"And finally, running any regexp match overwrites the values of all the number variables (for instance, $2), so after checking $2 =~ /css/i, $2 will be undef."

Not true. Perl doesn't garuntee that it will overright those unless they actualy match. Since that regex isn't doing any captures he should be safe, not that I think its the best way to do it, but its safe.

"Since that regex isn't doing any captures he should be safe" is wrong. Captures or not, $2 is cleared on a successful match.

$cfm = 'moo';
'moo' =~ /()(.*)/;

if (($2 =~ /css/i) || ($2 eq $cfm)) {
  print(defined($2)?1:0, $/);             # 1, cause match failed
  $uniques{$2}++;
}

'css' =~ /()(.*)/;

if (($2 =~ /css/i) || ($2 eq $cfm)) {
  print(defined($2)?1:0, $/);             # 0, cause match succeeded
  $uniques{$2}++;                         # $uniques{undef} ???
}
[download]

You're right, it only resets them on a successful match, not on an attempted match. But it will overwrite all of them if there's a successful match, even if the match doesn't include any parens. So if the string /css/ matches, $2 will be undef, and in that case the code won't work right.

/me eats his words.

You are 100% correct. Here is an example (that I had to run to beleive it).

my $test = "Hello world!";
print "Matched\n" if $test =~ /(Hello)/;
print "\$1 = $1\n";
print "Matched\n" if $test =~ /world/;
print "\$1 = $1\n";

__DATA__
Matched
$1 = Hello
Matched
$1 =
[download]


___________
Eric Hodges

You shouldn't ever need to use "$2". It's just the same thing as $2.

Another case where you need to double-quote a variable is when using the string bitwise operators on something that happens to have a cached numeric value:

$ perl -le'$x = "41"; 0+$x; print for ("$x" ^ "3"), ($x ^ "3")'
1
42
[download]