I would like to answer this in a way that discourages passing around arrays, when passing around cursors is a more sound approach. It doesn't exactly answer the question as posed, but it is an answer. Thinking cursors instead of thinking arrays is more efficient and thus more scalable. Also, it can be much easier code to maintain.
create table customer (
customer_id number(2) primary key,
customer_name varchar2(200) );
insert into customer values (1, 'Customer One');
insert into customer values (2, 'Customer Two');
insert into customer values (3, 'Customer Three');
insert into customer values (4, 'Customer Four');
insert into customer values (5, 'Customer Five');
insert into customer values (6, 'Customer Six');
insert into customer values (7, 'Customer Seven');
CREATE OR REPLACE PACKAGE cursor_not_array IS
FUNCTION get_customers(p_max_records INTEGER, p_id_start INTEGER, p_id_end INTEGER DEFAULT NULL) RETURN SYS_REFCURSOR;
END cursor_not_array;
CREATE OR REPLACE PACKAGE BODY cursor_not_array IS
c_max_customer_id CONSTANT NUMBER(2) := 99;
FUNCTION get_customers(p_max_records INTEGER, p_id_start INTEGER, p_id_end INTEGER DEFAULT NULL) RETURN SYS_REFCURSOR IS
v_result SYS_REFCURSOR;
BEGIN
OPEN v_result FOR
SELECT customer_id,
customer_name
FROM customer
WHERE customer_id BETWEEN p_id_start AND nvl(p_id_end, c_max_customer_id)
ORDER BY customer_id;
RETURN v_result;
END;
END cursor_not_array;