Passing cursor record values to procedures - looking for advice

Question

I frequently use cursors in my programs and need to pass the values from said cursor around in the program to other procedures/functions.  I am looking for advice on the best or recommended way of doing this.
Currently the majority of my programs do something like this:
CREATE OR REPLACE PACKAGE BODY some_pkg AS
     CURSOR some_cursor IS
          SELECT t.col1,
                           t.col2,
                          t.col3,
               FROM some_table t;
PROCEDURE sub_proc
     (p_cursor               IN      some_cursor%ROWTYPE)
IS
BEGIN
     dbms_output.put_line(p_cursor.col1);
END sub_proc;
PROCEDURE main_process
IS
     l_cursor          some_cursor%ROWTYPE;
BEGIN
     FOR l_cursor IN some_cursor LOOP
          sub_proc(l_cursor);
     END LOOP;
END main_process;
END some_pkg;
Basically I create the cursor as global so I can use it's %ROWTYPE as a parameter datatype in my procedures. 
I understand there are other options such as using a defined record type, or ref cursors, but I'm not totally clear on how ref cursors would help me, I don't really understand them well.
Can anyone provide some feedback on how I currently handle this situation vs. the alternatives?
Thanks

Answer 1

I am not sure what you need to do.
You can base records on cursors or tables using the %ROWTYPE operator.
Cursors can be defined in the package header and used elsewhere by using normal cursor functionality (untesed)
create or replace package my_package as
  cursor my_cursor is
    select *
      from dual;
  my_procedure;
end my_package;
create or replace package body my_package as
  procedure my_procedure is
  my_record my_package.my_cursor%rowtype;
  begin
    for r in my_package.my_cursor loop
      my_record := r;
    end;
  end my_procedure;
end my_package;
  This can allow you to define the cursor in one place and reference it elsewhere.
Reference cursors are pointers to result sets.  They reference actual datasets and can be passed between routines if defined in package headers.  There is a little more overhead using reference cursors but they support dynamic sql if needed (it usually isn't). and you MUST carefully coordinate FETCH commands to receive the values the cursor is selecting (this is harder than it looks)(untested)
create or replace package my_package as
  type my_refcur_type is sys_refcursor;
  my_procedure;
end my_package;
create or replace package body my_package as
  procedure my_procedure(p_refcur out my_package.my_refcur_type) is
  begin
    open p_refcur for 'select * from dual';
end my_package;
--elsewhere
my_package.my_procedure(my_cursor);
fetch my_cursor into v_dummy;
Hope this helps