Function as column in select query

Question

Recently I came upon a couple of applications using Oracle database 10g that use functions as a column in a select query for reporting purposes. These queries were taking about 10 minutes to return 55,000 records. How can such queries be optimized where the functions in the queries cannot be modified?
Can the performance be improved at all without a re-design of the query?
SELECT
          col1
         ,col2
         ,fn_get_val(col3)
         ,col4
         ,col5
   FROM
          table a,
          table b
   WHERE .....; The above function "fn_get_val" is defined as:
CREATE OR REPLACE fn_get_val(i_col IN VARCHAR2)  return varchar2
as
       CURSOR c ( i_var  VARCHAR2 )
          business logic here again
begin
     OPEN c( i_col );
     LOOP
         Loop through the above cursor
         EXIT;  -- exit when condition satisfied.
    END LOOP;
    CLOSE c;
end;
/Thanks in advance!

Answer 1

user1980 wrote:
Can the performance be improved at all without a re-design of the query?Possibly if function is deterministic.
Simplistic example:
SQL> create or replace package Global as
  2          counter integer := 0;
  3  end;
  4  /
Package created.
SQL>
SQL> --// non-deterministic
SQL> create or replace function EmpName( empID number ) return varchar2 is
  2          cursor c is
  3                  select ename from emp where empno = empID;
  4          empName         varchar2(10);
  5  begin
  6          Global.counter := Global.counter + 1;
  7          --// simulating your explicit cursor loop
  8          open c;
  9          fetch c into empName;
10          close c;
11 
12          return(
13                  InitCap(empName)
14          );
15  end;
16  /
Function created.
SQL>
SQL>
SQL> exec Global.counter := 0;
PL/SQL procedure successfully completed.
SQL> select
  2          empName( 7369 ) as NAME
  3  from       dual
  4  connect by level <= 100;
NAME
Smith
Smith
100 rows selected.
SQL> exec dbms_output.put_line( 'Counter='||Global.counter );
Counter=100
PL/SQL procedure successfully completed.
SQL>
SQL> --// deterministic
SQL> create or replace function EmpName( empID number ) return varchar2 deterministic is
  2          cursor c is
  3                  select ename from emp where empno = empID;
  4          empName         varchar2(10);
  5  begin
  6          Global.counter := Global.counter + 1;
  7          --// simulating your explicit cursor loop
  8          open c;
  9          fetch c into empName;
10          close c;
11 
12          return(
13                  InitCap(empName)
14          );
15  end;
16  /
Function created.
SQL>
SQL> exec Global.counter := 0;
PL/SQL procedure successfully completed.
SQL> select
  2          empName( 7369 ) as NAME
  3  from       dual
  4  connect by level <= 100;
NAME
Smith
Smith
100 rows selected.
SQL> exec dbms_output.put_line( 'Counter='||Global.counter );
Counter=1
PL/SQL procedure successfully completed.
SQL> So 1 execution of the function versus a 100 executions. But it is not always that straightforward and simplistic.
The ideal would be to not use a user function in the first place - but to instead apply that function's logic directly in SQL using native SQL statements.