Let's say you know you will be making a hash of 120,000 hashes, each with (up to) 2 values. Instead of making 120,000 tiny hashes, with the overhead of each hash, you can go ahead and make 120,000 blessed scalar-refs and one huge array of 360,000 scalars. Each object will be a blessed scalar. Something like (untested!):

package CoolObject;

my @objects;
my $next_index = 0;

sub new
{
    my $class = shift;

    my $x = $next_index;
    my $self = bless \$x, $class;
    $next_index += 3;

    return $self;
}

sub name
{
    my $self = shift;
    $object[$$self] = shift if @_;
    return $object[$$self];
}

sub attr1
{
    my $self = shift;
    $object[$$self + 1] = shift if @_;
    return $object[$$self + 1];
}

sub attr2
{
    my $self = shift;
    $object[$$self + 2] = shift if @_;
    return $object[$$self + 2];
}

1;
[download]

Those having read the Panther book will recognize this. Some obvious improvements would include:

1. Adding the kind of garbage collection he had to re-use entries. (This would be in DESTROY.)
2. Initialization of attributes. (This would be in new)
3. Making it into more of a base class that will take attributes. (This would involve adding some sort of attribute registration.)
4. Adding attribute inheritance. (This would involve attribute registration that walks the ISA-tree.)

------
We are the carpenters and bricklayers of the Information Age.

Don't go borrowing trouble. For programmers, this means Worry only about what you need to implement.