3 perlfaq6 - Regexps ($Revision: 1.14 $)
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 the section on Data and the Networking one on
12 networking, to be precise).
14 =head2 How can I hope to use regular expressions without creating illegible and unmaintainable code?
16 Three techniques can make regular expressions maintainable and
21 =item Comments Outside the Regexp
23 Describe what you're doing and how you're doing it, using normal Perl
26 # turn the line into the first word, a colon, and the
27 # number of characters on the rest of the line
28 s/^(\w+)(.*)/ lc($1) . ":" . length($2) /ge;
30 =item Comments Inside the Regexp
32 The C</x> modifier causes whitespace to be ignored in a regexp pattern
33 (except in a character class), and also allows you to use normal
34 comments there, too. As you can imagine, whitespace and comments help
37 C</x> lets you turn this:
39 s{<(?:[^>'"]*|".*?"|'.*?')+>}{}gs;
43 s{ < # opening angle bracket
44 (?: # Non-backreffing grouping paren
45 [^>'"] * # 0 or more things that are neither > nor ' nor "
47 ".*?" # a section between double quotes (stingy match)
49 '.*?' # a section between single quotes (stingy match)
50 ) + # all occurring one or more times
51 > # closing angle bracket
52 }{}gsx; # replace with nothing, i.e. delete
54 It's still not quite so clear as prose, but it is very useful for
55 describing the meaning of each part of the pattern.
57 =item Different Delimiters
59 While we normally think of patterns as being delimited with C</>
60 characters, they can be delimited by almost any character. L<perlre>
61 describes this. For example, the C<s///> above uses braces as
62 delimiters. Selecting another delimiter can avoid quoting the
63 delimiter within the pattern:
65 s/\/usr\/local/\/usr\/share/g; # bad delimiter choice
66 s#/usr/local#/usr/share#g; # better
70 =head2 I'm having trouble matching over more than one line. What's wrong?
72 Either you don't have newlines in your string, or you aren't using the
73 correct modifier(s) on your pattern.
75 There are many ways to get multiline data into a string. If you want
76 it to happen automatically while reading input, you'll want to set $/
77 (probably to '' for paragraphs or C<undef> for the whole file) to
78 allow you to read more than one line at a time.
80 Read L<perlre> to help you decide which of C</s> and C</m> (or both)
81 you might want to use: C</s> allows dot to include newline, and C</m>
82 allows caret and dollar to match next to a newline, not just at the
83 end of the string. You do need to make sure that you've actually
84 got a multiline string in there.
86 For example, this program detects duplicate words, even when they span
87 line breaks (but not paragraph ones). For this example, we don't need
88 C</s> because we aren't using dot in a regular expression that we want
89 to cross line boundaries. Neither do we need C</m> because we aren't
90 wanting caret or dollar to match at any point inside the record next
91 to newlines. But it's imperative that $/ be set to something other
92 than the default, or else we won't actually ever have a multiline
95 $/ = ''; # read in more whole paragraph, not just one line
97 while ( /\b(\w\S+)(\s+\1)+\b/gi ) {
98 print "Duplicate $1 at paragraph $.\n";
102 Here's code that finds sentences that begin with "From " (which would
103 be mangled by many mailers):
105 $/ = ''; # read in more whole paragraph, not just one line
107 while ( /^From /gm ) { # /m makes ^ match next to \n
108 print "leading from in paragraph $.\n";
112 Here's code that finds everything between START and END in a paragraph:
114 undef $/; # read in whole file, not just one line or paragraph
116 while ( /START(.*?)END/sm ) { # /s makes . cross line boundaries
121 =head2 How can I pull out lines between two patterns that are themselves on different lines?
123 You can use Perl's somewhat exotic C<..> operator (documented in
126 perl -ne 'print if /START/ .. /END/' file1 file2 ...
128 If you wanted text and not lines, you would use
130 perl -0777 -pe 'print "$1\n" while /START(.*?)END/gs' file1 file2 ...
132 But if you want nested occurrences of C<START> through C<END>, you'll
133 run up against the problem described in the question in this section
134 on matching balanced text.
136 =head2 I put a regular expression into $/ but it didn't work. What's wrong?
138 $/ must be a string, not a regular expression. Awk has to be better
141 Actually, you could do this if you don't mind reading the whole file into
144 @records = split /your_pattern/, <FH>;
146 =head2 How do I substitute case insensitively on the LHS, but preserving case on the RHS?
