exussum0 didn't object to delegation in general, just to the way I did it in my example. And he has a good point too: It's usually better to access the delegated object through an accessor method, because of the arguments he gave, and because it allows lazy loading. In that way, the delegated object is only constructed on first use. It also allows easier overriding.

If I were to rewrite my example using an accessor, it'd look something like this:

package My::Module;

sub new {
    my $class = shift;
    my $self = {};    # no clutter!

    bless $self, $class;
}

sub some_method {
    my $self = shift;

    # we use it here
    my $result = $self->_csv->some_csv_action( @input ); # uses access
+or
}

sub _csv {
    my $self = shift;

    # allow to be set from outside
    if(@_) {
        $self->{_csv} = shift;
    }

    # default to Text::CSV
    if( !defined $self->{_csv} ) {
        require Text::CSV;
        $self->{_csv} = Text::CSV->new( 'some args' );
    }

    return $self->{_csv};
}
[download]

I'm sure that could be refined further, but it demonstrates the basic idea. The benefit should be clear: it's now possible to supply a My::Module object with a different CSV object, which makes testing easier, as well as customizing the behavior of our class.