By leaving your data structure anonymous, you lose the means of access to it. The fix is simple:

package foo;
sub new 
{
   my $class = shift;
   # Default initialization
   my $self = { 'some_scalar'        => "",
                'another_scalar'     => "",
                'some_array'         => [],
                'an_array_of_arrays' => [], # also
           };
   # initialize $self from @_.
   # you may be calling package functions which expect to shift in a r
+ef to data.
   bless $self, $class;
}
[download]

Now methods can use $self to deal with the data.

Update: crazyinsomniac and chromatic have both objected to this. An instance of foo can provide itself with a name for the ref returned by bless. foo::bar() needs to shift in the instance from its first argument, that's standard. Why ever name something? I was assuming that the real constructor would find named data handy, as in setting up closures as class methods (e.g. accessors ala Class::Struct). If there is an implicit way of doing that on an anonymous ref, I'd enjoy learning it.

Update2: Added a couple of comments to code. How do you call a package function from the constructor without a named reference? A complex data struct will have special methods for dealing with its components, and some of them will be wanted in new(). In C++ you can't use object methods in a ctor, but perl is different.

After Compline,
Zaxo