Is there a change in the way 10g sees the rowid column.

Question

i have two databases in my test environment
1. Oracle 10g on CentOS 5.3
2. Oracle 9i on Windows Server 2003
i am practicing simple SQL with the default HR schema created while creating databases using dbca.
all i am doing is trying to display the employee with the 10th highest salary.
in Oracle 10g i run the following command:
SQL> select first_name || ' ' || last_name, salary from employees where
rowid =
(select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum <=10
minus
select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum < 10);
FIRST_NAME||''||LAST_NAME                          SALARY
Lisa Ozer                                           11500when i run the same command in 9i
select first_name || ' ' || last_name, salary from employees where
rowid =
(select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum <=10
minus
select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum < 10)
no rows selectedon digging deeper i found out that in 9i .... when i run the two sub-queries, give the same rowid as output.
SQL> select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum <=10;
ROWID
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
10 rows selected.
SQL> select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum < 10;
ROWID
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
AAAHM7AAFAAAABXAAI
9 rows selected.the same two queries in 10g give different rowids (this is correct)
SQL> select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum <=10;
ROWID
AAAMg3AAFAAAABYAAA
AAAMg3AAFAAAABYAAB
AAAMg3AAFAAAABYAAC
AAAMg3AAFAAAABYAAt
AAAMg3AAFAAAABYAAu
AAAMg3AAFAAAABUAAD
AAAMg3AAFAAAABUAAH
AAAMg3AAFAAAABYAAI
AAAMg3AAFAAAABYAAv
AAAMg3AAFAAAABYABE
10 rows selected.
SQL> select rowid from (select first_name || ' ' || last_name from employees order by salary desc) where rownum < 10;
ROWID
AAAMg3AAFAAAABYAAA
AAAMg3AAFAAAABYAAB
AAAMg3AAFAAAABYAAC
AAAMg3AAFAAAABYAAt
AAAMg3AAFAAAABYAAu
AAAMg3AAFAAAABUAAD
AAAMg3AAFAAAABUAAH
AAAMg3AAFAAAABYAAI
AAAMg3AAFAAAABYAAv
9 rows selected.is there a change in the way Oracle 10g treats rowids from subqueries?
when i change the query to include the rowid in the innermost sub-query things work as they should in both 10g and 9i
select first_name || ' ' || last_name, salary from employees where
rowid =
(select rowid from (select rowid,first_name || ' ' || last_name from employees order by salary desc) where rownum <=10
minus
select rowid from (select rowid,first_name || ' ' || last_name from employees order by salary desc) where rownum < 10)

Answer 1

OK,
I didn't seem to have made myself clear. You state
'the above query uses rownum to filter the top 10 and top 9 salaries and picks the person with the 10th highest salary.'
As you are using row*id* , and rowids are random strings, and can change during the life of your table, you can not rely on the results of your query.
Your usage of the rownum pseudo function (as opposed to the rownum analytical function is also incorrect, as rownum is assigned prior to ordering it.
This
        select * from
         (select rownum rn, ename, sal
          from
          (select ename, sal from scott.emp order by sal desc)
          where rn=10is quite as sufficient, and doesn't use hacks like rowid.
Sybrand Bakker
Senior Oracle DBA
Experts: those who did read documentation.