This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlretut: incorrect output in "Non-capturing groupings"
[perl5.git] / pod / perlretut.pod
CommitLineData
47f9c88b
GS
1=head1 NAME
2
3perlretut - Perl regular expressions tutorial
4
5=head1 DESCRIPTION
6
7This page provides a basic tutorial on understanding, creating and
8using regular expressions in Perl. It serves as a complement to the
9reference page on regular expressions L<perlre>. Regular expressions
10are an integral part of the C<m//>, C<s///>, C<qr//> and C<split>
11operators and so this tutorial also overlaps with
12L<perlop/"Regexp Quote-Like Operators"> and L<perlfunc/split>.
13
14Perl is widely renowned for excellence in text processing, and regular
15expressions are one of the big factors behind this fame. Perl regular
16expressions display an efficiency and flexibility unknown in most
17other computer languages. Mastering even the basics of regular
18expressions will allow you to manipulate text with surprising ease.
19
20What is a regular expression? A regular expression is simply a string
21that describes a pattern. Patterns are in common use these days;
22examples are the patterns typed into a search engine to find web pages
23and the patterns used to list files in a directory, e.g., C<ls *.txt>
24or C<dir *.*>. In Perl, the patterns described by regular expressions
25are used to search strings, extract desired parts of strings, and to
26do search and replace operations.
27
28Regular expressions have the undeserved reputation of being abstract
29and difficult to understand. Regular expressions are constructed using
30simple concepts like conditionals and loops and are no more difficult
31to understand than the corresponding C<if> conditionals and C<while>
32loops in the Perl language itself. In fact, the main challenge in
33learning regular expressions is just getting used to the terse
34notation used to express these concepts.
35
36This tutorial flattens the learning curve by discussing regular
37expression concepts, along with their notation, one at a time and with
38many examples. The first part of the tutorial will progress from the
39simplest word searches to the basic regular expression concepts. If
40you master the first part, you will have all the tools needed to solve
41about 98% of your needs. The second part of the tutorial is for those
42comfortable with the basics and hungry for more power tools. It
43discusses the more advanced regular expression operators and
44introduces the latest cutting edge innovations in 5.6.0.
45
46A note: to save time, 'regular expression' is often abbreviated as
47regexp or regex. Regexp is a more natural abbreviation than regex, but
48is harder to pronounce. The Perl pod documentation is evenly split on
49regexp vs regex; in Perl, there is more than one way to abbreviate it.
50We'll use regexp in this tutorial.
51
52=head1 Part 1: The basics
53
54=head2 Simple word matching
55
56The simplest regexp is simply a word, or more generally, a string of
57characters. A regexp consisting of a word matches any string that
58contains that word:
59
60 "Hello World" =~ /World/; # matches
61
7638d2dc 62What is this Perl statement all about? C<"Hello World"> is a simple
47f9c88b 63double quoted string. C<World> is the regular expression and the
7638d2dc 64C<//> enclosing C</World/> tells Perl to search a string for a match.
47f9c88b
GS
65The operator C<=~> associates the string with the regexp match and
66produces a true value if the regexp matched, or false if the regexp
67did not match. In our case, C<World> matches the second word in
68C<"Hello World">, so the expression is true. Expressions like this
69are useful in conditionals:
70
71 if ("Hello World" =~ /World/) {
72 print "It matches\n";
73 }
74 else {
75 print "It doesn't match\n";
76 }
77
78There are useful variations on this theme. The sense of the match can
7638d2dc 79be reversed by using the C<!~> operator:
47f9c88b
GS
80
81 if ("Hello World" !~ /World/) {
82 print "It doesn't match\n";
83 }
84 else {
85 print "It matches\n";
86 }
87
88The literal string in the regexp can be replaced by a variable:
89
90 $greeting = "World";
91 if ("Hello World" =~ /$greeting/) {
92 print "It matches\n";
93 }
94 else {
95 print "It doesn't match\n";
96 }
97
98If you're matching against the special default variable C<$_>, the
99C<$_ =~> part can be omitted:
100
101 $_ = "Hello World";
102 if (/World/) {
103 print "It matches\n";
104 }
105 else {
106 print "It doesn't match\n";
107 }
108
109And finally, the C<//> default delimiters for a match can be changed
110to arbitrary delimiters by putting an C<'m'> out front:
111
112 "Hello World" =~ m!World!; # matches, delimited by '!'
113 "Hello World" =~ m{World}; # matches, note the matching '{}'
a6b2f353
GS
114 "/usr/bin/perl" =~ m"/perl"; # matches after '/usr/bin',
115 # '/' becomes an ordinary char
47f9c88b
GS
116
117C</World/>, C<m!World!>, and C<m{World}> all represent the
7638d2dc
WL
118same thing. When, e.g., the quote (C<">) is used as a delimiter, the forward
119slash C<'/'> becomes an ordinary character and can be used in this regexp
47f9c88b
GS
120without trouble.
121
122Let's consider how different regexps would match C<"Hello World">:
123
124 "Hello World" =~ /world/; # doesn't match
125 "Hello World" =~ /o W/; # matches
126 "Hello World" =~ /oW/; # doesn't match
127 "Hello World" =~ /World /; # doesn't match
128
129The first regexp C<world> doesn't match because regexps are
130case-sensitive. The second regexp matches because the substring
7638d2dc 131S<C<'o W'>> occurs in the string S<C<"Hello World">>. The space
47f9c88b
GS
132character ' ' is treated like any other character in a regexp and is
133needed to match in this case. The lack of a space character is the
134reason the third regexp C<'oW'> doesn't match. The fourth regexp
135C<'World '> doesn't match because there is a space at the end of the
136regexp, but not at the end of the string. The lesson here is that
137regexps must match a part of the string I<exactly> in order for the
138statement to be true.
139
7638d2dc 140If a regexp matches in more than one place in the string, Perl will
47f9c88b
GS
141always match at the earliest possible point in the string:
142
143 "Hello World" =~ /o/; # matches 'o' in 'Hello'
144 "That hat is red" =~ /hat/; # matches 'hat' in 'That'
145
146With respect to character matching, there are a few more points you
147need to know about. First of all, not all characters can be used 'as
7638d2dc 148is' in a match. Some characters, called I<metacharacters>, are reserved
47f9c88b
GS
149for use in regexp notation. The metacharacters are
150
151 {}[]()^$.|*+?\
152
153The significance of each of these will be explained
154in the rest of the tutorial, but for now, it is important only to know
155that a metacharacter can be matched by putting a backslash before it:
156
157 "2+2=4" =~ /2+2/; # doesn't match, + is a metacharacter
158 "2+2=4" =~ /2\+2/; # matches, \+ is treated like an ordinary +
159 "The interval is [0,1)." =~ /[0,1)./ # is a syntax error!
160 "The interval is [0,1)." =~ /\[0,1\)\./ # matches
7638d2dc 161 "#!/usr/bin/perl" =~ /#!\/usr\/bin\/perl/; # matches
47f9c88b
GS
162
163In the last regexp, the forward slash C<'/'> is also backslashed,
164because it is used to delimit the regexp. This can lead to LTS
165(leaning toothpick syndrome), however, and it is often more readable
166to change delimiters.
167
7638d2dc 168 "#!/usr/bin/perl" =~ m!#\!/usr/bin/perl!; # easier to read
47f9c88b
GS
169
170The backslash character C<'\'> is a metacharacter itself and needs to
171be backslashed:
172
173 'C:\WIN32' =~ /C:\\WIN/; # matches
174
175In addition to the metacharacters, there are some ASCII characters
176which don't have printable character equivalents and are instead
7638d2dc 177represented by I<escape sequences>. Common examples are C<\t> for a
47f9c88b
GS
178tab, C<\n> for a newline, C<\r> for a carriage return and C<\a> for a
179bell. If your string is better thought of as a sequence of arbitrary
180bytes, the octal escape sequence, e.g., C<\033>, or hexadecimal escape
181sequence, e.g., C<\x1B> may be a more natural representation for your
182bytes. Here are some examples of escapes:
183
184 "1000\t2000" =~ m(0\t2) # matches
185 "1000\n2000" =~ /0\n20/ # matches
186 "1000\t2000" =~ /\000\t2/ # doesn't match, "0" ne "\000"
f0a2b745
KW
187 "cat" =~ /\o{143}\x61\x74/ # matches in ASCII, but a weird way
188 # to spell cat
47f9c88b
GS
189
190If you've been around Perl a while, all this talk of escape sequences
191may seem familiar. Similar escape sequences are used in double-quoted
192strings and in fact the regexps in Perl are mostly treated as
193double-quoted strings. This means that variables can be used in
194regexps as well. Just like double-quoted strings, the values of the
195variables in the regexp will be substituted in before the regexp is
196evaluated for matching purposes. So we have:
197
198 $foo = 'house';
199 'housecat' =~ /$foo/; # matches
200 'cathouse' =~ /cat$foo/; # matches
47f9c88b
GS
201 'housecat' =~ /${foo}cat/; # matches
202
203So far, so good. With the knowledge above you can already perform
204searches with just about any literal string regexp you can dream up.
205Here is a I<very simple> emulation of the Unix grep program:
206
207 % cat > simple_grep
208 #!/usr/bin/perl
209 $regexp = shift;
210 while (<>) {
211 print if /$regexp/;
212 }
213 ^D
214
215 % chmod +x simple_grep
216
217 % simple_grep abba /usr/dict/words
218 Babbage
219 cabbage
220 cabbages
221 sabbath
222 Sabbathize
223 Sabbathizes
224 sabbatical
225 scabbard
226 scabbards
227
228This program is easy to understand. C<#!/usr/bin/perl> is the standard
229way to invoke a perl program from the shell.
7638d2dc 230S<C<$regexp = shift;>> saves the first command line argument as the
47f9c88b 231regexp to be used, leaving the rest of the command line arguments to
7638d2dc
WL
232be treated as files. S<C<< while (<>) >>> loops over all the lines in
233all the files. For each line, S<C<print if /$regexp/;>> prints the
47f9c88b
GS
234line if the regexp matches the line. In this line, both C<print> and
235C</$regexp/> use the default variable C<$_> implicitly.
236
237With all of the regexps above, if the regexp matched anywhere in the
238string, it was considered a match. Sometimes, however, we'd like to
239specify I<where> in the string the regexp should try to match. To do
7638d2dc 240this, we would use the I<anchor> metacharacters C<^> and C<$>. The
47f9c88b
GS
241anchor C<^> means match at the beginning of the string and the anchor
242C<$> means match at the end of the string, or before a newline at the
243end of the string. Here is how they are used:
244
245 "housekeeper" =~ /keeper/; # matches
246 "housekeeper" =~ /^keeper/; # doesn't match
247 "housekeeper" =~ /keeper$/; # matches
248 "housekeeper\n" =~ /keeper$/; # matches
249
250The second regexp doesn't match because C<^> constrains C<keeper> to
251match only at the beginning of the string, but C<"housekeeper"> has
252keeper starting in the middle. The third regexp does match, since the
253C<$> constrains C<keeper> to match only at the end of the string.
254
255When both C<^> and C<$> are used at the same time, the regexp has to
256match both the beginning and the end of the string, i.e., the regexp
257matches the whole string. Consider
258
259 "keeper" =~ /^keep$/; # doesn't match
260 "keeper" =~ /^keeper$/; # matches
261 "" =~ /^$/; # ^$ matches an empty string
262
263The first regexp doesn't match because the string has more to it than
264C<keep>. Since the second regexp is exactly the string, it
265matches. Using both C<^> and C<$> in a regexp forces the complete
266string to match, so it gives you complete control over which strings
267match and which don't. Suppose you are looking for a fellow named
268bert, off in a string by himself:
269
270 "dogbert" =~ /bert/; # matches, but not what you want
271
272 "dilbert" =~ /^bert/; # doesn't match, but ..
273 "bertram" =~ /^bert/; # matches, so still not good enough
274
275 "bertram" =~ /^bert$/; # doesn't match, good
276 "dilbert" =~ /^bert$/; # doesn't match, good
277 "bert" =~ /^bert$/; # matches, perfect
278
279Of course, in the case of a literal string, one could just as easily
7638d2dc 280use the string comparison S<C<$string eq 'bert'>> and it would be
47f9c88b
GS
281more efficient. The C<^...$> regexp really becomes useful when we
282add in the more powerful regexp tools below.
283
284=head2 Using character classes
285
286Although one can already do quite a lot with the literal string
287regexps above, we've only scratched the surface of regular expression
288technology. In this and subsequent sections we will introduce regexp
289concepts (and associated metacharacter notations) that will allow a
290regexp to not just represent a single character sequence, but a I<whole
291class> of them.
292
7638d2dc 293One such concept is that of a I<character class>. A character class
47f9c88b
GS
294allows a set of possible characters, rather than just a single
295character, to match at a particular point in a regexp. Character
296classes are denoted by brackets C<[...]>, with the set of characters
297to be possibly matched inside. Here are some examples:
298
299 /cat/; # matches 'cat'
300 /[bcr]at/; # matches 'bat, 'cat', or 'rat'
301 /item[0123456789]/; # matches 'item0' or ... or 'item9'
a6b2f353 302 "abc" =~ /[cab]/; # matches 'a'
47f9c88b
GS
303
304In the last statement, even though C<'c'> is the first character in
305the class, C<'a'> matches because the first character position in the
306string is the earliest point at which the regexp can match.
307
308 /[yY][eE][sS]/; # match 'yes' in a case-insensitive way
309 # 'yes', 'Yes', 'YES', etc.
310
da75cd15 311This regexp displays a common task: perform a case-insensitive
28c3722c 312match. Perl provides a way of avoiding all those brackets by simply
47f9c88b
GS
313appending an C<'i'> to the end of the match. Then C</[yY][eE][sS]/;>
314can be rewritten as C</yes/i;>. The C<'i'> stands for
7638d2dc 315case-insensitive and is an example of a I<modifier> of the matching
47f9c88b
GS
316operation. We will meet other modifiers later in the tutorial.
317
318We saw in the section above that there were ordinary characters, which
319represented themselves, and special characters, which needed a
320backslash C<\> to represent themselves. The same is true in a
321character class, but the sets of ordinary and special characters
322inside a character class are different than those outside a character
7638d2dc 323class. The special characters for a character class are C<-]\^$> (and
353c6505 324the pattern delimiter, whatever it is).
7638d2dc 325C<]> is special because it denotes the end of a character class. C<$> is
47f9c88b
GS
326special because it denotes a scalar variable. C<\> is special because
327it is used in escape sequences, just like above. Here is how the
328special characters C<]$\> are handled:
329
330 /[\]c]def/; # matches ']def' or 'cdef'
331 $x = 'bcr';
a6b2f353 332 /[$x]at/; # matches 'bat', 'cat', or 'rat'
47f9c88b
GS
333 /[\$x]at/; # matches '$at' or 'xat'
334 /[\\$x]at/; # matches '\at', 'bat, 'cat', or 'rat'
335
353c6505 336The last two are a little tricky. In C<[\$x]>, the backslash protects
47f9c88b
GS
337the dollar sign, so the character class has two members C<$> and C<x>.
338In C<[\\$x]>, the backslash is protected, so C<$x> is treated as a
339variable and substituted in double quote fashion.
340
341The special character C<'-'> acts as a range operator within character
342classes, so that a contiguous set of characters can be written as a
343range. With ranges, the unwieldy C<[0123456789]> and C<[abc...xyz]>
344become the svelte C<[0-9]> and C<[a-z]>. Some examples are
345
346 /item[0-9]/; # matches 'item0' or ... or 'item9'
347 /[0-9bx-z]aa/; # matches '0aa', ..., '9aa',
348 # 'baa', 'xaa', 'yaa', or 'zaa'
349 /[0-9a-fA-F]/; # matches a hexadecimal digit
36bbe248 350 /[0-9a-zA-Z_]/; # matches a "word" character,
7638d2dc 351 # like those in a Perl variable name
47f9c88b
GS
352
353If C<'-'> is the first or last character in a character class, it is
354treated as an ordinary character; C<[-ab]>, C<[ab-]> and C<[a\-b]> are
355all equivalent.
356
357The special character C<^> in the first position of a character class
7638d2dc 358denotes a I<negated character class>, which matches any character but
a6b2f353 359those in the brackets. Both C<[...]> and C<[^...]> must match a
47f9c88b
GS
360character, or the match fails. Then
361
362 /[^a]at/; # doesn't match 'aat' or 'at', but matches
363 # all other 'bat', 'cat, '0at', '%at', etc.
364 /[^0-9]/; # matches a non-numeric character
365 /[a^]at/; # matches 'aat' or '^at'; here '^' is ordinary
366
28c3722c 367Now, even C<[0-9]> can be a bother to write multiple times, so in the
47f9c88b 368interest of saving keystrokes and making regexps more readable, Perl
7638d2dc
WL
369has several abbreviations for common character classes, as shown below.