148 It depends on what you mean by "preserving case". The following
149 script makes the substitution have the same case, letter by letter, as
150 the original. If the substitution has more characters than the string
151 being substituted, the case of the last character is used for the rest
154 # Original by Nathan Torkington, massaged by Jeffrey Friedl
156 sub preserve_case($$)
158 my ($old, $new) = @_;
159 my ($state) = 0; # 0 = no change; 1 = lc; 2 = uc
160 my ($i, $oldlen, $newlen, $c) = (0, length($old), length($new));
161 my ($len) = $oldlen < $newlen ? $oldlen : $newlen;
163 for ($i = 0; $i < $len; $i++) {
164 if ($c = substr($old, $i, 1), $c =~ /[\W\d_]/) {
166 } elsif (lc $c eq $c) {
167 substr($new, $i, 1) = lc(substr($new, $i, 1));
170 substr($new, $i, 1) = uc(substr($new, $i, 1));
174 # finish up with any remaining new (for when new is longer than old)
175 if ($newlen > $oldlen) {
177 substr($new, $oldlen) = lc(substr($new, $oldlen));
178 } elsif ($state == 2) {
179 substr($new, $oldlen) = uc(substr($new, $oldlen));
185 $a = "this is a TEsT case";
186 $a =~ s/(test)/preserve_case($1, "success")/gie;
191 this is a SUcCESS case
193 =head2 How can I make C<\w> match accented characters?
197 =head2 How can I match a locale-smart version of C</[a-zA-Z]/>?
199 One alphabetic character would be C</[^\W\d_]/>, no matter what locale
200 you're in. Non-alphabetics would be C</[\W\d_]/> (assuming you don't
201 consider an underscore a letter).
203 =head2 How can I quote a variable to use in a regexp?
205 The Perl parser will expand $variable and @variable references in
206 regular expressions unless the delimiter is a single quote. Remember,
207 too, that the right-hand side of a C<s///> substitution is considered
208 a double-quoted string (see L<perlop> for more details). Remember
209 also that any regexp special characters will be acted on unless you
210 precede the substitution with \Q. Here's an example:
214 $rhs = "sleep no more";
216 $string =~ s/\Q$lhs/$rhs/;
217 # $string is now "to sleep no more"
219 Without the \Q, the regexp would also spuriously match "di".
221 =head2 What is C</o> really for?
223 Using a variable in a regular expression match forces a re-evaluation
224 (and perhaps recompilation) each time through. The C</o> modifier
225 locks in the regexp the first time it's used. This always happens in a
226 constant regular expression, and in fact, the pattern was compiled
227 into the internal format at the same time your entire program was.
229 Use of C</o> is irrelevant unless variable interpolation is used in
230 the pattern, and if so, the regexp engine will neither know nor care
231 whether the variables change after the pattern is evaluated the I<very
234 C</o> is often used to gain an extra measure of efficiency by not
235 performing subsequent evaluations when you know it won't matter
236 (because you know the variables won't change), or more rarely, when
237 you don't want the regexp to notice if they do.
239 For example, here's a "paragrep" program:
241 $/ = ''; # paragraph mode
247 =head2 How do I use a regular expression to strip C style comments from a file?
249 While this actually can be done, it's much harder than you'd think.
250 For example, this one-liner
252 perl -0777 -pe 's{/\*.*?\*/}{}gs' foo.c
254 will work in many but not all cases. You see, it's too simple-minded for
255 certain kinds of C programs, in particular, those with what appear to be
256 comments in quoted strings. For that, you'd need something like this,
257 created by Jeffrey Friedl:
261 s#/\*[^*]*\*+([^/*][^*]*\*+)*/|("(\\.|[^"\\])*"|'(\\.|[^'\\])*'|\n+|.[^/"'\\]*)#$2#g;
264 This could, of course, be more legibly written with the C</x> modifier, adding
265 whitespace and comments.
267 =head2 Can I use Perl regular expressions to match balanced text?
269 Although Perl regular expressions are more powerful than "mathematical"
270 regular expressions, because they feature conveniences like backreferences
271 (C<\1> and its ilk), they still aren't powerful enough. You still need
272 to use non-regexp techniques to parse balanced text, such as the text
273 enclosed between matching parentheses or braces, for example.
275 An elaborate subroutine (for 7-bit ASCII only) to pull out balanced
276 and possibly nested single chars, like C<`> and C<'>, C<{> and C<}>,
277 or C<(> and C<)> can be found in
278 http://www.perl.com/CPAN/authors/id/TOMC/scripts/pull_quotes.gz .
280 The C::Scan module from CPAN contains such subs for internal usage,
281 but they are undocumented.
283 =head2 What does it mean that regexps are greedy? How can I get around it?
