package Toolkit::DB;
#############################################
# new - constructor
#
#
# returns blessed db object
#############################################
sub new
{
    my $self = shift;
    my %params = @_;
    if (!$params{dbtype})
    {
        $params{dbtype} = "Toolkit::FileDB";
    }
    $params{delegate} = "$params{dbtype}"->new(%params);
     
    return bless \%params,$self;
}

#############################################
# insert - add a record to the db
#
# takes hash of fieldname/fieldcontent
#
# returns success/fail
#############################################
sub insert
{
    my $self = shift;
    my %data = @_;

    my $result = $self->{delegate}->insert(%data);

    return $result; 
}
[download]

(With similar methods for all the other function calls needed). Then at run time you can choose which implementation class to use, or take the default ('Toolkit::FileDB' in this instance).

Update: It's just occurred to me - that way you could also create stubs for the functionality that return 'this function is not available without X::Y::Z installed' instead of a fatal error.

Update: There's some possibly related discussion about using use and require in Use and Require - making an informed choice.