370Since the introduction of Unicode, these character classes match more
371than just a few characters in the ISO 8859-1 range.
47f9c88b
GS
372
373=over 4
374
375=item *
551e1d92 376
7638d2dc 377\d matches a digit, not just [0-9] but also digits from non-roman scripts
47f9c88b
GS
378
379=item *
551e1d92 380
7638d2dc 381\s matches a whitespace character, the set [\ \t\r\n\f] and others
47f9c88b
GS
382
383=item *
551e1d92 384
7638d2dc
WL
385\w matches a word character (alphanumeric or _), not just [0-9a-zA-Z_]
386but also digits and characters from non-roman scripts
47f9c88b
GS
387
388=item *
551e1d92 389
7638d2dc 390\D is a negated \d; it represents any other character than a digit, or [^\d]
47f9c88b
GS
391
392=item *
551e1d92 393
47f9c88b
GS
394\S is a negated \s; it represents any non-whitespace character [^\s]
395
396=item *
551e1d92 397
47f9c88b
GS
398\W is a negated \w; it represents any non-word character [^\w]
399
400=item *
551e1d92 401
7638d2dc
WL
402The period '.' matches any character but "\n" (unless the modifier C<//s> is
403in effect, as explained below).
47f9c88b
GS
404
405=back
406
407The C<\d\s\w\D\S\W> abbreviations can be used both inside and outside
408of character classes. Here are some in use:
409
410 /\d\d:\d\d:\d\d/; # matches a hh:mm:ss time format
411 /[\d\s]/; # matches any digit or whitespace character
412 /\w\W\w/; # matches a word char, followed by a
413 # non-word char, followed by a word char
414 /..rt/; # matches any two chars, followed by 'rt'
415 /end\./; # matches 'end.'
416 /end[.]/; # same thing, matches 'end.'
417
418Because a period is a metacharacter, it needs to be escaped to match
419as an ordinary period. Because, for example, C<\d> and C<\w> are sets
420of characters, it is incorrect to think of C<[^\d\w]> as C<[\D\W]>; in
421fact C<[^\d\w]> is the same as C<[^\w]>, which is the same as
422C<[\W]>. Think DeMorgan's laws.
423
7638d2dc 424An anchor useful in basic regexps is the I<word anchor>
47f9c88b
GS
425C<\b>. This matches a boundary between a word character and a non-word
426character C<\w\W> or C<\W\w>:
427
428 $x = "Housecat catenates house and cat";
429 $x =~ /cat/; # matches cat in 'housecat'
430 $x =~ /\bcat/; # matches cat in 'catenates'
431 $x =~ /cat\b/; # matches cat in 'housecat'
432 $x =~ /\bcat\b/; # matches 'cat' at end of string
433
434Note in the last example, the end of the string is considered a word
435boundary.
436
437You might wonder why C<'.'> matches everything but C<"\n"> - why not
438every character? The reason is that often one is matching against
439lines and would like to ignore the newline characters. For instance,
440while the string C<"\n"> represents one line, we would like to think
28c3722c 441of it as empty. Then
47f9c88b
GS
442
443 "" =~ /^$/; # matches
7638d2dc 444 "\n" =~ /^$/; # matches, $ anchors before "\n"
47f9c88b
GS
445
446 "" =~ /./; # doesn't match; it needs a char
447 "" =~ /^.$/; # doesn't match; it needs a char
448 "\n" =~ /^.$/; # doesn't match; it needs a char other than "\n"
449 "a" =~ /^.$/; # matches
7638d2dc 450 "a\n" =~ /^.$/; # matches, $ anchors before "\n"
47f9c88b
GS
451
452This behavior is convenient, because we usually want to ignore
453newlines when we count and match characters in a line. Sometimes,
454however, we want to keep track of newlines. We might even want C<^>
455and C<$> to anchor at the beginning and end of lines within the
456string, rather than just the beginning and end of the string. Perl
457allows us to choose between ignoring and paying attention to newlines
458by using the C<//s> and C<//m> modifiers. C<//s> and C<//m> stand for
459single line and multi-line and they determine whether a string is to
460be treated as one continuous string, or as a set of lines. The two
461modifiers affect two aspects of how the regexp is interpreted: 1) how
462the C<'.'> character class is defined, and 2) where the anchors C<^>
463and C<$> are able to match. Here are the four possible combinations:
464
465=over 4
466
467=item *
551e1d92 468
47f9c88b
GS
469no modifiers (//): Default behavior. C<'.'> matches any character
470except C<"\n">. C<^> matches only at the beginning of the string and
471C<$> matches only at the end or before a newline at the end.
472
473=item *
551e1d92 474
47f9c88b
GS
475s modifier (//s): Treat string as a single long line. C<'.'> matches
476any character, even C<"\n">. C<^> matches only at the beginning of
477the string and C<$> matches only at the end or before a newline at the
478end.
479
480=item *
551e1d92 481
47f9c88b
GS
482m modifier (//m): Treat string as a set of multiple lines. C<'.'>
483matches any character except C<"\n">. C<^> and C<$> are able to match
484at the start or end of I<any> line within the string.
485
486=item *
551e1d92 487
47f9c88b
GS
488both s and m modifiers (//sm): Treat string as a single long line, but
489detect multiple lines. C<'.'> matches any character, even
490C<"\n">. C<^> and C<$>, however, are able to match at the start or end
491of I<any> line within the string.
492
493=back
494
495Here are examples of C<//s> and C<//m> in action:
496
497 $x = "There once was a girl\nWho programmed in Perl\n";
498
499 $x =~ /^Who/; # doesn't match, "Who" not at start of string
500 $x =~ /^Who/s; # doesn't match, "Who" not at start of string
501 $x =~ /^Who/m; # matches, "Who" at start of second line
502 $x =~ /^Who/sm; # matches, "Who" at start of second line
503
504 $x =~ /girl.Who/; # doesn't match, "." doesn't match "\n"
505 $x =~ /girl.Who/s; # matches, "." matches "\n"
506 $x =~ /girl.Who/m; # doesn't match, "." doesn't match "\n"
507 $x =~ /girl.Who/sm; # matches, "." matches "\n"
508
3c12f9b9 509Most of the time, the default behavior is what is wanted, but C<//s> and
47f9c88b 510C<//m> are occasionally very useful. If C<//m> is being used, the start
28c3722c 511of the string can still be matched with C<\A> and the end of the string
47f9c88b
GS
512can still be matched with the anchors C<\Z> (matches both the end and
513the newline before, like C<$>), and C<\z> (matches only the end):
514
515 $x =~ /^Who/m; # matches, "Who" at start of second line
516 $x =~ /\AWho/m; # doesn't match, "Who" is not at start of string
517
518 $x =~ /girl$/m; # matches, "girl" at end of first line
519 $x =~ /girl\Z/m; # doesn't match, "girl" is not at end of string
520
521 $x =~ /Perl\Z/m; # matches, "Perl" is at newline before end
522 $x =~ /Perl\z/m; # doesn't match, "Perl" is not at end of string
523
524We now know how to create choices among classes of characters in a
525regexp. What about choices among words or character strings? Such
526choices are described in the next section.
527
528=head2 Matching this or that
529
28c3722c 530Sometimes we would like our regexp to be able to match different
47f9c88b 531possible words or character strings. This is accomplished by using
7638d2dc
WL
532the I<alternation> metacharacter C<|>. To match C<dog> or C<cat>, we
533form the regexp C<dog|cat>. As before, Perl will try to match the
47f9c88b 534regexp at the earliest possible point in the string. At each
7638d2dc
WL
535character position, Perl will first try to match the first
536alternative, C<dog>. If C<dog> doesn't match, Perl will then try the
47f9c88b 537next alternative, C<cat>. If C<cat> doesn't match either, then the
7638d2dc 538match fails and Perl moves to the next position in the string. Some
47f9c88b
GS
539examples:
540
541 "cats and dogs" =~ /cat|dog|bird/; # matches "cat"
542 "cats and dogs" =~ /dog|cat|bird/; # matches "cat"
543
544Even though C<dog> is the first alternative in the second regexp,
545C<cat> is able to match earlier in the string.
546
547 "cats" =~ /c|ca|cat|cats/; # matches "c"
548 "cats" =~ /cats|cat|ca|c/; # matches "cats"
549
550Here, all the alternatives match at the first string position, so the
551first alternative is the one that matches. If some of the
552alternatives are truncations of the others, put the longest ones first
553to give them a chance to match.
554
555 "cab" =~ /a|b|c/ # matches "c"
556 # /a|b|c/ == /[abc]/
557
558The last example points out that character classes are like
559alternations of characters. At a given character position, the first
210b36aa 560alternative that allows the regexp match to succeed will be the one
47f9c88b
GS
561that matches.
562
563=head2 Grouping things and hierarchical matching
564
565Alternation allows a regexp to choose among alternatives, but by
7638d2dc 566itself it is unsatisfying. The reason is that each alternative is a whole
47f9c88b
GS
567regexp, but sometime we want alternatives for just part of a
568regexp. For instance, suppose we want to search for housecats or
569housekeepers. The regexp C<housecat|housekeeper> fits the bill, but is
570inefficient because we had to type C<house> twice. It would be nice to
da75cd15 571have parts of the regexp be constant, like C<house>, and some
47f9c88b
GS
572parts have alternatives, like C<cat|keeper>.
573
7638d2dc 574The I<grouping> metacharacters C<()> solve this problem. Grouping
47f9c88b
GS
575allows parts of a regexp to be treated as a single unit. Parts of a
576regexp are grouped by enclosing them in parentheses. Thus we could solve
577the C<housecat|housekeeper> by forming the regexp as
578C<house(cat|keeper)>. The regexp C<house(cat|keeper)> means match
579C<house> followed by either C<cat> or C<keeper>. Some more examples
580are
581
582 /(a|b)b/; # matches 'ab' or 'bb'
583 /(ac|b)b/; # matches 'acb' or 'bb'
584 /(^a|b)c/; # matches 'ac' at start of string or 'bc' anywhere
585 /(a|[bc])d/; # matches 'ad', 'bd', or 'cd'
586
587 /house(cat|)/; # matches either 'housecat' or 'house'
588 /house(cat(s|)|)/; # matches either 'housecats' or 'housecat' or
589 # 'house'. Note groups can be nested.
590
591 /(19|20|)\d\d/; # match years 19xx, 20xx, or the Y2K problem, xx
592 "20" =~ /(19|20|)\d\d/; # matches the null alternative '()\d\d',
593 # because '20\d\d' can't match
594
595Alternations behave the same way in groups as out of them: at a given
596string position, the leftmost alternative that allows the regexp to
210b36aa 597match is taken. So in the last example at the first string position,
47f9c88b 598C<"20"> matches the second alternative, but there is nothing left over
7638d2dc 599to match the next two digits C<\d\d>. So Perl moves on to the next
47f9c88b
GS
600alternative, which is the null alternative and that works, since
601C<"20"> is two digits.
602
603The process of trying one alternative, seeing if it matches, and
7638d2dc
WL
604moving on to the next alternative, while going back in the string
605from where the previous alternative was tried, if it doesn't, is called
606I<backtracking>. The term 'backtracking' comes from the idea that
47f9c88b
GS
607matching a regexp is like a walk in the woods. Successfully matching
608a regexp is like arriving at a destination. There are many possible
609trailheads, one for each string position, and each one is tried in
610order, left to right. From each trailhead there may be many paths,
611some of which get you there, and some which are dead ends. When you
612walk along a trail and hit a dead end, you have to backtrack along the
613trail to an earlier point to try another trail. If you hit your
614destination, you stop immediately and forget about trying all the
615other trails. You are persistent, and only if you have tried all the
616trails from all the trailheads and not arrived at your destination, do
617you declare failure. To be concrete, here is a step-by-step analysis
7638d2dc 618of what Perl does when it tries to match the regexp
47f9c88b
GS
619
620 "abcde" =~ /(abd|abc)(df|d|de)/;
621
622=over 4
623
551e1d92
RB
624=item 0
625
626Start with the first letter in the string 'a'.
627
628=item 1
47f9c88b 629
551e1d92 630Try the first alternative in the first group 'abd'.
47f9c88b 631
551e1d92 632=item 2
47f9c88b 633
551e1d92
RB
634Match 'a' followed by 'b'. So far so good.
635
636=item 3
637
638'd' in the regexp doesn't match 'c' in the string - a dead
47f9c88b
GS
639end. So backtrack two characters and pick the second alternative in
640the first group 'abc'.
641
551e1d92
RB
642=item 4
643
644Match 'a' followed by 'b' followed by 'c'. We are on a roll
47f9c88b
GS
645and have satisfied the first group. Set $1 to 'abc'.
646
551e1d92
RB
647=item 5
648
649Move on to the second group and pick the first alternative
47f9c88b
GS
650'df'.
651
551e1d92 652=item 6
47f9c88b 653
551e1d92
RB
654Match the 'd'.
655
656=item 7
657
658'f' in the regexp doesn't match 'e' in the string, so a dead
47f9c88b
GS
659end. Backtrack one character and pick the second alternative in the
660second group 'd'.
661
551e1d92
RB
662=item 8
663
664'd' matches. The second grouping is satisfied, so set $2 to
47f9c88b
GS
665'd'.
666
551e1d92
RB
667=item 9
668
669We are at the end of the regexp, so we are done! We have
47f9c88b
GS
670matched 'abcd' out of the string "abcde".
671
672=back
673
674There are a couple of things to note about this analysis. First, the
675third alternative in the second group 'de' also allows a match, but we
676stopped before we got to it - at a given character position, leftmost
677wins. Second, we were able to get a match at the first character
678position of the string 'a'. If there were no matches at the first
7638d2dc 679position, Perl would move to the second character position 'b' and
47f9c88b 680attempt the match all over again. Only when all possible paths at all
7638d2dc
WL
681possible character positions have been exhausted does Perl give
682up and declare S<C<$string =~ /(abd|abc)(df|d|de)/;>> to be false.
47f9c88b
GS
683
684Even with all this work, regexp matching happens remarkably fast. To
353c6505
DL
685speed things up, Perl compiles the regexp into a compact sequence of
686opcodes that can often fit inside a processor cache. When the code is
7638d2dc
WL
687executed, these opcodes can then run at full throttle and search very
688quickly.
47f9c88b
GS
689
690=head2 Extracting matches
691
692The grouping metacharacters C<()> also serve another completely
693different function: they allow the extraction of the parts of a string
694that matched. This is very useful to find out what matched and for
695text processing in general. For each grouping, the part that matched
696inside goes into the special variables C<$1>, C<$2>, etc. They can be
697used just as ordinary variables:
698
699 # extract hours, minutes, seconds
2275acdc
RGS
700 if ($time =~ /(\d\d):(\d\d):(\d\d)/) { # match hh:mm:ss format
701 $hours = $1;
702 $minutes = $2;
703 $seconds = $3;
704 }
47f9c88b
GS
705
706Now, we know that in scalar context,
7638d2dc 707S<C<$time =~ /(\d\d):(\d\d):(\d\d)/>> returns a true or false
47f9c88b
GS
708value. In list context, however, it returns the list of matched values
709C<($1,$2,$3)>. So we could write the code more compactly as
710
711 # extract hours, minutes, seconds
712 ($hours, $minutes, $second) = ($time =~ /(\d\d):(\d\d):(\d\d)/);
713
714If the groupings in a regexp are nested, C<$1> gets the group with the
715leftmost opening parenthesis, C<$2> the next opening parenthesis,
7638d2dc 716etc. Here is a regexp with nested groups:
47f9c88b
GS
717
718 /(ab(cd|ef)((gi)|j))/;
719 1 2 34
720
7638d2dc
WL
721If this regexp matches, C<$1> contains a string starting with
722C<'ab'>, C<$2> is either set to C<'cd'> or C<'ef'>, C<$3> equals either
723C<'gi'> or C<'j'>, and C<$4> is either set to C<'gi'>, just like C<$3>,
724or it remains undefined.
725
726For convenience, Perl sets C<$+> to the string held by the highest numbered
727C<$1>, C<$2>,... that got assigned (and, somewhat related, C<$^N> to the
728value of the C<$1>, C<$2>,... most-recently assigned; i.e. the C<$1>,
729C<$2>,... associated with the rightmost closing parenthesis used in the
a01268b5 730match).
