update using for loop in plsql

Question

i'm having problem updating and insert into below column. Please advise on this.

This is the input

depnto   extra    comm
----------------------------
20       300      NULL
20       300      400
20       NULL     NULL
20       500      NULL

This is the expected output

depnto  Extra    comm
---------------------
20      300      300
20      300      400
20      NULL     NULL           
20      500      500

I need to update comm column with extra column on below conditions.

• If comm Is null then extra value is updated to comm.
• If comm Is not null, no need to update,
• If both are null, leave as null,
• if comm column has a value no need to overwrite.

My program is below. Even I need to keep track which are rows are updated and to which value in another table.

PROCEDURE (dept_id )
AS
BEGIN
   FOR r IN (SELECT *
               FROM emp
              WHERE comm IS NULL AND extra IS NOT NULL AND deptno = dept_id)
   LOOP
      UPDATE emp
         SET comm = extra
       WHERE comm IS NULL AND extra IS NOT NULL AND deptno = dept_id;



      INSERT INTO changed_comm (deptno, oldval, newval)
           VALUES (dept_id, r.comm, r.extra);
   END LOOP;
EXCEPTION
   WHEN NO_DATA_FOUND
   THEN
      NULL;
END;

please provide some opinion on above. Its not inserting correctly.

Answer 1

You do not need FOR LOOP, just a single UPDATE does the work:

UPDATE emp
  SET comm = extra
WHERE comm IS NULL AND extra IS NOT NULL;

Here is a demo: http://www.sqlfiddle.com/#!4/aacc3/1

--- EDIT ----

I didn't notice, that in the expected output deptno 10 was updated to 20,
to update deptno an another query is needed:

UPDATE emp
   SET deptno = 20
WHERE deptno = 10;

---- EDIT -----

If you want to insert changed values to the other table, try a procedure with RETURNING..BULK COLLECT and FORALL:

CREATE OR REPLACE PROCEDURE pro_cedure( p_dept_id number  ) 
IS
      TYPE changed_table_type IS TABLE OF changed%ROWTYPE;
      changed_buff changed_table_type;
BEGIN
      SELECT deptno, comm, extra BULK COLLECT INTO changed_buff
      FROM emp
      WHERE comm IS NULL AND extra IS NOT NULL AND deptno = p_dept_id
      FOR UPDATE;
      UPDATE emp
      SET comm = extra
      WHERE comm IS NULL AND extra IS NOT NULL AND deptno = p_dept_id;
      FORALL i IN 1 .. changed_buff.count
        INSERT INTO changed VALUES changed_buff( i );
END;
/

The procedure should work if you are not going to process huge number of records in a one call (more than 1000 ... or maximum a few thousands). If one dept_id can contain ten thousands and more rows, then this procedure might be slow, becasue it will consume a huge amount of PGA memory. In such a case, an another approach with bulk collectiong in chunks is required.

-- EDIT --- how to store sequence values -------

I assume that the table changed has 4 columns, like this:

  CREATE TABLE "TEST"."CHANGED" 
   (    "DEPTNO" NUMBER, 
        "OLDVAL" NUMBER, 
        "NEWVAL" NUMBER, 
        "SEQ_NEXTVAL" NUMBER 
   ) ;

and we will store sequence values in the seq_nextval column.

In such a case the procedure might look like this:

create or replace 
PROCEDURE pro_cedure( p_dept_id number  ) 
IS
      TYPE changed_table_type IS TABLE OF changed%ROWTYPE;
      changed_buff changed_table_type;
BEGIN
      SELECT deptno, comm, extra, sequence_name.nextval 
        BULK COLLECT INTO changed_buff
        FROM emp
        WHERE comm IS NULL AND extra IS NOT NULL AND deptno = p_dept_id
        FOR UPDATE;
      UPDATE emp
        SET comm = extra
        WHERE comm IS NULL AND extra IS NOT NULL AND deptno = p_dept_id;
      FORALL i IN 1 .. changed_buff.count
        INSERT INTO changed VALUES changed_buff( i );
END;

--- EDIT --- version with cursor for small sets of data -----

Yes, for small sets of data bulk collecting doesn't give significant increase of the speed, and plain cursor with for..loop is sufficient in such a case.
Below is an example how tu use the cursor together with update, notice the FOR UPDATE clause, it is required when we plan to update a record fetched from the cursor using WHERE CURRENT OF clause.
This time a sequence value is evaluated within the INSERT statement.

create or replace 
PROCEDURE pro_cedure( p_dept_id number  ) 
IS
      CURSOR mycursor IS 
         SELECT deptno, comm, extra
         FROM emp
         WHERE comm IS NULL AND extra IS NOT NULL 
               AND deptno = p_dept_id
         FOR UPDATE;    
BEGIN
      FOR emp_rec IN  mycursor
      LOOP
         UPDATE emp 
            SET comm = extra
            WHERE CURRENT OF mycursor;
         INSERT INTO changed( deptno, oldval, newval, seq_nextval)
                VALUES( emp_rec.deptno, emp_rec.comm, 
                        emp_rec.extra, sequence_name.nextval );
      END LOOP;
END;