Maybe I should have included this in my post. Right now I have a hash of a hash of hash... where each character maps to a key. ex.

hello equals $hash{h}{e}{l}{l}{o} = {}.

I planned on walking through this hash as I went through the dfs sub, but wanted to get the sub correct before I made it more complicated. Same thing with the singular call to dfs. I figured it would easier to examine the output if I started from one point. I could have wrapped the initial dfs call inside two for loops, but in the intrest of getting dfs working I decided to start from one point only.

Update: Here is my preprocess code, if you have any suggestions.

my ($dict) = "/usr/share/dict/words";

open (FILE, $dict) or die ("Can't open $dict: $!\n");

while (<FILE>) {
    chomp;
    next if length() < 3;
    @char = split //, $_;
    $pointer = \%{$word_tree{$char[0]}};
    for (1 .. $#char) {
        $pointer = \%{$$pointer{$char[$_]}};
    }  
}

close (FILE);
[download]