47f9c88b 731
7638d2dc
WL
732
733=head2 Backreferences
734
47f9c88b 735Closely associated with the matching variables C<$1>, C<$2>, ... are
d8b950dc 736the I<backreferences> C<\g1>, C<\g2>,... Backreferences are simply
47f9c88b 737matching variables that can be used I<inside> a regexp. This is a
ac036724 738really nice feature; what matches later in a regexp is made to depend on
47f9c88b 739what matched earlier in the regexp. Suppose we wanted to look
7638d2dc 740for doubled words in a text, like 'the the'. The following regexp finds
47f9c88b
GS
741all 3-letter doubles with a space in between:
742
d8b950dc 743 /\b(\w\w\w)\s\g1\b/;
47f9c88b 744
d8b950dc 745The grouping assigns a value to \g1, so that the same 3 letter sequence
7638d2dc
WL
746is used for both parts.
747
748A similar task is to find words consisting of two identical parts:
47f9c88b 749
d8b950dc 750 % simple_grep '^(\w\w\w\w|\w\w\w|\w\w|\w)\g1$' /usr/dict/words
47f9c88b
GS
751 beriberi
752 booboo
753 coco
754 mama
755 murmur
756 papa
757
758The regexp has a single grouping which considers 4-letter
d8b950dc
KW
759combinations, then 3-letter combinations, etc., and uses C<\g1> to look for
760a repeat. Although C<$1> and C<\g1> represent the same thing, care should be
7638d2dc 761taken to use matched variables C<$1>, C<$2>,... only I<outside> a regexp
d8b950dc 762and backreferences C<\g1>, C<\g2>,... only I<inside> a regexp; not doing
7638d2dc
WL
763so may lead to surprising and unsatisfactory results.
764
765
766=head2 Relative backreferences
767
768Counting the opening parentheses to get the correct number for a
353c6505 769backreference is errorprone as soon as there is more than one
7638d2dc
WL
770capturing group. A more convenient technique became available
771with Perl 5.10: relative backreferences. To refer to the immediately
772preceding capture group one now may write C<\g{-1}>, the next but
773last is available via C<\g{-2}>, and so on.
774
775Another good reason in addition to readability and maintainability
776for using relative backreferences is illustrated by the following example,
777where a simple pattern for matching peculiar strings is used:
778
d8b950dc 779 $a99a = '([a-z])(\d)\g2\g1'; # matches a11a, g22g, x33x, etc.
7638d2dc
WL
780
781Now that we have this pattern stored as a handy string, we might feel
782tempted to use it as a part of some other pattern:
783
784 $line = "code=e99e";
785 if ($line =~ /^(\w+)=$a99a$/){ # unexpected behavior!
786 print "$1 is valid\n";
787 } else {
788 print "bad line: '$line'\n";
789 }
790
ac036724 791But this doesn't match, at least not the way one might expect. Only
7638d2dc
WL
792after inserting the interpolated C<$a99a> and looking at the resulting
793full text of the regexp is it obvious that the backreferences have
ac036724 794backfired. The subexpression C<(\w+)> has snatched number 1 and
7638d2dc
WL
795demoted the groups in C<$a99a> by one rank. This can be avoided by
796using relative backreferences:
797
798 $a99a = '([a-z])(\d)\g{-1}\g{-2}'; # safe for being interpolated
799
800
801=head2 Named backreferences
802
c27a5cfe 803Perl 5.10 also introduced named capture groups and named backreferences.
7638d2dc
WL
804To attach a name to a capturing group, you write either
805C<< (?<name>...) >> or C<< (?'name'...) >>. The backreference may
806then be written as C<\g{name}>. It is permissible to attach the
807same name to more than one group, but then only the leftmost one of the
808eponymous set can be referenced. Outside of the pattern a named
c27a5cfe 809capture group is accessible through the C<%+> hash.
7638d2dc 810
353c6505 811Assuming that we have to match calendar dates which may be given in one
7638d2dc 812of the three formats yyyy-mm-dd, mm/dd/yyyy or dd.mm.yyyy, we can write
353c6505 813three suitable patterns where we use 'd', 'm' and 'y' respectively as the
c27a5cfe 814names of the groups capturing the pertaining components of a date. The
7638d2dc
WL
815matching operation combines the three patterns as alternatives:
816
817 $fmt1 = '(?<y>\d\d\d\d)-(?<m>\d\d)-(?<d>\d\d)';
818 $fmt2 = '(?<m>\d\d)/(?<d>\d\d)/(?<y>\d\d\d\d)';
819 $fmt3 = '(?<d>\d\d)\.(?<m>\d\d)\.(?<y>\d\d\d\d)';
820 for my $d qw( 2006-10-21 15.01.2007 10/31/2005 ){
821 if ( $d =~ m{$fmt1|$fmt2|$fmt3} ){
822 print "day=$+{d} month=$+{m} year=$+{y}\n";
823 }
824 }
825
826If any of the alternatives matches, the hash C<%+> is bound to contain the
827three key-value pairs.
828
829
830=head2 Alternative capture group numbering
831
832Yet another capturing group numbering technique (also as from Perl 5.10)
833deals with the problem of referring to groups within a set of alternatives.
834Consider a pattern for matching a time of the day, civil or military style:
47f9c88b 835
7638d2dc
WL
836 if ( $time =~ /(\d\d|\d):(\d\d)|(\d\d)(\d\d)/ ){
837 # process hour and minute
838 }
839
840Processing the results requires an additional if statement to determine
353c6505 841whether C<$1> and C<$2> or C<$3> and C<$4> contain the goodies. It would
c27a5cfe 842be easier if we could use group numbers 1 and 2 in second alternative as
353c6505 843well, and this is exactly what the parenthesized construct C<(?|...)>,
7638d2dc
WL
844set around an alternative achieves. Here is an extended version of the
845previous pattern:
846
847 if ( $time =~ /(?|(\d\d|\d):(\d\d)|(\d\d)(\d\d))\s+([A-Z][A-Z][A-Z])/ ){
848 print "hour=$1 minute=$2 zone=$3\n";
849 }
850
c27a5cfe 851Within the alternative numbering group, group numbers start at the same
7638d2dc 852position for each alternative. After the group, numbering continues
353c6505 853with one higher than the maximum reached across all the alternatives.
7638d2dc
WL
854
855=head2 Position information
856
857In addition to what was matched, Perl (since 5.6.0) also provides the
858positions of what was matched as contents of the C<@-> and C<@+>
47f9c88b
GS
859arrays. C<$-[0]> is the position of the start of the entire match and
860C<$+[0]> is the position of the end. Similarly, C<$-[n]> is the
861position of the start of the C<$n> match and C<$+[n]> is the position
862of the end. If C<$n> is undefined, so are C<$-[n]> and C<$+[n]>. Then
863this code
864
865 $x = "Mmm...donut, thought Homer";
866 $x =~ /^(Mmm|Yech)\.\.\.(donut|peas)/; # matches
867 foreach $expr (1..$#-) {
868 print "Match $expr: '${$expr}' at position ($-[$expr],$+[$expr])\n";
869 }
870
871prints
872
873 Match 1: 'Mmm' at position (0,3)
874 Match 2: 'donut' at position (6,11)
875
876Even if there are no groupings in a regexp, it is still possible to
7638d2dc 877find out what exactly matched in a string. If you use them, Perl
47f9c88b
GS
878will set C<$`> to the part of the string before the match, will set C<$&>
879to the part of the string that matched, and will set C<$'> to the part
880of the string after the match. An example:
881
882 $x = "the cat caught the mouse";
883 $x =~ /cat/; # $` = 'the ', $& = 'cat', $' = ' caught the mouse'
884 $x =~ /the/; # $` = '', $& = 'the', $' = ' cat caught the mouse'
885
7638d2dc
WL
886In the second match, C<$`> equals C<''> because the regexp matched at the
887first character position in the string and stopped; it never saw the
47f9c88b 888second 'the'. It is important to note that using C<$`> and C<$'>
7638d2dc 889slows down regexp matching quite a bit, while C<$&> slows it down to a
47f9c88b 890lesser extent, because if they are used in one regexp in a program,
7638d2dc 891they are generated for I<all> regexps in the program. So if raw
47f9c88b 892performance is a goal of your application, they should be avoided.
7638d2dc
WL
893If you need to extract the corresponding substrings, use C<@-> and
894C<@+> instead:
47f9c88b
GS
895
896 $` is the same as substr( $x, 0, $-[0] )
897 $& is the same as substr( $x, $-[0], $+[0]-$-[0] )
898 $' is the same as substr( $x, $+[0] )
899
7638d2dc
WL
900
901=head2 Non-capturing groupings
902
353c6505 903A group that is required to bundle a set of alternatives may or may not be
7638d2dc 904useful as a capturing group. If it isn't, it just creates a superfluous
c27a5cfe 905addition to the set of available capture group values, inside as well as
7638d2dc 906outside the regexp. Non-capturing groupings, denoted by C<(?:regexp)>,
353c6505 907still allow the regexp to be treated as a single unit, but don't establish
c27a5cfe 908a capturing group at the same time. Both capturing and non-capturing
7638d2dc
WL
909groupings are allowed to co-exist in the same regexp. Because there is
910no extraction, non-capturing groupings are faster than capturing
911groupings. Non-capturing groupings are also handy for choosing exactly
912which parts of a regexp are to be extracted to matching variables:
913
914 # match a number, $1-$4 are set, but we only want $1
915 /([+-]?\ *(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?)/;
916
917 # match a number faster , only $1 is set
918 /([+-]?\ *(?:\d+(?:\.\d*)?|\.\d+)(?:[eE][+-]?\d+)?)/;
919
920 # match a number, get $1 = whole number, $2 = exponent
921 /([+-]?\ *(?:\d+(?:\.\d*)?|\.\d+)(?:[eE]([+-]?\d+))?)/;
922
923Non-capturing groupings are also useful for removing nuisance
924elements gathered from a split operation where parentheses are
925required for some reason:
926
927 $x = '12aba34ba5';
9b846e30 928 @num = split /(a|b)+/, $x; # @num = ('12','a','34','a','5')
7638d2dc
WL
929 @num = split /(?:a|b)+/, $x; # @num = ('12','34','5')
930
931
47f9c88b
GS
932=head2 Matching repetitions
933
934The examples in the previous section display an annoying weakness. We
7638d2dc
WL
935were only matching 3-letter words, or chunks of words of 4 letters or
936less. We'd like to be able to match words or, more generally, strings
937of any length, without writing out tedious alternatives like
47f9c88b
GS
938C<\w\w\w\w|\w\w\w|\w\w|\w>.
939
7638d2dc
WL
940This is exactly the problem the I<quantifier> metacharacters C<?>,
941C<*>, C<+>, and C<{}> were created for. They allow us to delimit the
942number of repeats for a portion of a regexp we consider to be a
47f9c88b
GS
943match. Quantifiers are put immediately after the character, character
944class, or grouping that we want to specify. They have the following
945meanings:
946
947=over 4
948
551e1d92 949=item *
47f9c88b 950
7638d2dc 951C<a?> means: match 'a' 1 or 0 times
47f9c88b 952
551e1d92
RB
953=item *
954
7638d2dc 955C<a*> means: match 'a' 0 or more times, i.e., any number of times
551e1d92
RB
956
957=item *
47f9c88b 958
7638d2dc 959C<a+> means: match 'a' 1 or more times, i.e., at least once
551e1d92
RB
960
961=item *
962
7638d2dc 963C<a{n,m}> means: match at least C<n> times, but not more than C<m>
47f9c88b
GS
964times.
965
551e1d92
RB
966=item *
967
7638d2dc 968C<a{n,}> means: match at least C<n> or more times
551e1d92
RB
969
970=item *
47f9c88b 971
7638d2dc 972C<a{n}> means: match exactly C<n> times
47f9c88b
GS
973
974=back
975
976Here are some examples:
977
7638d2dc 978 /[a-z]+\s+\d*/; # match a lowercase word, at least one space, and
47f9c88b 979 # any number of digits
d8b950dc 980 /(\w+)\s+\g1/; # match doubled words of arbitrary length
47f9c88b
GS
981 /y(es)?/i; # matches 'y', 'Y', or a case-insensitive 'yes'
982 $year =~ /\d{2,4}/; # make sure year is at least 2 but not more
983 # than 4 digits
984 $year =~ /\d{4}|\d{2}/; # better match; throw out 3 digit dates
985 $year =~ /\d{2}(\d{2})?/; # same thing written differently. However,
986 # this produces $1 and the other does not.
987
d8b950dc 988 % simple_grep '^(\w+)\g1$' /usr/dict/words # isn't this easier?
47f9c88b
GS
989 beriberi
990 booboo
991 coco
992 mama
993 murmur
994 papa
995
7638d2dc 996For all of these quantifiers, Perl will try to match as much of the
47f9c88b 997string as possible, while still allowing the regexp to succeed. Thus
7638d2dc
WL
998with C</a?.../>, Perl will first try to match the regexp with the C<a>
999present; if that fails, Perl will try to match the regexp without the
47f9c88b
GS
1000C<a> present. For the quantifier C<*>, we get the following:
1001
1002 $x = "the cat in the hat";
1003 $x =~ /^(.*)(cat)(.*)$/; # matches,
1004 # $1 = 'the '
1005 # $2 = 'cat'
1006 # $3 = ' in the hat'
1007
1008Which is what we might expect, the match finds the only C<cat> in the
1009string and locks onto it. Consider, however, this regexp:
1010
1011 $x =~ /^(.*)(at)(.*)$/; # matches,
1012 # $1 = 'the cat in the h'
1013 # $2 = 'at'
7638d2dc 1014 # $3 = '' (0 characters match)
47f9c88b 1015
7638d2dc 1016One might initially guess that Perl would find the C<at> in C<cat> and
47f9c88b
GS
1017stop there, but that wouldn't give the longest possible string to the
1018first quantifier C<.*>. Instead, the first quantifier C<.*> grabs as
1019much of the string as possible while still having the regexp match. In
a6b2f353 1020this example, that means having the C<at> sequence with the final C<at>
47f9c88b
GS
1021in the string. The other important principle illustrated here is that
1022when there are two or more elements in a regexp, the I<leftmost>
1023quantifier, if there is one, gets to grab as much the string as
1024possible, leaving the rest of the regexp to fight over scraps. Thus in
1025our example, the first quantifier C<.*> grabs most of the string, while
1026the second quantifier C<.*> gets the empty string. Quantifiers that
7638d2dc
WL
1027grab as much of the string as possible are called I<maximal match> or
1028I<greedy> quantifiers.
47f9c88b
GS
1029
1030When a regexp can match a string in several different ways, we can use
1031the principles above to predict which way the regexp will match:
1032
1033=over 4
1034
1035=item *
551e1d92 1036
47f9c88b
GS
1037Principle 0: Taken as a whole, any regexp will be matched at the
1038earliest possible position in the string.
1039
1040=item *
551e1d92 1041
47f9c88b
GS
1042Principle 1: In an alternation C<a|b|c...>, the leftmost alternative
1043that allows a match for the whole regexp will be the one used.
1044
1045=item *
551e1d92 1046
47f9c88b
GS
1047Principle 2: The maximal matching quantifiers C<?>, C<*>, C<+> and
1048C<{n,m}> will in general match as much of the string as possible while
1049still allowing the whole regexp to match.
1050
1051=item *
551e1d92 1052
47f9c88b
GS
1053Principle 3: If there are two or more elements in a regexp, the
1054leftmost greedy quantifier, if any, will match as much of the string
1055as possible while still allowing the whole regexp to match. The next
1056leftmost greedy quantifier, if any, will try to match as much of the
1057string remaining available to it as possible, while still allowing the
1058whole regexp to match. And so on, until all the regexp elements are
1059satisfied.
1060
1061=back
1062
ac036724 1063As we have seen above, Principle 0 overrides the others. The regexp
47f9c88b
GS
1064will be matched as early as possible, with the other principles
1065determining how the regexp matches at that earliest character
1066position.
1067
1068Here is an example of these principles in action:
1069
1070 $x = "The programming republic of Perl";
1071 $x =~ /^(.+)(e|r)(.*)$/; # matches,
1072 # $1 = 'The programming republic of Pe'
1073 # $2 = 'r'
1074 # $3 = 'l'
1075
1076This regexp matches at the earliest string position, C<'T'>. One
1077might think that C<e>, being leftmost in the alternation, would be
1078matched, but C<r> produces the longest string in the first quantifier.
1079
1080 $x =~ /(m{1,2})(.*)$/; # matches,
1081 # $1 = 'mm'
1082 # $2 = 'ing republic of Perl'
1083
1084Here, The earliest possible match is at the first C<'m'> in
1085C<programming>. C<m{1,2}> is the first quantifier, so it gets to match
1086a maximal C<mm>.
1087
1088 $x =~ /.*(m{1,2})(.*)$/; # matches,
1089 # $1 = 'm'
1090 # $2 = 'ing republic of Perl'
1091
1092Here, the regexp matches at the start of the string. The first
1093quantifier C<.*> grabs as much as possible, leaving just a single
1094C<'m'> for the second quantifier C<m{1,2}>.
