SQL%rowcount 导致 PL/SQL 性能下降

Question

I have a PL/SQL script in which I am updating some record and depending upon the result of this query I am updating another table.

In a loop I am executing below queries which can loop up to 100 000 records.

update employee set dept='department' where empid=a.id;

IF SQL%rowcount = 1 THEN
    -- updating other table
END IF;

Is there any alternative logic to achieve this?

Answer 1

Most probably there is another option.

From what you described, you're doing that in a loop which suggests that it affects only one row at a time (based on where clause and sql%rowcount = 1 ; if there were many rows, you'd check >= 1 ). Such an approach usually means row-by-row which is slow-by-slow.

See whether you can do that on sets , not single rows . Remove the loop entirely.

For example, if this is what you have now:

begin
  for cur_r in (select deptno, empno 
                from emp 
                where job = 'CLERK'
               ) 
  loop
    update emp set sal = sal + 100
      where empno = cur_r.empno;

    if sql%rowcount = 1 then
       update dept set
         dname = dname || 'x'
         where deptno = cur_r.deptno
    end if;
  end loop;
end;

rewrite it to

begin
  update emp e set
    e.sal = e.sal + 100
    where e.job = 'CLERK';           --> condition from the cursor FOR loop

  update dept d set
    d.dname = d.dname || 'x'
    where d.deptno in (select e.deptno           --> SELECT used in
                       from emp e                --> cursor FOR
                       where e.job = 'CLERK'     --> loop
                      );
end;

Answer 2

Use an SQL collection and the RETURNING ... [BULK COLLECT] INTO clause of the UPDATE statement:

CREATE TYPE int_array IS TABLE OF NUMBER(10,0);

Then if you were using:

BEGIN
  FOR a IN (
    SELECT id
    FROM   first_table
    WHERE  some_condition = 'MET'
  ) 
  LOOP
    update employee set dept='department' where empid=a.id;

    IF sql%rowcount = 1 THEN
       UPDATE other_table
       SET    something = 'X'
       WHERE  employee_id = a.id;
    END IF;
  END LOOP;
END;
/

Change it to:

DECLARE
  p_ids int_array;
BEGIN
  update employee
  set    dept = 'department'
  where  empid IN (
    SELECT id
    FROM   first_table
    WHERE  some_condition = 'MET'
  )
  RETURNING empid BULK COLLECT INTO int_array;

  UPDATE other_table
  SET    something = 'X'
  WHERE  employee_id MEMBER OF int_array;
END;
/

And you will have got rid of the loop entirely and directly use the primary keys that have been updated.