Best Practice Exception Handling.

Question

Hi,
Please consider two scenarios:
Scenario 1:
DECLARE
  l_emp   scott.emp.ename%TYPE;
  l_dname scott.dept.dname%TYPE;
BEGIN
  BEGIN
    SELECT ename INTO l_emp FROM emp WHERE empno = 7934;
  EXCEPTION
    WHEN no_data_found THEN
      dbms_output.put_line('No big deal');
      NULL;
    WHEN too_many_rows THEN
      dbms_output.put_line('It is a big deal');
      RAISE;
    WHEN OTHERS THEN
      RAISE;
  END;
  BEGIN
    SELECT dname INTO l_dname FROM dept WHERE deptno = 1;
  EXCEPTION
    WHEN no_data_found THEN
      dbms_output.put_line('It is a big deal');
      RAISE;
    WHEN too_many_rows THEN
      dbms_output.put_line('It is a big deal');
      RAISE;
    WHEN OTHERS THEN
      RAISE;
  END;
EXCEPTION
  WHEN OTHERS THEN
    RAISE;
END;
Scenario 2:
DECLARE
  l_point_of_error NUMBER;
  l_emp            scott.emp.ename%TYPE;
  l_dname          scott.dept.dname%TYPE;
BEGIN
  l_point_of_error := 1;
  now write some implicit cursors-
  SELECT ename INTO l_emp FROM emp WHERE empno = 7934;
  l_point_of_error := 2;
  SELECT dname INTO l_dname FROM dept WHERE deptno = 1;
EXCEPTION
  WHEN no_data_found THEN
    CASE l_point_of_error
      WHEN 1 THEN
        dbms_output.put_line('No big deal');
        NULL;
      WHEN 2 THEN
        dbms_output.put_line('It is a big deal');
        RAISE;
      ELSE
        dbms_output.put_line('I have an idea which block this errored out on...but I want or HAVE TO raise the error');
        RAISE;
    END CASE;
    NULL;
  WHEN too_many_rows THEN
    CASE l_point_of_error
      WHEN 1 THEN
        dbms_output.put_line('It is a big deal');
        RAISE;
      WHEN 2 THEN
        dbms_output.put_line('It is a big deal');
        RAISE;
      ELSE
        dbms_output.put_line('I have an idea which block this errored out on...but I want or HAVE TO raise the error');
        RAISE;
    END CASE;
    NULL;
  WHEN OTHERS THEN
    RAISE;
END;
/What do you think is the right approach?
The one thing I can think of using Scenario 2, it will be a nightmare to handle the 'case' statements in the final exception catcher when the number of blocks to handle are more.
Also, Scenario 2 also uses ONE more variable to be assigned (more processing time maybe??).
I am also told that Scenario 2 is used extensively used by Oracle Applications PL/SQL Api's.
But, if you can suggest which do you think is the best practice is? I would really appreciate it.
Thank you,
Rahul

Answer 1

My bad 3360:
I didn't mention the error logging think in OP and I use it in my code all the time and forgot to put it in OP.
Anyway, I use Tom Kyte's who_am_i and who_called_me in my error logging to figure out the line number of the error.
* One failure was caused by the error log table running out of space because no one ever looked at it and didn't realize it had hundreds of thousands of exceptions in it. Yes, you are absolutely right. Yes somebody(the developer(s) ) have to take care of the error logging table (try to see what errors out and why).
P.S: I am gonna stick with Scenario 1 (with the error logging of course :) ) according to Todd Barry's comments and I have no idea why Oracle Applications uses Scenario 2. Can anybody tell me why (maybe it's just bad coding on Oracle Apps part?).
Thank you,
Rahul.