1095
1096 $x =~ /(.?)(m{1,2})(.*)$/; # matches,
1097 # $1 = 'a'
1098 # $2 = 'mm'
1099 # $3 = 'ing republic of Perl'
1100
1101Here, C<.?> eats its maximal one character at the earliest possible
1102position in the string, C<'a'> in C<programming>, leaving C<m{1,2}>
1103the opportunity to match both C<m>'s. Finally,
1104
1105 "aXXXb" =~ /(X*)/; # matches with $1 = ''
1106
1107because it can match zero copies of C<'X'> at the beginning of the
1108string. If you definitely want to match at least one C<'X'>, use
1109C<X+>, not C<X*>.
1110
1111Sometimes greed is not good. At times, we would like quantifiers to
1112match a I<minimal> piece of string, rather than a maximal piece. For
7638d2dc
WL
1113this purpose, Larry Wall created the I<minimal match> or
1114I<non-greedy> quantifiers C<??>, C<*?>, C<+?>, and C<{}?>. These are
47f9c88b
GS
1115the usual quantifiers with a C<?> appended to them. They have the
1116following meanings:
1117
1118=over 4
1119
551e1d92
RB
1120=item *
1121
7638d2dc 1122C<a??> means: match 'a' 0 or 1 times. Try 0 first, then 1.
47f9c88b 1123
551e1d92
RB
1124=item *
1125
7638d2dc 1126C<a*?> means: match 'a' 0 or more times, i.e., any number of times,
47f9c88b
GS
1127but as few times as possible
1128
551e1d92
RB
1129=item *
1130
7638d2dc 1131C<a+?> means: match 'a' 1 or more times, i.e., at least once, but
47f9c88b
GS
1132as few times as possible
1133
551e1d92
RB
1134=item *
1135
7638d2dc 1136C<a{n,m}?> means: match at least C<n> times, not more than C<m>
47f9c88b
GS
1137times, as few times as possible
1138
551e1d92
RB
1139=item *
1140
7638d2dc 1141C<a{n,}?> means: match at least C<n> times, but as few times as
47f9c88b
GS
1142possible
1143
551e1d92
RB
1144=item *
1145
7638d2dc 1146C<a{n}?> means: match exactly C<n> times. Because we match exactly
47f9c88b
GS
1147C<n> times, C<a{n}?> is equivalent to C<a{n}> and is just there for
1148notational consistency.
1149
1150=back
1151
1152Let's look at the example above, but with minimal quantifiers:
1153
1154 $x = "The programming republic of Perl";
1155 $x =~ /^(.+?)(e|r)(.*)$/; # matches,
1156 # $1 = 'Th'
1157 # $2 = 'e'
1158 # $3 = ' programming republic of Perl'
1159
1160The minimal string that will allow both the start of the string C<^>
1161and the alternation to match is C<Th>, with the alternation C<e|r>
1162matching C<e>. The second quantifier C<.*> is free to gobble up the
1163rest of the string.
1164
1165 $x =~ /(m{1,2}?)(.*?)$/; # matches,
1166 # $1 = 'm'
1167 # $2 = 'ming republic of Perl'
1168
1169The first string position that this regexp can match is at the first
1170C<'m'> in C<programming>. At this position, the minimal C<m{1,2}?>
1171matches just one C<'m'>. Although the second quantifier C<.*?> would
1172prefer to match no characters, it is constrained by the end-of-string
1173anchor C<$> to match the rest of the string.
1174
1175 $x =~ /(.*?)(m{1,2}?)(.*)$/; # matches,
1176 # $1 = 'The progra'
1177 # $2 = 'm'
1178 # $3 = 'ming republic of Perl'
1179
1180In this regexp, you might expect the first minimal quantifier C<.*?>
1181to match the empty string, because it is not constrained by a C<^>
1182anchor to match the beginning of the word. Principle 0 applies here,
1183however. Because it is possible for the whole regexp to match at the
1184start of the string, it I<will> match at the start of the string. Thus
1185the first quantifier has to match everything up to the first C<m>. The
1186second minimal quantifier matches just one C<m> and the third
1187quantifier matches the rest of the string.
1188
1189 $x =~ /(.??)(m{1,2})(.*)$/; # matches,
1190 # $1 = 'a'
1191 # $2 = 'mm'
1192 # $3 = 'ing republic of Perl'
1193
1194Just as in the previous regexp, the first quantifier C<.??> can match
1195earliest at position C<'a'>, so it does. The second quantifier is
1196greedy, so it matches C<mm>, and the third matches the rest of the
1197string.
1198
1199We can modify principle 3 above to take into account non-greedy
1200quantifiers:
1201
1202=over 4
1203
1204=item *
551e1d92 1205
47f9c88b
GS
1206Principle 3: If there are two or more elements in a regexp, the
1207leftmost greedy (non-greedy) quantifier, if any, will match as much
1208(little) of the string as possible while still allowing the whole
1209regexp to match. The next leftmost greedy (non-greedy) quantifier, if
1210any, will try to match as much (little) of the string remaining
1211available to it as possible, while still allowing the whole regexp to
1212match. And so on, until all the regexp elements are satisfied.
1213
1214=back
1215
1216Just like alternation, quantifiers are also susceptible to
1217backtracking. Here is a step-by-step analysis of the example
1218
1219 $x = "the cat in the hat";
1220 $x =~ /^(.*)(at)(.*)$/; # matches,
1221 # $1 = 'the cat in the h'
1222 # $2 = 'at'
1223 # $3 = '' (0 matches)
1224
1225=over 4
1226
551e1d92
RB
1227=item 0
1228
1229Start with the first letter in the string 't'.
47f9c88b 1230
551e1d92
RB
1231=item 1
1232
1233The first quantifier '.*' starts out by matching the whole
47f9c88b
GS
1234string 'the cat in the hat'.
1235
551e1d92
RB
1236=item 2
1237
1238'a' in the regexp element 'at' doesn't match the end of the
47f9c88b
GS
1239string. Backtrack one character.
1240
551e1d92
RB
1241=item 3
1242
1243'a' in the regexp element 'at' still doesn't match the last
47f9c88b
GS
1244letter of the string 't', so backtrack one more character.
1245
551e1d92
RB
1246=item 4
1247
1248Now we can match the 'a' and the 't'.
47f9c88b 1249
551e1d92
RB
1250=item 5
1251
1252Move on to the third element '.*'. Since we are at the end of
47f9c88b
GS
1253the string and '.*' can match 0 times, assign it the empty string.
1254
551e1d92
RB
1255=item 6
1256
1257We are done!
47f9c88b
GS
1258
1259=back
1260
1261Most of the time, all this moving forward and backtracking happens
7638d2dc 1262quickly and searching is fast. There are some pathological regexps,
47f9c88b
GS
1263however, whose execution time exponentially grows with the size of the
1264string. A typical structure that blows up in your face is of the form
1265
1266 /(a|b+)*/;
1267
1268The problem is the nested indeterminate quantifiers. There are many
1269different ways of partitioning a string of length n between the C<+>
1270and C<*>: one repetition with C<b+> of length n, two repetitions with
1271the first C<b+> length k and the second with length n-k, m repetitions
1272whose bits add up to length n, etc. In fact there are an exponential
7638d2dc 1273number of ways to partition a string as a function of its length. A
47f9c88b 1274regexp may get lucky and match early in the process, but if there is
7638d2dc 1275no match, Perl will try I<every> possibility before giving up. So be
47f9c88b 1276careful with nested C<*>'s, C<{n,m}>'s, and C<+>'s. The book
7638d2dc 1277I<Mastering Regular Expressions> by Jeffrey Friedl gives a wonderful
47f9c88b
GS
1278discussion of this and other efficiency issues.
1279
7638d2dc
WL
1280
1281=head2 Possessive quantifiers
1282
1283Backtracking during the relentless search for a match may be a waste
1284of time, particularly when the match is bound to fail. Consider
1285the simple pattern
1286
1287 /^\w+\s+\w+$/; # a word, spaces, a word
1288
1289Whenever this is applied to a string which doesn't quite meet the
1290pattern's expectations such as S<C<"abc ">> or S<C<"abc def ">>,
353c6505
DL
1291the regex engine will backtrack, approximately once for each character
1292in the string. But we know that there is no way around taking I<all>
1293of the initial word characters to match the first repetition, that I<all>
7638d2dc 1294spaces must be eaten by the middle part, and the same goes for the second
353c6505
DL
1295word.
1296
1297With the introduction of the I<possessive quantifiers> in Perl 5.10, we
1298have a way of instructing the regex engine not to backtrack, with the
1299usual quantifiers with a C<+> appended to them. This makes them greedy as
1300well as stingy; once they succeed they won't give anything back to permit
1301another solution. They have the following meanings:
7638d2dc
WL
1302
1303=over 4
1304
1305=item *
1306
353c6505
DL
1307C<a{n,m}+> means: match at least C<n> times, not more than C<m> times,
1308as many times as possible, and don't give anything up. C<a?+> is short
7638d2dc
WL
1309for C<a{0,1}+>
1310
1311=item *
1312
1313C<a{n,}+> means: match at least C<n> times, but as many times as possible,
353c6505 1314and don't give anything up. C<a*+> is short for C<a{0,}+> and C<a++> is
7638d2dc
WL
1315short for C<a{1,}+>.
1316
1317=item *
1318
1319C<a{n}+> means: match exactly C<n> times. It is just there for
1320notational consistency.
1321
1322=back
1323
353c6505
DL
1324These possessive quantifiers represent a special case of a more general
1325concept, the I<independent subexpression>, see below.
7638d2dc
WL
1326
1327As an example where a possessive quantifier is suitable we consider
1328matching a quoted string, as it appears in several programming languages.
1329The backslash is used as an escape character that indicates that the
1330next character is to be taken literally, as another character for the
1331string. Therefore, after the opening quote, we expect a (possibly
353c6505 1332empty) sequence of alternatives: either some character except an
7638d2dc
WL
1333unescaped quote or backslash or an escaped character.
1334
1335 /"(?:[^"\\]++|\\.)*+"/;
1336
1337
47f9c88b
GS
1338=head2 Building a regexp
1339
1340At this point, we have all the basic regexp concepts covered, so let's
1341give a more involved example of a regular expression. We will build a
1342regexp that matches numbers.
1343
1344The first task in building a regexp is to decide what we want to match
1345and what we want to exclude. In our case, we want to match both
1346integers and floating point numbers and we want to reject any string
1347that isn't a number.
1348
1349The next task is to break the problem down into smaller problems that
1350are easily converted into a regexp.
1351
1352The simplest case is integers. These consist of a sequence of digits,
1353with an optional sign in front. The digits we can represent with
1354C<\d+> and the sign can be matched with C<[+-]>. Thus the integer
1355regexp is
1356
1357 /[+-]?\d+/; # matches integers
1358
1359A floating point number potentially has a sign, an integral part, a
1360decimal point, a fractional part, and an exponent. One or more of these
1361parts is optional, so we need to check out the different
1362possibilities. Floating point numbers which are in proper form include
1363123., 0.345, .34, -1e6, and 25.4E-72. As with integers, the sign out
1364front is completely optional and can be matched by C<[+-]?>. We can
1365see that if there is no exponent, floating point numbers must have a
1366decimal point, otherwise they are integers. We might be tempted to
1367model these with C<\d*\.\d*>, but this would also match just a single
1368decimal point, which is not a number. So the three cases of floating
7638d2dc 1369point number without exponent are
47f9c88b
GS
1370
1371 /[+-]?\d+\./; # 1., 321., etc.
1372 /[+-]?\.\d+/; # .1, .234, etc.
1373 /[+-]?\d+\.\d+/; # 1.0, 30.56, etc.
1374
1375These can be combined into a single regexp with a three-way alternation:
1376
1377 /[+-]?(\d+\.\d+|\d+\.|\.\d+)/; # floating point, no exponent
1378
1379In this alternation, it is important to put C<'\d+\.\d+'> before
1380C<'\d+\.'>. If C<'\d+\.'> were first, the regexp would happily match that
1381and ignore the fractional part of the number.
1382
1383Now consider floating point numbers with exponents. The key
1384observation here is that I<both> integers and numbers with decimal
1385points are allowed in front of an exponent. Then exponents, like the
1386overall sign, are independent of whether we are matching numbers with
1387or without decimal points, and can be 'decoupled' from the
1388mantissa. The overall form of the regexp now becomes clear:
1389
1390 /^(optional sign)(integer | f.p. mantissa)(optional exponent)$/;
1391
1392The exponent is an C<e> or C<E>, followed by an integer. So the
1393exponent regexp is
1394
1395 /[eE][+-]?\d+/; # exponent
1396
1397Putting all the parts together, we get a regexp that matches numbers:
1398
1399 /^[+-]?(\d+\.\d+|\d+\.|\.\d+|\d+)([eE][+-]?\d+)?$/; # Ta da!
1400
1401Long regexps like this may impress your friends, but can be hard to
1402decipher. In complex situations like this, the C<//x> modifier for a
1403match is invaluable. It allows one to put nearly arbitrary whitespace
1404and comments into a regexp without affecting their meaning. Using it,
1405we can rewrite our 'extended' regexp in the more pleasing form
1406
1407 /^
1408 [+-]? # first, match an optional sign
1409 ( # then match integers or f.p. mantissas:
1410 \d+\.\d+ # mantissa of the form a.b
1411 |\d+\. # mantissa of the form a.
1412 |\.\d+ # mantissa of the form .b
1413 |\d+ # integer of the form a
1414 )
1415 ([eE][+-]?\d+)? # finally, optionally match an exponent
1416 $/x;
1417
1418If whitespace is mostly irrelevant, how does one include space
1419characters in an extended regexp? The answer is to backslash it
7638d2dc 1420S<C<'\ '>> or put it in a character class S<C<[ ]>>. The same thing
47f9c88b 1421goes for pound signs, use C<\#> or C<[#]>. For instance, Perl allows
7638d2dc 1422a space between the sign and the mantissa or integer, and we could add
47f9c88b
GS
1423this to our regexp as follows:
1424
1425 /^
1426 [+-]?\ * # first, match an optional sign *and space*
1427 ( # then match integers or f.p. mantissas:
1428 \d+\.\d+ # mantissa of the form a.b
1429 |\d+\. # mantissa of the form a.
1430 |\.\d+ # mantissa of the form .b
1431 |\d+ # integer of the form a
1432 )
1433 ([eE][+-]?\d+)? # finally, optionally match an exponent
1434 $/x;
1435
1436In this form, it is easier to see a way to simplify the
1437alternation. Alternatives 1, 2, and 4 all start with C<\d+>, so it
1438could be factored out:
1439
1440 /^
1441 [+-]?\ * # first, match an optional sign
1442 ( # then match integers or f.p. mantissas:
1443 \d+ # start out with a ...
1444 (
1445 \.\d* # mantissa of the form a.b or a.
1446 )? # ? takes care of integers of the form a
1447 |\.\d+ # mantissa of the form .b
1448 )
1449 ([eE][+-]?\d+)? # finally, optionally match an exponent
1450 $/x;
1451
1452or written in the compact form,
1453
1454 /^[+-]?\ *(\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?$/;
1455
1456This is our final regexp. To recap, we built a regexp by
1457
1458=over 4
1459
551e1d92
RB
1460=item *
1461
1462specifying the task in detail,
47f9c88b 1463
551e1d92
RB
1464=item *
1465
1466breaking down the problem into smaller parts,
1467
1468=item *
47f9c88b 1469
551e1d92 1470translating the small parts into regexps,
47f9c88b 1471
551e1d92
RB
1472=item *
1473
1474combining the regexps,
1475
1476=item *
47f9c88b 1477
551e1d92 1478and optimizing the final combined regexp.
47f9c88b
GS
1479
1480=back
1481
1482These are also the typical steps involved in writing a computer
1483program. This makes perfect sense, because regular expressions are
7638d2dc 1484essentially programs written in a little computer language that specifies
47f9c88b
GS
1485patterns.
1486
1487=head2 Using regular expressions in Perl
1488
1489The last topic of Part 1 briefly covers how regexps are used in Perl
1490programs. Where do they fit into Perl syntax?
1491
1492We have already introduced the matching operator in its default
1493C</regexp/> and arbitrary delimiter C<m!regexp!> forms. We have used
1494the binding operator C<=~> and its negation C<!~> to test for string
1495matches. Associated with the matching operator, we have discussed the
1496single line C<//s>, multi-line C<//m>, case-insensitive C<//i> and
353c6505
DL
1497extended C<//x> modifiers. There are a few more things you might
1498want to know about matching operators.
