as much as they can.
405 Technically speaking, it's actually the quantifiers (C<?>, C<*>, C<+>,
406 C<{}>) that are greedy rather than the whole pattern; Perl prefers local
407 greed and immediate gratification to overall greed. To get non-greedy
408 versions of the same quantifiers, use (C<??>, C<*?>, C<+?>, C<{}?>).
412 $s1 = $s2 = "I am very very cold";
413 $s1 =~ s/ve.*y //; # I am cold
414 $s2 =~ s/ve.*?y //; # I am very cold
416 Notice how the second substitution stopped matching as soon as it
417 encountered "y ". The C<*?> quantifier effectively tells the regular
418 expression engine to find a match as quickly as possible and pass
419 control on to whatever is next in line, like you would if you were
422 =head2 How do I process each word on each line?
424 Use the split function:
427 foreach $word ( split ) {
428 # do something with $word here
432 Note that this isn't really a word in the English sense; it's just
433 chunks of consecutive non-whitespace characters.
435 To work with only alphanumeric sequences (including underscores), you
439 foreach $word (m/(\w+)/g) {
440 # do something with $word here
444 =head2 How can I print out a word-frequency or line-frequency summary?
446 To do this, you have to parse out each word in the input stream. We'll
447 pretend that by word you mean chunk of alphabetics, hyphens, or
448 apostrophes, rather than the non-whitespace chunk idea of a word given
449 in the previous question:
452 while ( /(\b[^\W_\d][\w'-]+\b)/g ) { # misses "`sheep'"
456 while ( ($word, $count) = each %seen ) {
457 print "$count $word\n";
460 If you wanted to do the same thing for lines, you wouldn't need a
466 while ( ($line, $count) = each %seen ) {
467 print "$count $line";
470 If you want these output in a sorted order, see L<perlfaq4>: ``How do I
471 sort a hash (optionally by value instead of key)?''.
473 =head2 How can I do approximate matching?
475 See the module String::Approx available from CPAN.
477 =head2 How do I efficiently match many regular expressions at once?
479 The following is extremely inefficient:
481 # slow but obvious way
482 @popstates = qw(CO ON MI WI MN);
483 while (defined($line = <>)) {
484 for $state (@popstates) {
485 if ($line =~ /\b$state\b/i) {
492 That's because Perl has to recompile all those patterns for each of
493 the lines of the file. As of the 5.005 release, there's a much better
494 approach, one which makes use of the new C<qr//> operator:
496 # use spiffy new qr// operator, with /i flag even
498 @popstates = qw(CO ON MI WI MN);
499 @poppats = map { qr/\b$_\b/i } @popstates;
500 while (defined($line = <>)) {
501 for $patobj (@poppats) {
502 print $line if $line =~ /$patobj/;
506 =head2 Why don't word-boundary searches with C<\b> work for me?
508 Two common misconceptions are that C<\b> is a synonym for C<\s+> and
509 that it's the edge between whitespace characters and non-whitespace
510 characters. Neither is correct. C<\b> is the place between a C<\w>
511 character and a C<\W> character (that is, C<\b> is the edge of a
512 "word"). It's a zero-width assertion, just like C<^>, C<$>, and all
513 the other anchors, so it doesn't consume any characters. L<perlre>
514 describes the behavior of all the regex metacharacters.
516 Here are examples of the incorrect application of C<\b>, with fixes:
518 "two words" =~ /(\w+)\b(\w+)/; # WRONG
519 "two words" =~ /(\w+)\s+(\w+)/; # right
521 " =matchless= text" =~ /\b=(\w+)=\b/; # WRONG
522 " =matchless= text" =~ /=(\w+)=/; # right
524 Although they may not do what you thought they did, C<\b> and C<\B>
525 can still be quite useful. For an example of the correct use of
526 C<\b>, see the example of matching duplicate words over multiple
529 An example of using C<\B> is the pattern C<\Bis\B>. This will find
530 occurrences of "is" on the insides of words only, as in "thistle", but
531 not "this" or "island".
533 =head2 Why does using $&, $`, or $' slow my program down?
535 Once Perl sees that you need one of these variables anywhere in
536 the program, it provides them on each and every pattern match.
537 The same mechanism that handles these provides for the use of $1, $2,
538 etc., so you pay the same price for each regex that contains capturing
539 parentheses. If you never use $&, etc., in your script, then regexes
540 I<without> capturing parentheses won't be penalized. So avoid $&, $',
541 and $` if you can, but if you can't, once you've used them at all, use
542 them at will because you've already paid the price. Remember that some
543 algorithms really appreciate them. As of the 5.005 release. the $&
544 variable is no longer "expensive" the way the other two are.
546 =head2 What good is C<\G> in a regular expression?
548 The notation C<\G> is used in a match or substitution in conjunction with
549 the C</g> modifier to anchor the regular expression to the point just past
550 where the last match occurred, i.e. the pos() point. A failed match resets
551 the position of C<\G> unless the C</c> modifier is in effect. C<\G> can be
552 used in a match without the C</g> modifier; it acts the same (i.e. still
553 anchors at the pos() point) but of course only matches once and does not
554 update pos(), as non-C</g> expressions never do. C<\G> in an expression
555 applied to a target string that has never been matched against a C</g>
556 expression before or has had its pos() reset is functionally equivalent to
557 C<\A>, which matches at the beginning of the string.
559 For example, suppose you had a line of text quoted in standard mail
560 and Usenet notation, (that is, with leading C<< > >> characters), and
561 you want change each leading C<< > >> into a corresponding C<:>. You
562 could do so in this way:
564 s/^(>+)/':' x length($1)/gem;
566 Or, using C<\G>, the much simpler (and faster):
570 A more sophisticated use might involve a tokenizer. The following
571 lex-like example is courtesy of Jeffrey Friedl. It did not work in
572 5.003 due to bugs in that release, but does work in 5.004 or better.
