How about...

package FileHash;
use strict;
use warnings;
use Tie::Hash;
use Data::Dumper;
our @ISA = ("Tie::Hash");

sub TIEHASH {
   my ($class, $file) = @_;
   my $s = bless {
        FILE => $file,
        DATA => {},
    },$class;
   $s->_read();
   return $s;
}


sub STORE    { $_[0]->{DATA}->{$_[1]} = $_[2] }
sub CLEAR    { $_[0]->{DATA} = () }
sub FETCH    { $_[0]->{DATA}->{$_[1]} }
sub FIRSTKEY { my $a = scalar keys %{$_[0]}; each %{$_[0]->{DATA}} }
sub NEXTKEY  { each %{$_[0]->{DATA}} }
sub EXISTS   { exists $_[0]->{DATA}->{$_[1]} }
sub DELETE   { delete $_[0]->{DATA}->{$_[1]} }
sub DESTROY  { $_[0]->_write }

# ----- private subs -----
sub _write { # still need to deal with race conditions
   my $file =   $_[0]->{FILE} ;
   open (FH, "> $file") or die "Can't open $file for FileHash _write: 
+$!";
   print FH Dumper($_[0]->{DATA});  # writes out as $VAR1 = { ..... }
   close FH or die "Can't close $file for FileHash _write: $!";
}

sub _read { # still need to deal with race conditions
   my $self = shift;
   return unless -f $self->{FILE};
   $self->{DATA} = do $self->{FILE};

}


1;
[download]

Yup. Thats much better....

;-)

Yves / DeMerphq
---
Writing a good benchmark isnt as easy as it might look.