Updating billions of rows from historical data

Question

Hi,
we have a task of updating billions of rows from some tables ( one at a time ) for some past DATE's.
Could you please suggest what would be the best approach to follow in such cases where simple UPDATE statements would take dont know how much time.
The scenario is something like this..
TEST table.
col1 col2 col3
Now we added one more column col4.
Now all the records in this table needs to be updated for col3=col4 values for past data ( past in the sense the time when this will be executed in the Prodn DB, the sysdate-1 will be past data)
I thought of using FORALL kind of clause but still I dont know if that is going to be useful completely or not.
I am using Oracle 10g rel 2 version
Plase give me your expert opinions on this scenario so that the script to update such a voluminous data can execute successfully with better performance.
Thanks,
Aashish

Answer 1

Hi Mohamed,
thanks for your help.
However, in my case, its not possible to drop and recreate the table only to update single column for values in another column.
I am trying to do some POC for this using BULK UPDATE but I am not able to do so...
I did something like this to check for 50,00,000 records.
create table test
( col1 varchar2(100),
  col2 varchar2(100));
inserted 50,00,000 records to col1 column of this table.Now when I tried to do something like
declare
CURSOR s_cur IS
SELECT col1
FROM test;
TYPE fetch_array IS TABLE OF s_cur%ROWTYPE;
s_array fetch_array;
begin
OPEN s_cur;
  LOOP
    FETCH s_cur BULK COLLECT INTO s_array LIMIT 10000;
    FORALL i IN 1..s_array.COUNT
    UPDATE TEST
    SET col2 = s_array(i); -- dont know if this is correct
    EXIT WHEN s_cur%NOTFOUND;
  END LOOP;
  CLOSE s_cur;
  COMMIT;
END;It gives me some error.
Can you please correct me in this?
Still I am looking for better way...
Rgds,
Aashish