Help on Procedure

Question

I have two tables.
ID columns exist in both tables and same record has the same id in both tables. Which mean if the record with ID = 3 in TABLE 1, the record with ID = 3 in TABLE 2 will represent the same record.
TABLE 1 has more data then TABLE 2. Which means TABLE 2 only contains part of the records. No duplicate records.
I want to update some of the columns in TABLE 1 by using TABLE 2 information.
Both tables contain more then 50 million records.
TABLE 1
ID                    name                    title                    level                    office                    location
1                     John                     VP                      9                        18133                        SH
2                     Steve                    MG                      6                        16231                        SA
3                     Jane                                             3                        14267                      
4                                                                      4                        208                           PB
5                                              SL                      6                        192                           BZ
6                     John                     PM                                               133                          SH
7                     Smith                    PM                      2                        183                          SH
8                     Joe                                              2                        18133                        SH
TABLE 2
ID                    name                    title                    level                    office                    location
3                     Jane                                             3                        247                      
6                     John                     PM                      3                        153                          SH
1                     John                     VP                      6                        18133                        SH
4                                                                      4                        208                           PB
2                     Steve                    MG                      4                        16571                        SA
8                     Joe                      TL                      2                        18003                        SH
5                                              SL                      7                        192                           BZ
7                     Smith                    PM                      4                        196                          SH
.I want to use update TABLE 1 by using TABLE2 (Column "level", "office" & "title"). Here is the procedure.
CREATE OR REPLACE PROCEDURE PROC_4 IS
step_id    CONSTANT NUMBER(10) := 100000;
maximum_id          NUMBER(10) := 75000000;
low_id              NUMBER(10) := 1;
high_id             NUMBER(10);
BEGIN
high_id := low_id + step_id;
DBMS_OUTPUT.PUT_LINE('LOW ID = ' || TO_CHAR(low_id) || ', HIGH ID = ' || TO_CHAR(high_id) || ', MAXIMUM ID = ' || TO_CHAR(maximum_id));
WHILE maximum_id > low_id LOOP
UPDATE TABLE_1 A
SET
    A.TITEL = (SELECT B.TITLE FROM TABLE_2 B WHERE A.ID = B.ID),
    A.OFFICE = (SELECT B.OFFICE FROM TABLE_2 B WHERE A.ID = B.ID),
    A.LEVEL = (SELECT B.LEVEL FROM TABLE_2 B WHERE A.ID = B.ID)
WHERE EXISTS (SELECT B.ID FROM TABLE_2 B WHERE A.ID = B.ID);
    COMMIT;
    low_id  := high_id;
    high_id := high_id + step_id;
    DBMS_OUTPUT.PUT_LINE('LOW ID = ' || TO_CHAR(low_id) || ',HIGH ID = ' || TO_CHAR(high_id));
END LOOP;
END PROC_4;
/I tried this, but it is kind of slow, can anyone help me to speed it up?
PS, these two tables are huge.
Thanks at advance.
Edited by: user644467 on Mar 10, 2011 3:02 PM
Edited by: user644467 on Mar 10, 2011 9:28 PM

Answer 1

user644467 wrote:
These are huge tables, my original purpose is to do it piece by piece (every 100000). I thought this will faster.Do what piece by piece? Faster than what? Your posted code will do the same thing 750 times for no clear reason.
My understanding is that your SQL will full table scan for every record, right?My version has two subqueries instead of the four in your version, and it also avoids 750 unnecessary commits. I'm not sure what you think full scans have to do with it.