285 Most people mean that greedy regexps match as much as they can.
286 Technically speaking, it's actually the quantifiers (C<?>, C<*>, C<+>,
287 C<{}>) that are greedy rather than the whole pattern; Perl prefers local
288 greed and immediate gratification to overall greed. To get non-greedy
289 versions of the same quantifiers, use (C<??>, C<*?>, C<+?>, C<{}?>).
293 $s1 = $s2 = "I am very very cold";
294 $s1 =~ s/ve.*y //; # I am cold
295 $s2 =~ s/ve.*?y //; # I am very cold
297 Notice how the second substitution stopped matching as soon as it
298 encountered "y ". The C<*?> quantifier effectively tells the regular
299 expression engine to find a match as quickly as possible and pass
300 control on to whatever is next in line, like you would if you were
303 =head2 How do I process each word on each line?
305 Use the split function:
308 foreach $word ( split ) {
309 # do something with $word here
313 Note that this isn't really a word in the English sense; it's just
314 chunks of consecutive non-whitespace characters.
316 To work with only alphanumeric sequences, you might consider
319 foreach $word (m/(\w+)/g) {
320 # do something with $word here
324 =head2 How can I print out a word-frequency or line-frequency summary?
326 To do this, you have to parse out each word in the input stream. We'll
327 pretend that by word you mean chunk of alphabetics, hyphens, or
328 apostrophes, rather than the non-whitespace chunk idea of a word given
329 in the previous question:
332 while ( /(\b[^\W_\d][\w'-]+\b)/g ) { # misses "`sheep'"
336 while ( ($word, $count) = each %seen ) {
337 print "$count $word\n";
340 If you wanted to do the same thing for lines, you wouldn't need a
346 while ( ($line, $count) = each %seen ) {
347 print "$count $line";
350 If you want these output in a sorted order, see the section on Hashes.
352 =head2 How can I do approximate matching?
354 See the module String::Approx available from CPAN.
356 =head2 How do I efficiently match many regular expressions at once?
358 The following is super-inefficient:
361 foreach $pat (@patterns) {
368 Instead, you either need to use one of the experimental Regexp extension
369 modules from CPAN (which might well be overkill for your purposes),
370 or else put together something like this, inspired from a routine
371 in Jeffrey Friedl's book:
374 my $condition = shift;
375 my @regexp = @_; # this MUST not be local(); need my()
376 my $expr = join $condition => map { "m/\$regexp[$_]/o" } (0..$#regexp);
377 my $match_func = eval "sub { $expr }";
378 die if $@; # propagate $@; this shouldn't happen!
382 sub bm_and { _bm_build('&&', @_) }
383 sub bm_or { _bm_build('||', @_) }
393 (?i)sys(tem)?\s*[V5]\b
396 # feed me /etc/termcap, prolly
398 print "1: $_" if &$f1;
399 print "2: $_" if &$f2;
402 =head2 Why don't word-boundary searches with C<\b> work for me?
404 Two common misconceptions are that C<\b> is a synonym for C<\s+>, and
405 that it's the edge between whitespace characters and non-whitespace
406 characters. Neither is correct. C<\b> is the place between a C<\w>
407 character and a C<\W> character (that is, C<\b> is the edge of a
408 "word"). It's a zero-width assertion, just like C<^>, C<$>, and all
409 the other anchors, so it doesn't consume any characters. L<perlre>
410 describes the behaviour of all the regexp metacharacters.
412 Here are examples of the incorrect application of C<\b>, with fixes:
414 "two words" =~ /(\w+)\b(\w+)/; # WRONG
415 "two words" =~ /(\w+)\s+(\w+)/; # right
417 " =matchless= text" =~ /\b=(\w+)=\b/; # WRONG
418 " =matchless= text" =~ /=(\w+)=/; # right
420 Although they may not do what you thought they did, C<\b> and C<\B>
421 can still be quite useful. For an example of the correct use of
422 C<\b>, see the example of matching duplicate words over multiple
425 An example of using C<\B> is the pattern C<\Bis\B>. This will find
426 occurrences of "is" on the insides of words only, as in "thistle", but
427 not "this" or "island".
429 =head2 Why does using $&, $`, or $' slow my program down?
431 Because once Perl sees that you need one of these variables anywhere
432 in the program, it has to provide them on each and every pattern
433 match. The same mechanism that handles these provides for the use of
434 $1, $2, etc., so you pay the same price for each regexp that contains
435 capturing parentheses. But if you never use $&, etc., in your script,
436 then regexps I<without> capturing parentheses won't be penalized. So
437 avoid $&, $', and $` if you can, but if you can't (and some algorithms
438 really appreciate them), once you've used them once, use them at will,
439 because you've already paid the price.
441 =head2 What good is C<\G> in a regular expression?
