Trying to optimize the following update query

Question

I am trying to update a table based on the values of another table in the following manner. I have two tables A and B. Say A has 300,000 records and B has 100,000 which is basically a set of the 300,000 in A. Therefore, B has the exact same records as A. There are two columns in both tables (status and my_status). Currently for all the records in A these two values are 0 and null respectively. In B, the same records have different values for the two columns. Table A needs to updated with these values currently in B. I have the following the query that I am hesitant to use since the explain shows a very high cost and a full table scan of B.
update A a
set (status,my_status) = (select b.status, b.my_status from B b, A a where a.id = b.id)
where a.date >= '01-JAN-2003' and a.cd = 'FD'.
As the above query shows, the where condition in the outer part(where a.date >= '01-JAN-2003' and a.cd = 'FD'.) ensures that only those records present in B are updated in A. Is there any way to join at the outer part where I can just specify A.id = B.id rather than having two conditions? Or is there any other route that would optimize this sort of a query?

Answer 1

Hi,
Be sure to put unique constraints on A.ID and B.ID like so before running the UPDATE view statement, like this:
ALTER TABLE a ADD CONSTRAINT a_uk1 UNIQUE (id);
ALTER TABLE b ADD CONSTRAINT b_uk1 UNIQUE (id);Then, remember to gather stats like so:
BEGIN
   FOR x IN (SELECT table_name
             FROM user_tables
             WHERE table_name IN ('A', 'B')) LOOP
     DBMS_STATS.gather_table_stats (ownname               => USER
                                  , tabname               => x.table_name
                                  , partname              => NULL
                                  , estimate_percent      => DBMS_STATS.auto_sample_size
                                  , block_sample          => FALSE
                                  , method_opt            => 'FOR ALL INDEXED COLUMNS SIZE 254'
                                  , degree                => NULL
                                  , granularity           => 'ALL'
                                  , cascade               => TRUE
                                  , no_invalidate         => FALSE
   END LOOP;
END;
/Good luck!