47f9c88b 1499
7638d2dc
WL
1500=head3 Optimizing pattern evaluation
1501
353c6505 1502We pointed out earlier that variables in regexps are substituted
7638d2dc 1503before the regexp is evaluated:
47f9c88b
GS
1504
1505 $pattern = 'Seuss';
1506 while (<>) {
1507 print if /$pattern/;
1508 }
1509
1510This will print any lines containing the word C<Seuss>. It is not as
7638d2dc
WL
1511efficient as it could be, however, because Perl has to re-evaluate
1512(or compile) C<$pattern> each time through the loop. If C<$pattern> won't be
47f9c88b 1513changing over the lifetime of the script, we can add the C<//o>
7638d2dc 1514modifier, which directs Perl to only perform variable substitutions
47f9c88b
GS
1515once:
1516
1517 #!/usr/bin/perl
1518 # Improved simple_grep
1519 $regexp = shift;
1520 while (<>) {
1521 print if /$regexp/o; # a good deal faster
1522 }
1523
7638d2dc
WL
1524
1525=head3 Prohibiting substitution
1526
1527If you change C<$pattern> after the first substitution happens, Perl
47f9c88b
GS
1528will ignore it. If you don't want any substitutions at all, use the
1529special delimiter C<m''>:
1530
16e8b840 1531 @pattern = ('Seuss');
47f9c88b 1532 while (<>) {
16e8b840 1533 print if m'@pattern'; # matches literal '@pattern', not 'Seuss'
47f9c88b
GS
1534 }
1535
353c6505 1536Similar to strings, C<m''> acts like apostrophes on a regexp; all other
7638d2dc 1537C<m> delimiters act like quotes. If the regexp evaluates to the empty string,
47f9c88b
GS
1538the regexp in the I<last successful match> is used instead. So we have
1539
1540 "dog" =~ /d/; # 'd' matches
1541 "dogbert =~ //; # this matches the 'd' regexp used before
1542
7638d2dc
WL
1543
1544=head3 Global matching
1545
47f9c88b 1546The final two modifiers C<//g> and C<//c> concern multiple matches.
da75cd15 1547The modifier C<//g> stands for global matching and allows the
47f9c88b
GS
1548matching operator to match within a string as many times as possible.
1549In scalar context, successive invocations against a string will have
1550`C<//g> jump from match to match, keeping track of position in the
1551string as it goes along. You can get or set the position with the
1552C<pos()> function.
1553
1554The use of C<//g> is shown in the following example. Suppose we have
1555a string that consists of words separated by spaces. If we know how
1556many words there are in advance, we could extract the words using
1557groupings:
1558
1559 $x = "cat dog house"; # 3 words
1560 $x =~ /^\s*(\w+)\s+(\w+)\s+(\w+)\s*$/; # matches,
1561 # $1 = 'cat'
1562 # $2 = 'dog'
1563 # $3 = 'house'
1564
1565But what if we had an indeterminate number of words? This is the sort
1566of task C<//g> was made for. To extract all words, form the simple
1567regexp C<(\w+)> and loop over all matches with C</(\w+)/g>:
1568
1569 while ($x =~ /(\w+)/g) {
1570 print "Word is $1, ends at position ", pos $x, "\n";
1571 }
1572
1573prints
1574
1575 Word is cat, ends at position 3
1576 Word is dog, ends at position 7
1577 Word is house, ends at position 13
1578
1579A failed match or changing the target string resets the position. If
1580you don't want the position reset after failure to match, add the
1581C<//c>, as in C</regexp/gc>. The current position in the string is
1582associated with the string, not the regexp. This means that different
1583strings have different positions and their respective positions can be
1584set or read independently.
1585
1586In list context, C<//g> returns a list of matched groupings, or if
1587there are no groupings, a list of matches to the whole regexp. So if
1588we wanted just the words, we could use
1589
1590 @words = ($x =~ /(\w+)/g); # matches,
1591 # $word[0] = 'cat'
1592 # $word[1] = 'dog'
1593 # $word[2] = 'house'
1594
1595Closely associated with the C<//g> modifier is the C<\G> anchor. The
1596C<\G> anchor matches at the point where the previous C<//g> match left
1597off. C<\G> allows us to easily do context-sensitive matching:
1598
1599 $metric = 1; # use metric units
1600 ...
1601 $x = <FILE>; # read in measurement
1602 $x =~ /^([+-]?\d+)\s*/g; # get magnitude
1603 $weight = $1;
1604 if ($metric) { # error checking
1605 print "Units error!" unless $x =~ /\Gkg\./g;
1606 }
1607 else {
1608 print "Units error!" unless $x =~ /\Glbs\./g;
1609 }
1610 $x =~ /\G\s+(widget|sprocket)/g; # continue processing
1611
1612The combination of C<//g> and C<\G> allows us to process the string a
1613bit at a time and use arbitrary Perl logic to decide what to do next.
25cf8c22
HS
1614Currently, the C<\G> anchor is only fully supported when used to anchor
1615to the start of the pattern.
47f9c88b
GS
1616
1617C<\G> is also invaluable in processing fixed length records with
1618regexps. Suppose we have a snippet of coding region DNA, encoded as
1619base pair letters C<ATCGTTGAAT...> and we want to find all the stop
1620codons C<TGA>. In a coding region, codons are 3-letter sequences, so
1621we can think of the DNA snippet as a sequence of 3-letter records. The
1622naive regexp
1623
1624 # expanded, this is "ATC GTT GAA TGC AAA TGA CAT GAC"
1625 $dna = "ATCGTTGAATGCAAATGACATGAC";
1626 $dna =~ /TGA/;
1627
d1be9408 1628doesn't work; it may match a C<TGA>, but there is no guarantee that
47f9c88b 1629the match is aligned with codon boundaries, e.g., the substring
7638d2dc 1630S<C<GTT GAA>> gives a match. A better solution is
47f9c88b
GS
1631
1632 while ($dna =~ /(\w\w\w)*?TGA/g) { # note the minimal *?
1633 print "Got a TGA stop codon at position ", pos $dna, "\n";
1634 }
1635
1636which prints
1637
1638 Got a TGA stop codon at position 18
1639 Got a TGA stop codon at position 23
1640
1641Position 18 is good, but position 23 is bogus. What happened?
1642
1643The answer is that our regexp works well until we get past the last
1644real match. Then the regexp will fail to match a synchronized C<TGA>
1645and start stepping ahead one character position at a time, not what we
1646want. The solution is to use C<\G> to anchor the match to the codon
1647alignment:
1648
1649 while ($dna =~ /\G(\w\w\w)*?TGA/g) {
1650 print "Got a TGA stop codon at position ", pos $dna, "\n";
1651 }
1652
1653This prints
1654
1655 Got a TGA stop codon at position 18
1656
1657which is the correct answer. This example illustrates that it is
1658important not only to match what is desired, but to reject what is not
1659desired.
1660
7638d2dc 1661=head3 Search and replace
47f9c88b 1662
7638d2dc 1663Regular expressions also play a big role in I<search and replace>
47f9c88b
GS
1664operations in Perl. Search and replace is accomplished with the
1665C<s///> operator. The general form is
1666C<s/regexp/replacement/modifiers>, with everything we know about
1667regexps and modifiers applying in this case as well. The
1668C<replacement> is a Perl double quoted string that replaces in the
1669string whatever is matched with the C<regexp>. The operator C<=~> is
1670also used here to associate a string with C<s///>. If matching
7638d2dc 1671against C<$_>, the S<C<$_ =~>> can be dropped. If there is a match,
47f9c88b
GS
1672C<s///> returns the number of substitutions made, otherwise it returns
1673false. Here are a few examples:
1674
1675 $x = "Time to feed the cat!";
1676 $x =~ s/cat/hacker/; # $x contains "Time to feed the hacker!"
1677 if ($x =~ s/^(Time.*hacker)!$/$1 now!/) {
1678 $more_insistent = 1;
1679 }
1680 $y = "'quoted words'";
1681 $y =~ s/^'(.*)'$/$1/; # strip single quotes,
1682 # $y contains "quoted words"
1683
1684In the last example, the whole string was matched, but only the part
1685inside the single quotes was grouped. With the C<s///> operator, the
1686matched variables C<$1>, C<$2>, etc. are immediately available for use
1687in the replacement expression, so we use C<$1> to replace the quoted
1688string with just what was quoted. With the global modifier, C<s///g>
1689will search and replace all occurrences of the regexp in the string:
1690
1691 $x = "I batted 4 for 4";
1692 $x =~ s/4/four/; # doesn't do it all:
1693 # $x contains "I batted four for 4"
1694 $x = "I batted 4 for 4";
1695 $x =~ s/4/four/g; # does it all:
1696 # $x contains "I batted four for four"
1697
1698If you prefer 'regex' over 'regexp' in this tutorial, you could use
1699the following program to replace it:
1700
1701 % cat > simple_replace
1702 #!/usr/bin/perl
1703 $regexp = shift;
1704 $replacement = shift;
1705 while (<>) {
1706 s/$regexp/$replacement/go;
1707 print;
1708 }
1709 ^D
1710
1711 % simple_replace regexp regex perlretut.pod
1712
1713In C<simple_replace> we used the C<s///g> modifier to replace all
1714occurrences of the regexp on each line and the C<s///o> modifier to
1715compile the regexp only once. As with C<simple_grep>, both the
1716C<print> and the C<s/$regexp/$replacement/go> use C<$_> implicitly.
1717
4f4d7508
DC
1718If you don't want C<s///> to change your original variable you can use
1719the non-destructive substitute modifier, C<s///r>. This changes the
1720behavior so that C<s///r> returns the final substituted string:
1721
1722 $x = "I like dogs.";
1723 $y = $x =~ s/dogs/cats/r;
1724 print "$x $y\n";
1725
1726That example will print "I like dogs. I like cats". Notice the original
1727C<$x> variable has not been affected by the substitute. The overall
1728result of the substitution is instead stored in C<$y>. If the
1729substitution doesn't affect anything then the original string is
1730returned:
1731
1732 $x = "I like dogs.";
1733 $y = $x =~ s/elephants/cougars/r;
1734 print "$x $y\n"; # prints "I like dogs. I like dogs."
1735
1736One other interesting thing that the C<s///r> flag allows is chaining
1737substitutions:
1738
1739 $x = "Cats are great.";
1740 print $x =~ s/Cats/Dogs/r =~ s/Dogs/Frogs/r =~ s/Frogs/Hedgehogs/r, "\n";
1741 # prints "Hedgehogs are great."
1742
47f9c88b
GS
1743A modifier available specifically to search and replace is the
1744C<s///e> evaluation modifier. C<s///e> wraps an C<eval{...}> around
1745the replacement string and the evaluated result is substituted for the
1746matched substring. C<s///e> is useful if you need to do a bit of
1747computation in the process of replacing text. This example counts
1748character frequencies in a line:
1749
1750 $x = "Bill the cat";
1751 $x =~ s/(.)/$chars{$1}++;$1/eg; # final $1 replaces char with itself
1752 print "frequency of '$_' is $chars{$_}\n"
1753 foreach (sort {$chars{$b} <=> $chars{$a}} keys %chars);
1754
1755This prints
1756
1757 frequency of ' ' is 2
1758 frequency of 't' is 2
1759 frequency of 'l' is 2
1760 frequency of 'B' is 1
1761 frequency of 'c' is 1
1762 frequency of 'e' is 1
1763 frequency of 'h' is 1
1764 frequency of 'i' is 1
1765 frequency of 'a' is 1
1766
1767As with the match C<m//> operator, C<s///> can use other delimiters,
1768such as C<s!!!> and C<s{}{}>, and even C<s{}//>. If single quotes are
1769used C<s'''>, then the regexp and replacement are treated as single
1770quoted strings and there are no substitutions. C<s///> in list context
1771returns the same thing as in scalar context, i.e., the number of
1772matches.
1773
7638d2dc 1774=head3 The split function
47f9c88b 1775
7638d2dc 1776The C<split()> function is another place where a regexp is used.
353c6505
DL
1777C<split /regexp/, string, limit> separates the C<string> operand into
1778a list of substrings and returns that list. The regexp must be designed
7638d2dc 1779to match whatever constitutes the separators for the desired substrings.
353c6505 1780The C<limit>, if present, constrains splitting into no more than C<limit>
7638d2dc 1781number of strings. For example, to split a string into words, use
47f9c88b
GS
1782
1783 $x = "Calvin and Hobbes";
1784 @words = split /\s+/, $x; # $word[0] = 'Calvin'
1785 # $word[1] = 'and'
1786 # $word[2] = 'Hobbes'
1787
1788If the empty regexp C<//> is used, the regexp always matches and
1789the string is split into individual characters. If the regexp has
7638d2dc 1790groupings, then the resulting list contains the matched substrings from the
47f9c88b
GS
1791groupings as well. For instance,
1792
1793 $x = "/usr/bin/perl";
1794 @dirs = split m!/!, $x; # $dirs[0] = ''
1795 # $dirs[1] = 'usr'
1796 # $dirs[2] = 'bin'
1797 # $dirs[3] = 'perl'
1798 @parts = split m!(/)!, $x; # $parts[0] = ''
1799 # $parts[1] = '/'
1800 # $parts[2] = 'usr'
1801 # $parts[3] = '/'
1802 # $parts[4] = 'bin'
1803 # $parts[5] = '/'
1804 # $parts[6] = 'perl'
1805
1806Since the first character of $x matched the regexp, C<split> prepended
1807an empty initial element to the list.
1808
1809If you have read this far, congratulations! You now have all the basic
1810tools needed to use regular expressions to solve a wide range of text
1811processing problems. If this is your first time through the tutorial,
1812why not stop here and play around with regexps a while... S<Part 2>
1813concerns the more esoteric aspects of regular expressions and those
1814concepts certainly aren't needed right at the start.
1815
1816=head1 Part 2: Power tools
1817
1818OK, you know the basics of regexps and you want to know more. If
1819matching regular expressions is analogous to a walk in the woods, then
1820the tools discussed in Part 1 are analogous to topo maps and a
1821compass, basic tools we use all the time. Most of the tools in part 2
da75cd15 1822are analogous to flare guns and satellite phones. They aren't used
47f9c88b
GS
1823too often on a hike, but when we are stuck, they can be invaluable.
1824
1825What follows are the more advanced, less used, or sometimes esoteric
7638d2dc 1826capabilities of Perl regexps. In Part 2, we will assume you are
47f9c88b
GS
1827comfortable with the basics and concentrate on the new features.
1828
1829=head2 More on characters, strings, and character classes
1830
1831There are a number of escape sequences and character classes that we
1832haven't covered yet.
1833
1834There are several escape sequences that convert characters or strings
7638d2dc 1835between upper and lower case, and they are also available within
353c6505 1836patterns. C<\l> and C<\u> convert the next character to lower or
7638d2dc 1837upper case, respectively:
47f9c88b
GS
1838
1839 $x = "perl";
1840 $string =~ /\u$x/; # matches 'Perl' in $string
1841 $x = "M(rs?|s)\\."; # note the double backslash
1842 $string =~ /\l$x/; # matches 'mr.', 'mrs.', and 'ms.',
1843
7638d2dc
WL
1844A C<\L> or C<\U> indicates a lasting conversion of case, until
1845terminated by C<\E> or thrown over by another C<\U> or C<\L>:
47f9c88b
GS
1846
1847 $x = "This word is in lower case:\L SHOUT\E";
1848 $x =~ /shout/; # matches
1849 $x = "I STILL KEYPUNCH CARDS FOR MY 360"
1850 $x =~ /\Ukeypunch/; # matches punch card string
1851
1852If there is no C<\E>, case is converted until the end of the
1853string. The regexps C<\L\u$word> or C<\u\L$word> convert the first
1854character of C<$word> to uppercase and the rest of the characters to
1855lowercase.
1856
1857Control characters can be escaped with C<\c>, so that a control-Z
1858character would be matched with C<\cZ>. The escape sequence
1859C<\Q>...C<\E> quotes, or protects most non-alphabetic characters. For
1860instance,
1861
1862 $x = "\QThat !^*&%~& cat!";
1863 $x =~ /\Q!^*&%~&\E/; # check for rough language
1864
1865It does not protect C<$> or C<@>, so that variables can still be
1866substituted.
1867
7638d2dc
WL
1868With the advent of 5.6.0, Perl regexps can handle more than just the
1869standard ASCII character set. Perl now supports I<Unicode>, a standard
1870for representing the alphabets from virtually all of the world's written
38a44b82 1871languages, and a host of symbols. Perl's text strings are Unicode strings, so
2575c402
JW
1872they can contain characters with a value (codepoint or character number) higher
1873than 255
47f9c88b
GS
1874
1875What does this mean for regexps? Well, regexp users don't need to know
7638d2dc 1876much about Perl's internal representation of strings. But they do need
2575c402
JW
1877to know 1) how to represent Unicode characters in a regexp and 2) that
1878a matching operation will treat the string to be searched as a sequence
1879of characters, not bytes. The answer to 1) is that Unicode characters
f0a2b745
KW
1880greater than C<chr(255)> are represented using the C<\x{hex}> notation, because
1881\x hex (without curly braces) doesn't go further than 255. (Starting in Perl
18825.14) if you're an octal fan, you can also use C<\o{oct}>.
