Listing 2-11 nameswap: Using backreferences

#!/usr/bin/perl -w

print 'Type your name: ';
chomp($_ = <>);

s/(\S+) (\S+)/$2 $1/;      # Swap the first and second words

print;

The two pairs of parentheses create two backreferences: $1 refers to the first (\S+), and $2 refers to the second.

% nameswap
RESULT: Type your name: Abraham Lincoln
Lincoln Abraham

\1, \2, \3...

The backreferences \1 and \2 and \3 (and so on) have the same meaning as $1 and $2 and $3, but are valid only inside s/// or m// expressions. In particular, they’re valid inside the pattern itself. That lets you represent repeated substrings inside the pattern.

Suppose you want to match any phrase that contains a double word, like Pago Pago or chocolate chocolate chip. You could use (\S+) to match the first word and \1 to match the second occurrence of the same word, as shown in Listing 2-12.

Listing 2-12 couplet: Using \1

#!/usr/bin/perl -w

while (<>) {
    print "Echo! \n" if /(\S+)\s\1/;
}

couplet matches double double but not Hello world! or soy product.

% couplet
RESULT: double double toil and trouble
Echo!
Hello world!
mime pantomime
A thousand thousand is a million.
Echo!

The programs you’ve seen so far in this chapter use \S+ to define words. That classifies strings such as “3.14159” and “:-)” and “world!” as words unto themselves, which might upset strict grammarians. They should use \w.

---

The Metacharacters \w and \W

\w matches any letter, number, or underscore.

\W matches any character not matched by \w.

---

Listing 2-13 shows advent, the world’s shortest and hardest adventure game. It lets you type an adventure-style command and then rejects whatever you typed, using /w to identify characters belonging to words and three backreferences.

Listing 2-13 advent: A simple yet difficult adventure game

#!/usr/bin/perl

while (<>) {
    chomp;

# Match a first word (the verb), any middle words,
# and an optional last word (the noun).

    ($verb, $other_words, $noun) = (/\b(\w+)\b\s+\b(.*)\b\s?\b(\w+)\b/);
    if ($noun) { print "Where is $noun?\n"; }
    else      { print "How do I $verb?\n"; }
}

advent’s regex matches a word, followed by an optional space, followed by zero or more words followed by a space, followed by a final optional word. That’s a fairly contorted pattern; what matters is that it plops the first word into $verb, the last word into $noun, and any other words in the middle into $extra_words. (For an uncluttered version of this regex, see page 111.)

Then, if a $noun exists, advent prints Where is $noun?. Otherwise, it prints How do I $verb?.

% advent
RESULT: look up
Where is up?
run
How do I run?
read the perl book
Where is book?

QUIZ 3

1.  All of these one-liners will print Soon if the user asks a question beginning with the word when. Which would you consider the best?

a.  perl -ne ‘print “Soon.\n” if /when\s/;’

b.  perl -ne ‘print “Soon.\n” if /when\s/i;’

c.  perl -ne ‘print “Soon.\n” if /when\s/g;’

d.  perl -ne ‘print “Soon.\n” if /when\S/i;’

2.  Imagine you have a calculator program with the following line, used to match an addition statement (such as “15+17”) stored in $calculator_entry:

$calculator_entry =~ /(\d+)\s*(.)\s*(\d+)/;

which stores 15 in $1, + in $2, and 17 in $3.
Which of the following is not matched by this pattern?

a.  1729 + 1089

b.  1729 + 1089

c.  Neither a nor b is matched.

d.  Both a and b are matched.

3.  What will the following program print?

#!/usr/bin/perl
$_ = 'a_b8';
print 'yes '  if /\b\w{2,4}\b/;
print 'yup '  if /\b\w{2,3}\b/;
print 'yeah ' if /\b\w{2,}\b/;
print 'yo '   if /\b\w{2,3}\b/;

a.  

yes yeah yo

b.  

yes yup yeah yo

c.  

yes yup yo

d.  

yes yup yeah

4.  What will this code fragment print? (Caution! This is tricky.)

$_ = "14:";
($hour, $minute) = /(\d+):(\d+)/;
print "--$hour hours, $minute minutes--";

a.  

-- hours,  minutes--

b.  

-- 14 hours, 0 minutes--

c.  

-- 14 hours,   minutes--

d.  Nothing—there’s a syntax error.

Exercise 3

Difficulty: Easy

Modify advent’s regular expression so that it uses the \b metacharacter.