573 (Note the use of C</c>, which prevents a failed match with C</g> from
574 resetting the search position back to the beginning of the string.)
579 m/ \G( \d+\b )/gcx && do { print "number: $1\n"; redo; };
580 m/ \G( \w+ )/gcx && do { print "word: $1\n"; redo; };
581 m/ \G( \s+ )/gcx && do { print "space: $1\n"; redo; };
582 m/ \G( [^\w\d]+ )/gcx && do { print "other: $1\n"; redo; };
586 Of course, that could have been written as
591 if ( /\G( \d+\b )/gcx {
592 print "number: $1\n";
595 if ( /\G( \w+ )/gcx {
599 if ( /\G( \s+ )/gcx {
603 if ( /\G( [^\w\d]+ )/gcx {
610 but then you lose the vertical alignment of the regular expressions.
612 =head2 Are Perl regexes DFAs or NFAs? Are they POSIX compliant?
614 While it's true that Perl's regular expressions resemble the DFAs
615 (deterministic finite automata) of the egrep(1) program, they are in
616 fact implemented as NFAs (non-deterministic finite automata) to allow
617 backtracking and backreferencing. And they aren't POSIX-style either,
618 because those guarantee worst-case behavior for all cases. (It seems
619 that some people prefer guarantees of consistency, even when what's
620 guaranteed is slowness.) See the book "Mastering Regular Expressions"
621 (from O'Reilly) by Jeffrey Friedl for all the details you could ever
622 hope to know on these matters (a full citation appears in
625 =head2 What's wrong with using grep or map in a void context?
627 Both grep and map build a return list, regardless of their context.
628 This means you're making Perl go to the trouble of building up a
629 return list that you then just ignore. That's no way to treat a
630 programming language, you insensitive scoundrel!
632 =head2 How can I match strings with multibyte characters?
634 This is hard, and there's no good way. Perl does not directly support
635 wide characters. It pretends that a byte and a character are
636 synonymous. The following set of approaches was offered by Jeffrey
637 Friedl, whose article in issue #5 of The Perl Journal talks about this
640 Let's suppose you have some weird Martian encoding where pairs of
641 ASCII uppercase letters encode single Martian letters (i.e. the two
642 bytes "CV" make a single Martian letter, as do the two bytes "SG",
643 "VS", "XX", etc.). Other bytes represent single characters, just like
646 So, the string of Martian "I am CVSGXX!" uses 12 bytes to encode the
647 nine characters 'I', ' ', 'a', 'm', ' ', 'CV', 'SG', 'XX', '!'.
649 Now, say you want to search for the single character C</GX/>. Perl
650 doesn't know about Martian, so it'll find the two bytes "GX" in the "I
651 am CVSGXX!" string, even though that character isn't there: it just
652 looks like it is because "SG" is next to "XX", but there's no real
653 "GX". This is a big problem.
655 Here are a few ways, all painful, to deal with it:
657 $martian =~ s/([A-Z][A-Z])/ $1 /g; # Make sure adjacent ``martian'' bytes
658 # are no longer adjacent.
659 print "found GX!\n" if $martian =~ /GX/;
663 @chars = $martian =~ m/([A-Z][A-Z]|[^A-Z])/g;
664 # above is conceptually similar to: @chars = $text =~ m/(.)/g;
666 foreach $char (@chars) {
667 print "found GX!\n", last if $char eq 'GX';
672 while ($martian =~ m/\G([A-Z][A-Z]|.)/gs) { # \G probably unneeded
673 print "found GX!\n", last if $1 eq 'GX';
678 die "sorry, Perl doesn't (yet) have Martian support )-:\n";
680 There are many double- (and multi-) byte encodings commonly used these
681 days. Some versions of these have 1-, 2-, 3-, and 4-byte characters,
684 =head2 How do I match a pattern that is supplied by the user?
686 Well, if it's really a pattern, then just use
688 chomp($pattern = <STDIN>);
689 if ($line =~ /$pattern/) { }
691 Alternatively, since you have no guarantee that your user entered
692 a valid regular expression, trap the exception this way:
694 if (eval { $line =~ /$pattern/ }) { }
696 If all you really want to search for a string, not a pattern,
697 then you should either use the index() function, which is made for
698 string searching, or if you can't be disabused of using a pattern
699 match on a non-pattern, then be sure to use C<\Q>...C<\E>, documented
704 open (FILE, $input) or die "Couldn't open input $input: $!; aborting";
706 print if /\Q$pattern\E/;
710 =head1 AUTHOR AND COPYRIGHT
712 Copyright (c) 1997-1999 Tom Christiansen and Nathan Torkington.
715 This documentation is free; you can redistribute it and/or modify it
716 under the same terms as Perl itself.
718 Irrespective of its distribution, all code examples in this file
719 are hereby placed into the public domain. You are permitted and
720 encouraged to use this code in your own programs for fun
721 or for profit as you see fit. A simple comment in the code giving
722 credit would be courteous but is not required.