443 The notation C<\G> is used in a match or substitution in conjunction the
444 C</g> modifier (and ignored if there's no C</g>) to anchor the regular
445 expression to the point just past where the last match occurred, i.e. the
448 For example, suppose you had a line of text quoted in standard mail
449 and Usenet notation, (that is, with leading C<E<gt>> characters), and
450 you want change each leading C<E<gt>> into a corresponding C<:>. You
451 could do so in this way:
453 s/^(>+)/':' x length($1)/gem;
455 Or, using C<\G>, the much simpler (and faster):
459 A more sophisticated use might involve a tokenizer. The following
460 lex-like example is courtesy of Jeffrey Friedl. It did not work in
461 5.003 due to bugs in that release, but does work in 5.004 or better:
466 m/ \G( \d+\b )/gx && do { print "number: $1\n"; redo; };
467 m/ \G( \w+ )/gx && do { print "word: $1\n"; redo; };
468 m/ \G( \s+ )/gx && do { print "space: $1\n"; redo; };
469 m/ \G( [^\w\d]+ )/gx && do { print "other: $1\n"; redo; };
473 Of course, that could have been written as
478 if ( /\G( \d+\b )/gx {
479 print "number: $1\n";
490 if ( /\G( [^\w\d]+ )/gx {
497 But then you lose the vertical alignment of the regular expressions.
499 =head2 Are Perl regexps DFAs or NFAs? Are they POSIX compliant?
501 While it's true that Perl's regular expressions resemble the DFAs
502 (deterministic finite automata) of the egrep(1) program, they are in
503 fact implemented as NFAs (non-deterministic finite automata) to allow
504 backtracking and backreferencing. And they aren't POSIX-style either,
505 because those guarantee worst-case behavior for all cases. (It seems
506 that some people prefer guarantees of consistency, even when what's
507 guaranteed is slowness.) See the book "Mastering Regular Expressions"
508 (from O'Reilly) by Jeffrey Friedl for all the details you could ever
509 hope to know on these matters (a full citation appears in
512 =head2 What's wrong with using grep or map in a void context?
514 Strictly speaking, nothing. Stylistically speaking, it's not a good
515 way to write maintainable code. That's because you're using these
516 constructs not for their return values but rather for their
517 side-effects, and side-effects can be mystifying. There's no void
518 grep() that's not better written as a C<for> (well, C<foreach>,
521 =head2 How can I match strings with multi-byte characters?
523 This is hard, and there's no good way. Perl does not directly support
524 wide characters. It pretends that a byte and a character are
525 synonymous. The following set of approaches was offered by Jeffrey
526 Friedl, whose article in issue #5 of The Perl Journal talks about this
529 Let's suppose you have some weird Martian encoding where pairs of ASCII
530 uppercase letters encode single Martian letters (i.e. the two bytes
531 "CV" make a single Martian letter, as do the two bytes "SG", "VS",
532 "XX", etc.). Other bytes represent single characters, just like ASCII.
534 So, the string of Martian "I am CVSGXX!" uses 12 bytes to encode the nine
535 characters 'I', ' ', 'a', 'm', ' ', 'CV', 'SG', 'XX', '!'.
537 Now, say you want to search for the single character C</GX/>. Perl
538 doesn't know about Martian, so it'll find the two bytes "GX" in the
539 "I am CVSGXX!" string, even though that character isn't there: it just
540 looks like it is because "SG" is next to "XX", but there's no real "GX".
541 This is a big problem.
543 Here are a few ways, all painful, to deal with it:
545 $martian =~ s/([A-Z][A-Z])/ $1 /g; # Make sure adjacent ``maritan'' bytes
546 # are no longer adjacent.
547 print "found GX!\n" if $martian =~ /GX/;
551 @chars = $martian =~ m/([A-Z][A-Z]|[^A-Z])/g;
552 # above is conceptually similar to: @chars = $text =~ m/(.)/g;
554 foreach $char (@chars) {
555 print "found GX!\n", last if $char eq 'GX';
560 while ($martian =~ m/\G([A-Z][A-Z]|.)/gs) { # \G probably unneeded
561 print "found GX!\n", last if $1 eq 'GX';
566 die "sorry, Perl doesn't (yet) have Martian support )-:\n";
568 In addition, a sample program which converts half-width to full-width
569 katakana (in Shift-JIS or EUC encoding) is available from CPAN as
573 There are many double- (and multi-) byte encodings commonly used these
574 days. Some versions of these have 1-, 2-, 3-, and 4-byte characters,
577 =head1 AUTHOR AND COPYRIGHT
579 Copyright (c) 1997 Tom Christiansen and Nathan Torkington.
580 All rights reserved. See L<perlfaq> for distribution information.