Refcursor and Normal Cursor

Question

Hi Folks,
As per my understanding there is no performance wise difference between the normal cursor and ref cursor the only difference is that only through the help of the refcur we can return the result set to the calling env or calling procedure.
But in OTN site one of the excerpts from Tom I came across a statement that the refcursors can not be global means we can not declare the refcur in the package specification
the link is http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:14188501024541
but many times in my project we have declaree the refcursor in the package specifications also
Can any one explain me what Tom really meant by that statement or whether I am misinterpreting the statement
Any suggestions will be highly appreciated
Thanks in advance
Prafulla

Answer 1

Prafulla wrote:
Please clarify me cause I am not able to make it out properluyHere's an example - but instead of ref cursors I've used DBMS_SQL cursors. Both are "+global+" cursors - but with DBMS_SQL cursor we can see the value of the cursor handle. In other words we can "print" the value of the cursor variable, which is what we cannot do with a ref cursor. This enables one to read and understand the sample code better as we can physically see the value of that cursor pointer/handle om the example.
The following 2 PL/SQL code units are defined:
-- // Creates a "global" cursor and returns the cursor handle (as a number for us)
create or replace function CreateCursor( sqlStatement varchar2 ) return number is
        c       number;
begin
        c := DBMS_SQL.open_cursor;
        DBMS_SQL.parse(
                c,
                sqlStatement,
                DBMS_SQL.native
        return( c );
end;
--// display the 1st column of a cursor, and we supply the cursor handle (number)
create or replace procedure DisplayCursor( c in out number ) is
        buffer  varchar2(4000);
        rc      number;
begin
        DBMS_SQL.define_column( c, 1, buffer, 4000 );
        DBMS_OUTPUT.put_line( '** displaying cursor '||c||' (1st column only)' );
        rc := DBMS_SQL.execute_and_fetch( c );
        loop
                DBMS_SQL.column_value( c, 1, buffer );
                DBMS_OUTPUT.put_line(  buffer );
                exit when DBMS_SQL.fetch_rows( c ) = 0;
        end loop;
        DBMS_OUTPUT.put_line( '** end of cursor display' );
        DBMS_SQL.close_cursor( c );
end;
/We can now create a cursor, get the handle and then later from another code unit (in the same session) use that cursor handle that was created. This means that the cursor existed from the time we've opened it, until we have explicitly closed it. In other words, a global cursor that exists for the duration of our session, or until we explicitly close the the cursor handle.
SQL> var c1 number
SQL> var c2 number
SQL>
SQL> --// we open a cursor
SQL> exec :c1 := CreateCursor( 'select object_name from user_objects where object_type =''TABLE'' and rownum <= 5' );
PL/SQL procedure successfully completed.
SQL> select 'Cursor '||:c1||' created.' as NOTE from dual;
NOTE
Cursor 710790198 created.
SQL> --// we now open a second cursor
SQL> exec :c2 := CreateCursor( 'select object_name from user_objects where object_type =''FUNCTION'' and rownum <= 5' );
PL/SQL procedure successfully completed.
SQL> select 'Cursor '||:c2||' created.' as NOTE from dual;
NOTE
Cursor 1228216362 created.
SQL>
SQL> --// we first display the 2nd cursor we've opened
SQL> exec DisplayCursor( :c2 );
** displaying cursor 1228216362 (1st column only)
CREATECURSOR
GETTOKEN
PLAINFOO
SPECIALFOO
TESTFUNC
** end of cursor display
PL/SQL procedure successfully completed.
SQL> --// we now display the 1st cursor we've opened
SQL> exec DisplayCursor( :c1 );
** displaying cursor 710790198 (1st column only)
EMPTAB
FOOTAB
FOOTAB_S
TESTRAW
TESTTAB
** end of cursor display
PL/SQL procedure successfully completed.
SQL> As you can see, these cursors are global. We create them using one unit of code. We use them via another unit of code. However, we cannot pass the DBMS_SQL handle (the number above) to our client session (e.g. Java, TOAD or SQL*Plus). We need to pass a ref cursor handle to such a client and not a DBMS_SQL handle. This allows our DisplayCursor() procedure we did in PL/SQL, to be written in client code like Java. Where the only difference is that this code will receive a ref cursor data type to process and not a DBMS_SQL cursor type.