Trouble with Bulk Collect to Ref cursor

Question

I'm trying to open a ref cursor to a dynamic query, and the fetch the cursor (BULK COLLECT)to the table type variable.But I keep getting the compilation error as 'PLS-00597: expression 'EMP_RESULTSET' in the INTO list is of wrong type'
But when I use a simple select from a table and Bulk Collect directly to the table type variable it works. But that is not what I want.
Can someone tell me where I have gone wrong in this stored proc I have listed below.
your help will be highly appreciated.
PROCEDURE SP_TEST_EMP_TABLE_TYPE (
      p_resultset OUT result_cursor          -- ref cursor as out parameter
   AS
    TYPE TYPE_REC_EMP is RECORD(EMPNO EMPLOYEE.EMPNO%TYPE, JOIN_DATE EMPLOYEE.JOIN_DATE%TYPE, SALARY EMPLOYEE.SALARY%TYPE); -- declare record type
    TYPE TYPE_TAB_EMP IS TABLE OF TYPE_REC_EMP;    -- declare table type
    EMP_RESULTSET TYPE_TAB_EMP; -- declare variable of type type_calendar_avail_resultset  
    SQUERY VARCHAR2(32000):='';
   BEGIN
    SQUERY:='SELECT EMPNO,JOIN_DATE,SALARY FROM EMPLOYEE WHERE EMPNO= 1001 AND JOIN_DATE=''20070530'' ';
    --select EMPNO,JOIN_DATE,SALARY BULK COLLECT INTO EMP_RESULTSET from  EMPLOYEE WHERE EMPNO=1001 AND JOIN_DATE='20070525';
    OPEN p_resultset FOR SQUERY;
      FETCH p_resultset BULK COLLECT INTO EMP_RESULTSET;
  EXCEPTION
      WHEN NO_DATA_FOUND
      THEN
         NULL;
      WHEN OTHERS
      THEN
         DBMS_OUTPUT.put_line (SQLERRM (SQLCODE));   
   END SP_TEST_EMP_TABLE_TYPE ;

Answer 1

> i) I use a ref cursor to return back to the java
front end, so I had to use a ref cursor.
What is a ref cursor? It is not a data set. It is a pointer to a "SQL program". This program is created by the SQL Engine and the CBO that parses the SQL source code and determine an execution plan to get to the requested rows.
When the client fetches from a (ref) cursor, the client is running this program and it find and returns the next row.
There is no such thing as a physical result set for a cursor - no "copy" of the rows found for the source code SQL is made as a result set.
> ii) I also use a dynamic sql, but I was thinking it
wasn't useful for this posting, so tried to write a
simple sql
What is dynamic SQL? SQL where object names (e.g name of the table) is only known at run-time. Or where the filter (e.g. WHERE conditions) can only be determined at run time.
If these are known, the SQL is static SQL.
For both static and dynamic SQL, bind variables are critical. It is the biggest performance mistake (in Oracle) to hardcode values and literals into a SQL.
> ii) I use a Bulk Collect to the table type
collection, since I use a loop, for which I had to
collect the results from each loop and finally send
the resultset thru the ref cursor.
Impossible. Nor does it make any sense. As stated, a ref cursor is a program and not a result set.
What you intend to do is run a SQL against data. Copy this data into private/local PL/SQL memory. Construct another SQL against this data - which means that it needs to be shipped back to the SQL engine as it cannot use/read local PL/SQL memory structures. And the pass that SQL cursor to the client.
What for?
> I had earlier used the logic to for this using a
temporary table, which works perfectly fine, but our
DBA says we should avoid temporary tables since it
makes additional read/write to the disk. This is the
reason I'm using table type collection here.
Your DBA is correct. One should so a single pass through a data set. Which is why simply passing a ref cursor for a SQL statement to the client is the simplest.
It makes no sense copying SQL data into a collection and then copying that back into the SQL engine in order to throw a ref cursor around it.
Basic client-server fundamentals.
And why RTFM the Oracle manuals I've indicated is important. You need to understand the memory and processing architectures in Oracle in order to make an informed and correct decision.