47f9c88b 1883
47f9c88b
GS
1884 /\x{263a}/; # match a Unicode smiley face :)
1885
7638d2dc 1886B<NOTE>: In Perl 5.6.0 it used to be that one needed to say C<use
72ff2908
JH
1887utf8> to use any Unicode features. This is no more the case: for
1888almost all Unicode processing, the explicit C<utf8> pragma is not
1889needed. (The only case where it matters is if your Perl script is in
1890Unicode and encoded in UTF-8, then an explicit C<use utf8> is needed.)
47f9c88b
GS
1891
1892Figuring out the hexadecimal sequence of a Unicode character you want
1893or deciphering someone else's hexadecimal Unicode regexp is about as
1894much fun as programming in machine code. So another way to specify
e526e8bb
KW
1895Unicode characters is to use the I<named character> escape
1896sequence C<\N{I<name>}>. I<name> is a name for the Unicode character, as
55eda711
JH
1897specified in the Unicode standard. For instance, if we wanted to
1898represent or match the astrological sign for the planet Mercury, we
1899could use
47f9c88b 1900
47f9c88b
GS
1901 use charnames ":full"; # use named chars with Unicode full names
1902 $x = "abc\N{MERCURY}def";
1903 $x =~ /\N{MERCURY}/; # matches
1904
1905One can also use short names or restrict names to a certain alphabet:
1906
47f9c88b
GS
1907 use charnames ':full';
1908 print "\N{GREEK SMALL LETTER SIGMA} is called sigma.\n";
1909
1910 use charnames ":short";
1911 print "\N{greek:Sigma} is an upper-case sigma.\n";
1912
1913 use charnames qw(greek);
1914 print "\N{sigma} is Greek sigma\n";
1915
7638d2dc
WL
1916A list of full names is found in the file NamesList.txt in the
1917lib/perl5/X.X.X/unicore directory (where X.X.X is the perl
1918version number as it is installed on your system).
47f9c88b 1919
38a44b82 1920The answer to requirement 2), as of 5.6.0, is that a regexp uses Unicode
2575c402
JW
1921characters. Internally, this is encoded to bytes using either UTF-8 or a
1922native 8 bit encoding, depending on the history of the string, but
1923conceptually it is a sequence of characters, not bytes. See
1924L<perlunitut> for a tutorial about that.
1925
1926Let us now discuss Unicode character classes. Just as with Unicode
1927characters, there are named Unicode character classes represented by the
1928C<\p{name}> escape sequence. Closely associated is the C<\P{name}>
1929character class, which is the negation of the C<\p{name}> class. For
1930example, to match lower and uppercase characters,
47f9c88b 1931
47f9c88b
GS
1932 use charnames ":full"; # use named chars with Unicode full names
1933 $x = "BOB";
1934 $x =~ /^\p{IsUpper}/; # matches, uppercase char class
1935 $x =~ /^\P{IsUpper}/; # doesn't match, char class sans uppercase
1936 $x =~ /^\p{IsLower}/; # doesn't match, lowercase char class
1937 $x =~ /^\P{IsLower}/; # matches, char class sans lowercase
1938
86929931
JH
1939Here is the association between some Perl named classes and the
1940traditional Unicode classes:
47f9c88b 1941
86929931 1942 Perl class name Unicode class name or regular expression
47f9c88b 1943
f5868911
JH
1944 IsAlpha /^[LM]/
1945 IsAlnum /^[LMN]/
1946 IsASCII $code <= 127
1947 IsCntrl /^C/
1948 IsBlank $code =~ /^(0020|0009)$/ || /^Z[^lp]/
47f9c88b 1949 IsDigit Nd
f5868911 1950 IsGraph /^([LMNPS]|Co)/
47f9c88b 1951 IsLower Ll
f5868911
JH
1952 IsPrint /^([LMNPS]|Co|Zs)/
1953 IsPunct /^P/
1954 IsSpace /^Z/ || ($code =~ /^(0009|000A|000B|000C|000D)$/
08ce8fc6 1955 IsSpacePerl /^Z/ || ($code =~ /^(0009|000A|000C|000D|0085|2028|2029)$/
f5868911
JH
1956 IsUpper /^L[ut]/
1957 IsWord /^[LMN]/ || $code eq "005F"
47f9c88b
GS
1958 IsXDigit $code =~ /^00(3[0-9]|[46][1-6])$/
1959
86929931
JH
1960You can also use the official Unicode class names with the C<\p> and
1961C<\P>, like C<\p{L}> for Unicode 'letters', or C<\p{Lu}> for uppercase
1962letters, or C<\P{Nd}> for non-digits. If a C<name> is just one
1963letter, the braces can be dropped. For instance, C<\pM> is the
98f22ffc 1964character class of Unicode 'marks', for example accent marks.
32293815
JH
1965For the full list see L<perlunicode>.
1966
fa11829f 1967The Unicode has also been separated into various sets of characters
7638d2dc
WL
1968which you can test with C<\p{...}> (in) and C<\P{...}> (not in).
1969To test whether a character is (or is not) an element of a script
353c6505 1970you would use the script name, for example C<\p{Latin}>, C<\p{Greek}>,
7638d2dc
WL
1971or C<\P{Katakana}>. Other sets are the Unicode blocks, the names
1972of which begin with "In". One such block is dedicated to mathematical
1973operators, and its pattern formula is <C\p{InMathematicalOperators>}>.
e1b711da
KW
1974For the full list see L<perluniprops>.
1975
1976What we have described so far is the single form of the C<\p{...}> character
1977classes. There is also a compound form which you may run into. These
1978look like C<\p{name=value}> or C<\p{name:value}> (the equals sign and colon
1979can be used interchangeably). These are more general than the single form,
1980and in fact most of the single forms are just Perl-defined shortcuts for common
1981compound forms. For example, the script examples in the previous paragraph
1982could be written equivalently as C<\p{Script=Latin}>, C<\p{Script:Greek}>, and
1983C<\P{script=katakana}> (case is irrelevant between the C<{}> braces). You may
1984never have to use the compound forms, but sometimes it is necessary, and their
1985use can make your code easier to understand.
47f9c88b 1986
7638d2dc 1987C<\X> is an abbreviation for a character class that comprises
e1b711da
KW
1988a Unicode I<extended grapheme cluster>. This represents a "logical character",
1989what appears to be a single character, but may be represented internally by more
1990than one. As an example, using the Unicode full names, e.g., S<C<A + COMBINING
1991RING>> is a grapheme cluster with base character C<A> and combining character
1992S<C<COMBINING RING>>, which translates in Danish to A with the circle atop it,
1993as in the word Angstrom.
47f9c88b 1994
da75cd15 1995For the full and latest information about Unicode see the latest
e1b711da 1996Unicode standard, or the Unicode Consortium's website L<http://www.unicode.org>
5e42d7b4 1997
47f9c88b
GS
1998As if all those classes weren't enough, Perl also defines POSIX style
1999character classes. These have the form C<[:name:]>, with C<name> the
aaa51d5e
JF
2000name of the POSIX class. The POSIX classes are C<alpha>, C<alnum>,
2001C<ascii>, C<cntrl>, C<digit>, C<graph>, C<lower>, C<print>, C<punct>,
2002C<space>, C<upper>, and C<xdigit>, and two extensions, C<word> (a Perl
2003extension to match C<\w>), and C<blank> (a GNU extension). If C<utf8>
2004is being used, then these classes are defined the same as their
7638d2dc 2005corresponding Perl Unicode classes: C<[:upper:]> is the same as
aaa51d5e
JF
2006C<\p{IsUpper}>, etc. The POSIX character classes, however, don't
2007require using C<utf8>. The C<[:digit:]>, C<[:word:]>, and
47f9c88b 2008C<[:space:]> correspond to the familiar C<\d>, C<\w>, and C<\s>
aaa51d5e
JF
2009character classes. To negate a POSIX class, put a C<^> in front of
2010the name, so that, e.g., C<[:^digit:]> corresponds to C<\D> and under
47f9c88b 2011C<utf8>, C<\P{IsDigit}>. The Unicode and POSIX character classes can
54c18d04
MK
2012be used just like C<\d>, with the exception that POSIX character
2013classes can only be used inside of a character class:
47f9c88b
GS
2014
2015 /\s+[abc[:digit:]xyz]\s*/; # match a,b,c,x,y,z, or a digit
54c18d04 2016 /^=item\s[[:digit:]]/; # match '=item',
47f9c88b 2017 # followed by a space and a digit
47f9c88b
GS
2018 use charnames ":full";
2019 /\s+[abc\p{IsDigit}xyz]\s+/; # match a,b,c,x,y,z, or a digit
2020 /^=item\s\p{IsDigit}/; # match '=item',
2021 # followed by a space and a digit
2022
2023Whew! That is all the rest of the characters and character classes.
2024
2025=head2 Compiling and saving regular expressions
2026
2027In Part 1 we discussed the C<//o> modifier, which compiles a regexp
2028just once. This suggests that a compiled regexp is some data structure
2029that can be stored once and used again and again. The regexp quote
2030C<qr//> does exactly that: C<qr/string/> compiles the C<string> as a
2031regexp and transforms the result into a form that can be assigned to a
2032variable:
2033
2034 $reg = qr/foo+bar?/; # reg contains a compiled regexp
2035
2036Then C<$reg> can be used as a regexp:
2037
2038 $x = "fooooba";
2039 $x =~ $reg; # matches, just like /foo+bar?/
2040 $x =~ /$reg/; # same thing, alternate form
2041
2042C<$reg> can also be interpolated into a larger regexp:
2043
2044 $x =~ /(abc)?$reg/; # still matches
2045
2046As with the matching operator, the regexp quote can use different
7638d2dc
WL
2047delimiters, e.g., C<qr!!>, C<qr{}> or C<qr~~>. Apostrophes
2048as delimiters (C<qr''>) inhibit any interpolation.
47f9c88b
GS
2049
2050Pre-compiled regexps are useful for creating dynamic matches that
2051don't need to be recompiled each time they are encountered. Using
7638d2dc
WL
2052pre-compiled regexps, we write a C<grep_step> program which greps
2053for a sequence of patterns, advancing to the next pattern as soon
2054as one has been satisfied.
47f9c88b 2055
7638d2dc 2056 % cat > grep_step
47f9c88b 2057 #!/usr/bin/perl
7638d2dc 2058 # grep_step - match <number> regexps, one after the other
47f9c88b
GS
2059 # usage: multi_grep <number> regexp1 regexp2 ... file1 file2 ...
2060
2061 $number = shift;
2062 $regexp[$_] = shift foreach (0..$number-1);
2063 @compiled = map qr/$_/, @regexp;
2064 while ($line = <>) {
7638d2dc
WL
2065 if ($line =~ /$compiled[0]/) {
2066 print $line;
2067 shift @compiled;
2068 last unless @compiled;
47f9c88b
GS
2069 }
2070 }
2071 ^D
2072
7638d2dc
WL
2073 % grep_step 3 shift print last grep_step
2074 $number = shift;
2075 print $line;
2076 last unless @compiled;
47f9c88b
GS
2077
2078Storing pre-compiled regexps in an array C<@compiled> allows us to
2079simply loop through the regexps without any recompilation, thus gaining
2080flexibility without sacrificing speed.
2081
7638d2dc
WL
2082
2083=head2 Composing regular expressions at runtime
2084
2085Backtracking is more efficient than repeated tries with different regular
2086expressions. If there are several regular expressions and a match with
353c6505 2087any of them is acceptable, then it is possible to combine them into a set
7638d2dc 2088of alternatives. If the individual expressions are input data, this
353c6505
DL
2089can be done by programming a join operation. We'll exploit this idea in
2090an improved version of the C<simple_grep> program: a program that matches
7638d2dc
WL
2091multiple patterns:
2092
2093 % cat > multi_grep
2094 #!/usr/bin/perl
2095 # multi_grep - match any of <number> regexps
2096 # usage: multi_grep <number> regexp1 regexp2 ... file1 file2 ...
2097
2098 $number = shift;
2099 $regexp[$_] = shift foreach (0..$number-1);
2100 $pattern = join '|', @regexp;
2101
2102 while ($line = <>) {
2103 print $line if $line =~ /$pattern/o;
2104 }
2105 ^D
2106
2107 % multi_grep 2 shift for multi_grep
2108 $number = shift;
2109 $regexp[$_] = shift foreach (0..$number-1);
2110
2111Sometimes it is advantageous to construct a pattern from the I<input>
2112that is to be analyzed and use the permissible values on the left
2113hand side of the matching operations. As an example for this somewhat
353c6505 2114paradoxical situation, let's assume that our input contains a command
7638d2dc 2115verb which should match one out of a set of available command verbs,
353c6505 2116with the additional twist that commands may be abbreviated as long as
7638d2dc
WL
2117the given string is unique. The program below demonstrates the basic
2118algorithm.
2119
2120 % cat > keymatch
2121 #!/usr/bin/perl
2122 $kwds = 'copy compare list print';
2123 while( $command = <> ){
2124 $command =~ s/^\s+|\s+$//g; # trim leading and trailing spaces
2125 if( ( @matches = $kwds =~ /\b$command\w*/g ) == 1 ){
92a24ac3 2126 print "command: '@matches'\n";
7638d2dc
WL
2127 } elsif( @matches == 0 ){
2128 print "no such command: '$command'\n";
2129 } else {
2130 print "not unique: '$command' (could be one of: @matches)\n";
2131 }
2132 }
2133 ^D
2134
2135 % keymatch
2136 li
2137 command: 'list'
2138 co
2139 not unique: 'co' (could be one of: copy compare)
2140 printer
2141 no such command: 'printer'
2142
2143Rather than trying to match the input against the keywords, we match the
2144combined set of keywords against the input. The pattern matching
353c6505
DL
2145operation S<C<$kwds =~ /\b($command\w*)/g>> does several things at the
2146same time. It makes sure that the given command begins where a keyword
2147begins (C<\b>). It tolerates abbreviations due to the added C<\w*>. It
2148tells us the number of matches (C<scalar @matches>) and all the keywords
7638d2dc 2149that were actually matched. You could hardly ask for more.
7638d2dc 2150
47f9c88b
GS
2151=head2 Embedding comments and modifiers in a regular expression
2152
2153Starting with this section, we will be discussing Perl's set of
7638d2dc 2154I<extended patterns>. These are extensions to the traditional regular
47f9c88b
GS
2155expression syntax that provide powerful new tools for pattern
2156matching. We have already seen extensions in the form of the minimal
2157matching constructs C<??>, C<*?>, C<+?>, C<{n,m}?>, and C<{n,}?>. The
2158rest of the extensions below have the form C<(?char...)>, where the
2159C<char> is a character that determines the type of extension.
2160
2161The first extension is an embedded comment C<(?#text)>. This embeds a
2162comment into the regular expression without affecting its meaning. The
2163comment should not have any closing parentheses in the text. An
2164example is
2165
2166 /(?# Match an integer:)[+-]?\d+/;
2167
2168This style of commenting has been largely superseded by the raw,
2169freeform commenting that is allowed with the C<//x> modifier.
2170
24549070
PG
2171The modifiers C<//i>, C<//m>, C<//s> and C<//x> (or any
2172combination thereof) can also be embedded in
47f9c88b
GS
2173a regexp using C<(?i)>, C<(?m)>, C<(?s)>, and C<(?x)>. For instance,
2174
2175 /(?i)yes/; # match 'yes' case insensitively
2176 /yes/i; # same thing
2177 /(?x)( # freeform version of an integer regexp
2178 [+-]? # match an optional sign
2179 \d+ # match a sequence of digits
2180 )
2181 /x;
2182
2183Embedded modifiers can have two important advantages over the usual
2184modifiers. Embedded modifiers allow a custom set of modifiers to
2185I<each> regexp pattern. This is great for matching an array of regexps
2186that must have different modifiers:
2187
2188 $pattern[0] = '(?i)doctor';
2189 $pattern[1] = 'Johnson';
2190 ...
2191 while (<>) {
2192 foreach $patt (@pattern) {
2193 print if /$patt/;
2194 }
2195 }
2196
24549070 2197The second advantage is that embedded modifiers (except C<//p>, which
7638d2dc 2198modifies the entire regexp) only affect the regexp
47f9c88b
GS
2199inside the group the embedded modifier is contained in. So grouping
2200can be used to localize the modifier's effects:
2201
2202 /Answer: ((?i)yes)/; # matches 'Answer: yes', 'Answer: YES', etc.
