DISTINCT in the query -- RBO vs CBO

Question

Hi all,
We are in a process of migrating our applications from RBO to CBO as we moved to 10g and recently I observed a simple query with DISTINCT clause showing different results in CBO. Here is my order of creating a table in 10g environment,
create table m_test(col1 varchar2(3));
insert into m_test values('ZYT');
insert into m_test values('ABC');
insert into m_test values('RKT');
select * from m_test
col1
ZYT
ABC
RKT
alter session set optimizer_mode = RULE
select distinct col1 from m_test;
col1
ABC
RKT
ZYT
Note: Explain plan shows that it's doing SORT UNIQUE
alter session set optimizer_mode = FIRST_ROWS_1
select distinct col1 from m_test;
col1
ZYT
ABC
RKT
Note: Explain plan shows that it's doing HASH UNIQUE
I would appreciate if someone can let me know why this is happening,
Thank you,
Madhu

Answer 1

It happened to be sorted when you were using the RBO because the query plan the RBO generated happened to involve a sort. The CBO knows that sorting isn't strictly necessary, so it may or may not choose to apply the sort.
The CBO has a lot of options for generating query plans, depending on table, column, index, and system statistics, histograms, etc. and has a lot more plan options than the old RBO did. The CBO is a lot more likely to change a query plan than the RBO was, which is normally a good thing (a plan that works for a 10 row table might need to be changed if the table grows to 1 million rows). Depending on the Oracle version, initialization parameters, table statistics, system statistics, etc. the CBO may happen to sort rows before returning them.
If you want Oracle to sort the rows, however, there is only one way to guarantee that-- add an ORDER BY clause.
Justin