2203
2204Embedded modifiers can also turn off any modifiers already present
2205by using, e.g., C<(?-i)>. Modifiers can also be combined into
2206a single expression, e.g., C<(?s-i)> turns on single line mode and
2207turns off case insensitivity.
2208
7638d2dc 2209Embedded modifiers may also be added to a non-capturing grouping.
47f9c88b
GS
2210C<(?i-m:regexp)> is a non-capturing grouping that matches C<regexp>
2211case insensitively and turns off multi-line mode.
2212
7638d2dc 2213
47f9c88b
GS
2214=head2 Looking ahead and looking behind
2215
2216This section concerns the lookahead and lookbehind assertions. First,
2217a little background.
2218
2219In Perl regular expressions, most regexp elements 'eat up' a certain
2220amount of string when they match. For instance, the regexp element
2221C<[abc}]> eats up one character of the string when it matches, in the
7638d2dc 2222sense that Perl moves to the next character position in the string
47f9c88b
GS
2223after the match. There are some elements, however, that don't eat up
2224characters (advance the character position) if they match. The examples
2225we have seen so far are the anchors. The anchor C<^> matches the
2226beginning of the line, but doesn't eat any characters. Similarly, the
7638d2dc 2227word boundary anchor C<\b> matches wherever a character matching C<\w>
353c6505 2228is next to a character that doesn't, but it doesn't eat up any
7638d2dc
WL
2229characters itself. Anchors are examples of I<zero-width assertions>.
2230Zero-width, because they consume
47f9c88b
GS
2231no characters, and assertions, because they test some property of the
2232string. In the context of our walk in the woods analogy to regexp
2233matching, most regexp elements move us along a trail, but anchors have
2234us stop a moment and check our surroundings. If the local environment
2235checks out, we can proceed forward. But if the local environment
2236doesn't satisfy us, we must backtrack.
2237
2238Checking the environment entails either looking ahead on the trail,
2239looking behind, or both. C<^> looks behind, to see that there are no
2240characters before. C<$> looks ahead, to see that there are no
2241characters after. C<\b> looks both ahead and behind, to see if the
7638d2dc 2242characters on either side differ in their "word-ness".
47f9c88b
GS
2243
2244The lookahead and lookbehind assertions are generalizations of the
2245anchor concept. Lookahead and lookbehind are zero-width assertions
2246that let us specify which characters we want to test for. The
2247lookahead assertion is denoted by C<(?=regexp)> and the lookbehind
a6b2f353 2248assertion is denoted by C<< (?<=fixed-regexp) >>. Some examples are
47f9c88b
GS
2249
2250 $x = "I catch the housecat 'Tom-cat' with catnip";
7638d2dc 2251 $x =~ /cat(?=\s)/; # matches 'cat' in 'housecat'
47f9c88b
GS
2252 @catwords = ($x =~ /(?<=\s)cat\w+/g); # matches,
2253 # $catwords[0] = 'catch'
2254 # $catwords[1] = 'catnip'
2255 $x =~ /\bcat\b/; # matches 'cat' in 'Tom-cat'
2256 $x =~ /(?<=\s)cat(?=\s)/; # doesn't match; no isolated 'cat' in
2257 # middle of $x
2258
a6b2f353 2259Note that the parentheses in C<(?=regexp)> and C<< (?<=regexp) >> are
47f9c88b
GS
2260non-capturing, since these are zero-width assertions. Thus in the
2261second regexp, the substrings captured are those of the whole regexp
a6b2f353
GS
2262itself. Lookahead C<(?=regexp)> can match arbitrary regexps, but
2263lookbehind C<< (?<=fixed-regexp) >> only works for regexps of fixed
2264width, i.e., a fixed number of characters long. Thus
2265C<< (?<=(ab|bc)) >> is fine, but C<< (?<=(ab)*) >> is not. The
2266negated versions of the lookahead and lookbehind assertions are
2267denoted by C<(?!regexp)> and C<< (?<!fixed-regexp) >> respectively.
2268They evaluate true if the regexps do I<not> match:
47f9c88b
GS
2269
2270 $x = "foobar";
2271 $x =~ /foo(?!bar)/; # doesn't match, 'bar' follows 'foo'
2272 $x =~ /foo(?!baz)/; # matches, 'baz' doesn't follow 'foo'
2273 $x =~ /(?<!\s)foo/; # matches, there is no \s before 'foo'
2274
f14c76ed
RGS
2275The C<\C> is unsupported in lookbehind, because the already
2276treacherous definition of C<\C> would become even more so
2277when going backwards.
2278
7638d2dc
WL
2279Here is an example where a string containing blank-separated words,
2280numbers and single dashes is to be split into its components.
2281Using C</\s+/> alone won't work, because spaces are not required between
2282dashes, or a word or a dash. Additional places for a split are established
2283by looking ahead and behind:
47f9c88b 2284
7638d2dc
WL
2285 $str = "one two - --6-8";
2286 @toks = split / \s+ # a run of spaces
2287 | (?<=\S) (?=-) # any non-space followed by '-'
2288 | (?<=-) (?=\S) # a '-' followed by any non-space
2289 /x, $str; # @toks = qw(one two - - - 6 - 8)
47f9c88b 2290
7638d2dc
WL
2291
2292=head2 Using independent subexpressions to prevent backtracking
2293
2294I<Independent subexpressions> are regular expressions, in the
47f9c88b
GS
2295context of a larger regular expression, that function independently of
2296the larger regular expression. That is, they consume as much or as
2297little of the string as they wish without regard for the ability of
2298the larger regexp to match. Independent subexpressions are represented
2299by C<< (?>regexp) >>. We can illustrate their behavior by first
2300considering an ordinary regexp:
2301
2302 $x = "ab";
2303 $x =~ /a*ab/; # matches
2304
2305This obviously matches, but in the process of matching, the
2306subexpression C<a*> first grabbed the C<a>. Doing so, however,
2307wouldn't allow the whole regexp to match, so after backtracking, C<a*>
2308eventually gave back the C<a> and matched the empty string. Here, what
2309C<a*> matched was I<dependent> on what the rest of the regexp matched.
2310
2311Contrast that with an independent subexpression:
2312
2313 $x =~ /(?>a*)ab/; # doesn't match!
2314
2315The independent subexpression C<< (?>a*) >> doesn't care about the rest
2316of the regexp, so it sees an C<a> and grabs it. Then the rest of the
2317regexp C<ab> cannot match. Because C<< (?>a*) >> is independent, there
da75cd15 2318is no backtracking and the independent subexpression does not give
47f9c88b
GS
2319up its C<a>. Thus the match of the regexp as a whole fails. A similar
2320behavior occurs with completely independent regexps:
2321
2322 $x = "ab";
2323 $x =~ /a*/g; # matches, eats an 'a'
2324 $x =~ /\Gab/g; # doesn't match, no 'a' available
2325
2326Here C<//g> and C<\G> create a 'tag team' handoff of the string from
2327one regexp to the other. Regexps with an independent subexpression are
2328much like this, with a handoff of the string to the independent
2329subexpression, and a handoff of the string back to the enclosing
2330regexp.
2331
2332The ability of an independent subexpression to prevent backtracking
2333can be quite useful. Suppose we want to match a non-empty string
2334enclosed in parentheses up to two levels deep. Then the following
2335regexp matches:
2336
2337 $x = "abc(de(fg)h"; # unbalanced parentheses
2338 $x =~ /\( ( [^()]+ | \([^()]*\) )+ \)/x;
2339
2340The regexp matches an open parenthesis, one or more copies of an
2341alternation, and a close parenthesis. The alternation is two-way, with
2342the first alternative C<[^()]+> matching a substring with no
2343parentheses and the second alternative C<\([^()]*\)> matching a
2344substring delimited by parentheses. The problem with this regexp is
2345that it is pathological: it has nested indeterminate quantifiers
07698885 2346of the form C<(a+|b)+>. We discussed in Part 1 how nested quantifiers
47f9c88b
GS
2347like this could take an exponentially long time to execute if there
2348was no match possible. To prevent the exponential blowup, we need to
2349prevent useless backtracking at some point. This can be done by
2350enclosing the inner quantifier as an independent subexpression:
2351
2352 $x =~ /\( ( (?>[^()]+) | \([^()]*\) )+ \)/x;
2353
2354Here, C<< (?>[^()]+) >> breaks the degeneracy of string partitioning
2355by gobbling up as much of the string as possible and keeping it. Then
2356match failures fail much more quickly.
2357
7638d2dc 2358
47f9c88b
GS
2359=head2 Conditional expressions
2360
7638d2dc 2361A I<conditional expression> is a form of if-then-else statement
47f9c88b
GS
2362that allows one to choose which patterns are to be matched, based on
2363some condition. There are two types of conditional expression:
2364C<(?(condition)yes-regexp)> and
2365C<(?(condition)yes-regexp|no-regexp)>. C<(?(condition)yes-regexp)> is
7638d2dc 2366like an S<C<'if () {}'>> statement in Perl. If the C<condition> is true,
47f9c88b 2367the C<yes-regexp> will be matched. If the C<condition> is false, the
7638d2dc
WL
2368C<yes-regexp> will be skipped and Perl will move onto the next regexp
2369element. The second form is like an S<C<'if () {} else {}'>> statement
47f9c88b
GS
2370in Perl. If the C<condition> is true, the C<yes-regexp> will be
2371matched, otherwise the C<no-regexp> will be matched.
2372
7638d2dc 2373The C<condition> can have several forms. The first form is simply an
47f9c88b 2374integer in parentheses C<(integer)>. It is true if the corresponding
7638d2dc 2375backreference C<\integer> matched earlier in the regexp. The same
c27a5cfe 2376thing can be done with a name associated with a capture group, written
7638d2dc 2377as C<< (<name>) >> or C<< ('name') >>. The second form is a bare
353c6505 2378zero width assertion C<(?...)>, either a lookahead, a lookbehind, or a
7638d2dc
WL
2379code assertion (discussed in the next section). The third set of forms
2380provides tests that return true if the expression is executed within
2381a recursion (C<(R)>) or is being called from some capturing group,
2382referenced either by number (C<(R1)>, C<(R2)>,...) or by name
2383(C<(R&name)>).
2384
2385The integer or name form of the C<condition> allows us to choose,
2386with more flexibility, what to match based on what matched earlier in the
2387regexp. This searches for words of the form C<"$x$x"> or C<"$x$y$y$x">:
47f9c88b 2388
d8b950dc 2389 % simple_grep '^(\w+)(\w+)?(?(2)\g2\g1|\g1)$' /usr/dict/words
47f9c88b
GS
2390 beriberi
2391 coco
2392 couscous
2393 deed
2394 ...
2395 toot
2396 toto
2397 tutu
2398
2399The lookbehind C<condition> allows, along with backreferences,
2400an earlier part of the match to influence a later part of the
2401match. For instance,
2402
2403 /[ATGC]+(?(?<=AA)G|C)$/;
2404
2405matches a DNA sequence such that it either ends in C<AAG>, or some
2406other base pair combination and C<C>. Note that the form is
a6b2f353
GS
2407C<< (?(?<=AA)G|C) >> and not C<< (?((?<=AA))G|C) >>; for the
2408lookahead, lookbehind or code assertions, the parentheses around the
2409conditional are not needed.
47f9c88b 2410
7638d2dc
WL
2411
2412=head2 Defining named patterns
2413
2414Some regular expressions use identical subpatterns in several places.
2415Starting with Perl 5.10, it is possible to define named subpatterns in
2416a section of the pattern so that they can be called up by name
2417anywhere in the pattern. This syntactic pattern for this definition
2418group is C<< (?(DEFINE)(?<name>pattern)...) >>. An insertion
2419of a named pattern is written as C<(?&name)>.
2420
2421The example below illustrates this feature using the pattern for
2422floating point numbers that was presented earlier on. The three
2423subpatterns that are used more than once are the optional sign, the
2424digit sequence for an integer and the decimal fraction. The DEFINE
2425group at the end of the pattern contains their definition. Notice
2426that the decimal fraction pattern is the first place where we can
2427reuse the integer pattern.
2428
353c6505 2429 /^ (?&osg)\ * ( (?&int)(?&dec)? | (?&dec) )
7638d2dc
WL
2430 (?: [eE](?&osg)(?&int) )?
2431 $
2432 (?(DEFINE)
2433 (?<osg>[-+]?) # optional sign
2434 (?<int>\d++) # integer
2435 (?<dec>\.(?&int)) # decimal fraction
2436 )/x
2437
2438
2439=head2 Recursive patterns
2440
2441This feature (introduced in Perl 5.10) significantly extends the
2442power of Perl's pattern matching. By referring to some other
2443capture group anywhere in the pattern with the construct
353c6505 2444C<(?group-ref)>, the I<pattern> within the referenced group is used
7638d2dc
WL
2445as an independent subpattern in place of the group reference itself.
2446Because the group reference may be contained I<within> the group it
2447refers to, it is now possible to apply pattern matching to tasks that
2448hitherto required a recursive parser.
2449
2450To illustrate this feature, we'll design a pattern that matches if
2451a string contains a palindrome. (This is a word or a sentence that,
2452while ignoring spaces, interpunctuation and case, reads the same backwards
2453as forwards. We begin by observing that the empty string or a string
2454containing just one word character is a palindrome. Otherwise it must
2455have a word character up front and the same at its end, with another
2456palindrome in between.
2457
fd2b7f55 2458 /(?: (\w) (?...Here be a palindrome...) \g{-1} | \w? )/x
7638d2dc 2459
e57a4e52 2460Adding C<\W*> at either end to eliminate what is to be ignored, we already
7638d2dc
WL
2461have the full pattern:
2462
2463 my $pp = qr/^(\W* (?: (\w) (?1) \g{-1} | \w? ) \W*)$/ix;
2464 for $s ( "saippuakauppias", "A man, a plan, a canal: Panama!" ){
2465 print "'$s' is a palindrome\n" if $s =~ /$pp/;
2466 }
2467
2468In C<(?...)> both absolute and relative backreferences may be used.
2469The entire pattern can be reinserted with C<(?R)> or C<(?0)>.
c27a5cfe
KW
2470If you prefer to name your groups, you can use C<(?&name)> to
2471recurse into that group.
7638d2dc
WL
2472
2473
47f9c88b
GS
2474=head2 A bit of magic: executing Perl code in a regular expression
2475
2476Normally, regexps are a part of Perl expressions.
7638d2dc 2477I<Code evaluation> expressions turn that around by allowing
da75cd15 2478arbitrary Perl code to be a part of a regexp. A code evaluation
7638d2dc 2479expression is denoted C<(?{code})>, with I<code> a string of Perl
47f9c88b
GS
2480statements.
2481
353c6505 2482Be warned that this feature is considered experimental, and may be
7638d2dc
WL
2483changed without notice.
2484
47f9c88b
GS
2485Code expressions are zero-width assertions, and the value they return
2486depends on their environment. There are two possibilities: either the
2487code expression is used as a conditional in a conditional expression
2488C<(?(condition)...)>, or it is not. If the code expression is a
2489conditional, the code is evaluated and the result (i.e., the result of
2490the last statement) is used to determine truth or falsehood. If the
2491code expression is not used as a conditional, the assertion always
2492evaluates true and the result is put into the special variable
2493C<$^R>. The variable C<$^R> can then be used in code expressions later
2494in the regexp. Here are some silly examples:
2495
2496 $x = "abcdef";
2497 $x =~ /abc(?{print "Hi Mom!";})def/; # matches,
2498 # prints 'Hi Mom!'
2499 $x =~ /aaa(?{print "Hi Mom!";})def/; # doesn't match,
2500 # no 'Hi Mom!'
745e1e41
DC
2501
2502Pay careful attention to the next example:
2503
47f9c88b
GS
2504 $x =~ /abc(?{print "Hi Mom!";})ddd/; # doesn't match,
2505 # no 'Hi Mom!'
745e1e41
DC
2506 # but why not?
2507
2508At first glance, you'd think that it shouldn't print, because obviously
2509the C<ddd> isn't going to match the target string. But look at this
2510example:
2511
87167316
RGS
2512 $x =~ /abc(?{print "Hi Mom!";})[dD]dd/; # doesn't match,
2513 # but _does_ print
745e1e41
DC
2514
2515Hmm. What happened here? If you've been following along, you know that
ac036724 2516the above pattern should be effectively (almost) the same as the last one;
2517enclosing the C<d> in a character class isn't going to change what it
745e1e41
DC
2518matches. So why does the first not print while the second one does?
2519
7638d2dc 2520The answer lies in the optimizations the regex engine makes. In the first
745e1e41
DC
2521case, all the engine sees are plain old characters (aside from the
2522C<?{}> construct). It's smart enough to realize that the string 'ddd'
2523doesn't occur in our target string before actually running the pattern
2524through. But in the second case, we've tricked it into thinking that our
87167316 2525pattern is more complicated. It takes a look, sees our
745e1e41
DC
2526character class, and decides that it will have to actually run the
2527pattern to determine whether or not it matches, and in the process of
2528running it hits the print statement before it discovers that we don't
2529have a match.
2530
2531To take a closer look at how the engine does optimizations, see the
2532section L<"Pragmas and debugging"> below.
2533
2534More fun with C<?{}>:
2535
47f9c88b
GS
2536 $x =~ /(?{print "Hi Mom!";})/; # matches,
2537 # prints 'Hi Mom!'
2538 $x =~ /(?{$c = 1;})(?{print "$c";})/; # matches,
2539 # prints '1'
2540 $x =~ /(?{$c = 1;})(?{print "$^R";})/; # matches,
2541 # prints '1'
2542
2543The bit of magic mentioned in the section title occurs when the regexp
2544backtracks in the process of searching for a match. If the regexp
2545backtracks over a code expression and if the variables used within are
2546localized using C<local>, the changes in the variables produced by the
2547code expression are undone! Thus, if we wanted to count how many times
2548a character got matched inside a group, we could use, e.g.,
2549
2550 $x = "aaaa";
2551 $count = 0; # initialize 'a' count
2552 $c = "bob"; # test if $c gets clobbered
2553 $x =~ /(?{local $c = 0;}) # initialize count
2554 ( a # match 'a'
2555 (?{local $c = $c + 1;}) # increment count
2556 )* # do this any number of times,
2557 aa # but match 'aa' at the end
2558 (?{$count = $c;}) # copy local $c var into $count
2559 /x;
2560 print "'a' count is $count, \$c variable is '$c'\n";
2561
2562This prints
2563
2564 'a' count is 2, $c variable is 'bob'
2565
7638d2dc
WL
2566If we replace the S<C< (?{local $c = $c + 1;})>> with
2567S<C< (?{$c = $c + 1;})>>, the variable changes are I<not> undone
47f9c88b
GS
2568during backtracking, and we get
2569
2570 'a' count is 4, $c variable is 'bob'
2571
2572Note that only localized variable changes are undone. Other side
2573effects of code expression execution are permanent. Thus
2574
2575 $x = "aaaa";
2576 $x =~ /(a(?{print "Yow\n";}))*aa/;
2577
2578produces
2579
2580 Yow
2581 Yow
2582 Yow
2583 Yow
2584
2585The result C<$^R> is automatically localized, so that it will behave
2586properly in the presence of backtracking.
2587
7638d2dc
WL
2588This example uses a code expression in a conditional to match a
2589definite article, either 'the' in English or 'der|die|das' in German:
47f9c88b 2590
47f9c88b
GS
2591 $lang = 'DE'; # use German
2592 ...
2593 $text = "das";
2594 print "matched\n"
2595 if $text =~ /(?(?{
2596 $lang eq 'EN'; # is the language English?
2597 })
2598 the | # if so, then match 'the'
7638d2dc 2599 (der|die|das) # else, match 'der|die|das'
47f9c88b
GS
2600 )
2601 /xi;
2602
2603Note that the syntax here is C<(?(?{...})yes-regexp|no-regexp)>, not
2604C<(?((?{...}))yes-regexp|no-regexp)>. In other words, in the case of a
2605code expression, we don't need the extra parentheses around the
2606conditional.
2607
7638d2dc 2608If you try to use code expressions with interpolating variables, Perl
a6b2f353
GS
2609may surprise you:
2610
2611 $bar = 5;
2612 $pat = '(?{ 1 })';
2613 /foo(?{ $bar })bar/; # compiles ok, $bar not interpolated
2614 /foo(?{ 1 })$bar/; # compile error!
2615 /foo${pat}bar/; # compile error!
2616
2617 $pat = qr/(?{ $foo = 1 })/; # precompile code regexp
2618 /foo${pat}bar/; # compiles ok
2619
fa11829f 2620If a regexp has (1) code expressions and interpolating variables, or
7638d2dc 2621(2) a variable that interpolates a code expression, Perl treats the
a6b2f353
GS
2622regexp as an error. If the code expression is precompiled into a
2623variable, however, interpolating is ok. The question is, why is this
2624an error?
2625
2626The reason is that variable interpolation and code expressions
2627together pose a security risk. The combination is dangerous because
2628many programmers who write search engines often take user input and
2629plug it directly into a regexp:
47f9c88b
GS
2630
2631 $regexp = <>; # read user-supplied regexp
2632 $chomp $regexp; # get rid of possible newline
2633 $text =~ /$regexp/; # search $text for the $regexp
2634
a6b2f353
GS
2635If the C<$regexp> variable contains a code expression, the user could
2636then execute arbitrary Perl code. For instance, some joker could
7638d2dc
WL
2637search for S<C<system('rm -rf *');>> to erase your files. In this
2638sense, the combination of interpolation and code expressions I<taints>
47f9c88b 2639your regexp. So by default, using both interpolation and code
a6b2f353
GS
2640expressions in the same regexp is not allowed. If you're not
2641concerned about malicious users, it is possible to bypass this
7638d2dc 2642security check by invoking S<C<use re 'eval'>>:
a6b2f353
GS
2643
2644 use re 'eval'; # throw caution out the door
2645 $bar = 5;
2646 $pat = '(?{ 1 })';
2647 /foo(?{ 1 })$bar/; # compiles ok
2648 /foo${pat}bar/; # compiles ok
47f9c88b 2649
7638d2dc 2650Another form of code expression is the I<pattern code expression>.
47f9c88b
GS
2651The pattern code expression is like a regular code expression, except
2652that the result of the code evaluation is treated as a regular
2653expression and matched immediately. A simple example is
2654
2655 $length = 5;
2656 $char = 'a';
2657 $x = 'aaaaabb';
2658 $x =~ /(??{$char x $length})/x; # matches, there are 5 of 'a'
2659
2660
2661This final example contains both ordinary and pattern code
7638d2dc 2662expressions. It detects whether a binary string C<1101010010001...> has a
47f9c88b
GS
2663Fibonacci spacing 0,1,1,2,3,5,... of the C<1>'s:
2664
47f9c88b 2665 $x = "1101010010001000001";
7638d2dc 2666 $z0 = ''; $z1 = '0'; # initial conditions
47f9c88b
GS
2667 print "It is a Fibonacci sequence\n"
2668 if $x =~ /^1 # match an initial '1'
7638d2dc
WL
2669 (?:
2670 ((??{ $z0 })) # match some '0'
2671 1 # and then a '1'
2672 (?{ $z0 = $z1; $z1 .= $^N; })
47f9c88b
GS
2673 )+ # repeat as needed
2674 $ # that is all there is
2675 /x;
7638d2dc 2676 printf "Largest sequence matched was %d\n", length($z1)-length($z0);
47f9c88b 2677
7638d2dc
WL
2678Remember that C<$^N> is set to whatever was matched by the last
2679completed capture group. This prints
47f9c88b
GS
2680
2681 It is a Fibonacci sequence
2682 Largest sequence matched was 5
2683
2684Ha! Try that with your garden variety regexp package...
2685
7638d2dc 2686Note that the variables C<$z0> and C<$z1> are not substituted when the
47f9c88b 2687regexp is compiled, as happens for ordinary variables outside a code
7638d2dc 2688expression. Rather, the code expressions are evaluated when Perl
47f9c88b
GS
2689encounters them during the search for a match.
2690
2691The regexp without the C<//x> modifier is
2692
7638d2dc
WL
2693 /^1(?:((??{ $z0 }))1(?{ $z0 = $z1; $z1 .= $^N; }))+$/
2694
2695which shows that spaces are still possible in the code parts. Nevertheless,
353c6505 2696when working with code and conditional expressions, the extended form of
7638d2dc
WL
2697regexps is almost necessary in creating and debugging regexps.
2698
2699
2700=head2 Backtracking control verbs
2701
2702Perl 5.10 introduced a number of control verbs intended to provide
2703detailed control over the backtracking process, by directly influencing
2704the regexp engine and by providing monitoring techniques. As all
2705the features in this group are experimental and subject to change or
2706removal in a future version of Perl, the interested reader is
2707referred to L<perlre/"Special Backtracking Control Verbs"> for a
2708detailed description.
2709
2710Below is just one example, illustrating the control verb C<(*FAIL)>,
2711which may be abbreviated as C<(*F)>. If this is inserted in a regexp
2712it will cause to fail, just like at some mismatch between the pattern
2713and the string. Processing of the regexp continues like after any "normal"
353c6505
DL
2714failure, so that, for instance, the next position in the string or another
2715alternative will be tried. As failing to match doesn't preserve capture
c27a5cfe 2716groups or produce results, it may be necessary to use this in
7638d2dc
WL
2717combination with embedded code.
2718
2719 %count = ();
2720 "supercalifragilisticexpialidoceous" =~
2721 /([aeiou])(?{ $count{$1}++; })(*FAIL)/oi;
2722 printf "%3d '%s'\n", $count{$_}, $_ for (sort keys %count);
2723
353c6505
DL
2724The pattern begins with a class matching a subset of letters. Whenever
2725this matches, a statement like C<$count{'a'}++;> is executed, incrementing
2726the letter's counter. Then C<(*FAIL)> does what it says, and
2727the regexp engine proceeds according to the book: as long as the end of
2728the string hasn't been reached, the position is advanced before looking
7638d2dc 2729for another vowel. Thus, match or no match makes no difference, and the
e1020413 2730regexp engine proceeds until the entire string has been inspected.
7638d2dc
WL
2731(It's remarkable that an alternative solution using something like
2732
2733 $count{lc($_)}++ for split('', "supercalifragilisticexpialidoceous");
2734 printf "%3d '%s'\n", $count2{$_}, $_ for ( qw{ a e i o u } );
2735
2736is considerably slower.)
47f9c88b 2737
47f9c88b
GS
2738
2739=head2 Pragmas and debugging
2740
2741Speaking of debugging, there are several pragmas available to control
2742and debug regexps in Perl. We have already encountered one pragma in
7638d2dc 2743the previous section, S<C<use re 'eval';>>, that allows variable
a6b2f353
GS
2744interpolation and code expressions to coexist in a regexp. The other
2745pragmas are
47f9c88b
GS
2746
2747 use re 'taint';
2748 $tainted = <>;
2749 @parts = ($tainted =~ /(\w+)\s+(\w+)/; # @parts is now tainted
2750
2751The C<taint> pragma causes any substrings from a match with a tainted
2752variable to be tainted as well. This is not normally the case, as
2753regexps are often used to extract the safe bits from a tainted
2754variable. Use C<taint> when you are not extracting safe bits, but are
2755performing some other processing. Both C<taint> and C<eval> pragmas
a6b2f353 2756are lexically scoped, which means they are in effect only until
47f9c88b
GS
2757the end of the block enclosing the pragmas.
2758
2759 use re 'debug';
2760 /^(.*)$/s; # output debugging info
2761
2762 use re 'debugcolor';
2763 /^(.*)$/s; # output debugging info in living color
2764
2765The global C<debug> and C<debugcolor> pragmas allow one to get
2766detailed debugging info about regexp compilation and
2767execution. C<debugcolor> is the same as debug, except the debugging
2768information is displayed in color on terminals that can display
2769termcap color sequences. Here is example output:
2770
2771 % perl -e 'use re "debug"; "abc" =~ /a*b+c/;'
2772 Compiling REx `a*b+c'
2773 size 9 first at 1
2774 1: STAR(4)
2775 2: EXACT <a>(0)
2776 4: PLUS(7)
2777 5: EXACT <b>(0)
2778 7: EXACT <c>(9)
2779 9: END(0)
2780 floating `bc' at 0..2147483647 (checking floating) minlen 2
2781 Guessing start of match, REx `a*b+c' against `abc'...
2782 Found floating substr `bc' at offset 1...
2783 Guessed: match at offset 0
2784 Matching REx `a*b+c' against `abc'
2785 Setting an EVAL scope, savestack=3
2786 0 <> <abc> | 1: STAR
2787 EXACT <a> can match 1 times out of 32767...
2788 Setting an EVAL scope, savestack=3
2789 1 <a> <bc> | 4: PLUS
2790 EXACT <b> can match 1 times out of 32767...
2791 Setting an EVAL scope, savestack=3
2792 2 <ab> <c> | 7: EXACT <c>
2793 3 <abc> <> | 9: END
2794 Match successful!
2795 Freeing REx: `a*b+c'
2796
2797If you have gotten this far into the tutorial, you can probably guess
2798what the different parts of the debugging output tell you. The first
2799part
2800
2801 Compiling REx `a*b+c'
2802 size 9 first at 1
2803 1: STAR(4)
2804 2: EXACT <a>(0)
2805 4: PLUS(7)
2806 5: EXACT <b>(0)
2807 7: EXACT <c>(9)
2808 9: END(0)
2809
2810describes the compilation stage. C<STAR(4)> means that there is a
2811starred object, in this case C<'a'>, and if it matches, goto line 4,
2812i.e., C<PLUS(7)>. The middle lines describe some heuristics and
2813optimizations performed before a match:
2814
2815 floating `bc' at 0..2147483647 (checking floating) minlen 2
2816 Guessing start of match, REx `a*b+c' against `abc'...
2817 Found floating substr `bc' at offset 1...
2818 Guessed: match at offset 0
2819
2820Then the match is executed and the remaining lines describe the
2821process:
2822
2823 Matching REx `a*b+c' against `abc'
2824 Setting an EVAL scope, savestack=3
2825 0 <> <abc> | 1: STAR
2826 EXACT <a> can match 1 times out of 32767...
2827 Setting an EVAL scope, savestack=3
2828 1 <a> <bc> | 4: PLUS
2829 EXACT <b> can match 1 times out of 32767...
2830 Setting an EVAL scope, savestack=3
2831 2 <ab> <c> | 7: EXACT <c>
2832 3 <abc> <> | 9: END
2833 Match successful!
2834 Freeing REx: `a*b+c'
2835
7638d2dc 2836Each step is of the form S<C<< n <x> <y> >>>, with C<< <x> >> the
47f9c88b 2837part of the string matched and C<< <y> >> the part not yet
7638d2dc 2838matched. The S<C<< | 1: STAR >>> says that Perl is at line number 1
47f9c88b
GS
2839n the compilation list above. See
2840L<perldebguts/"Debugging regular expressions"> for much more detail.
2841
2842An alternative method of debugging regexps is to embed C<print>
2843statements within the regexp. This provides a blow-by-blow account of
2844the backtracking in an alternation:
2845
2846 "that this" =~ m@(?{print "Start at position ", pos, "\n";})
2847 t(?{print "t1\n";})
2848 h(?{print "h1\n";})
2849 i(?{print "i1\n";})
2850 s(?{print "s1\n";})
2851 |
2852 t(?{print "t2\n";})
2853 h(?{print "h2\n";})
2854 a(?{print "a2\n";})
2855 t(?{print "t2\n";})
2856 (?{print "Done at position ", pos, "\n";})
2857 @x;
2858
2859prints
2860
2861 Start at position 0
2862 t1
2863 h1
2864 t2
2865 h2
2866 a2
2867 t2
2868 Done at position 4
2869
2870=head1 BUGS
2871
2872Code expressions, conditional expressions, and independent expressions
7638d2dc 2873are I<experimental>. Don't use them in production code. Yet.
47f9c88b
GS
2874
2875=head1 SEE ALSO
2876
7638d2dc 2877This is just a tutorial. For the full story on Perl regular
47f9c88b
GS
2878expressions, see the L<perlre> regular expressions reference page.
2879
2880For more information on the matching C<m//> and substitution C<s///>
2881operators, see L<perlop/"Regexp Quote-Like Operators">. For
2882information on the C<split> operation, see L<perlfunc/split>.
2883
2884For an excellent all-around resource on the care and feeding of
2885regular expressions, see the book I<Mastering Regular Expressions> by
2886Jeffrey Friedl (published by O'Reilly, ISBN 1556592-257-3).
2887
2888=head1 AUTHOR AND COPYRIGHT
2889
2890Copyright (c) 2000 Mark Kvale
2891All rights reserved.
2892
2893This document may be distributed under the same terms as Perl itself.
2894
2895=head2 Acknowledgments
2896
2897The inspiration for the stop codon DNA example came from the ZIP
2898code example in chapter 7 of I<Mastering Regular Expressions>.
2899
a6b2f353
GS
2900The author would like to thank Jeff Pinyan, Andrew Johnson, Peter
2901Haworth, Ronald J Kimball, and Joe Smith for all their helpful
2902comments.
47f9c88b
GS
2903
2904=cut
a6b2f353 2905