How to make sql to use index/make to query to perform better

Question

Hi,
I have 2 sql query which results the same.
But both has difference in SQL trace.
create table test_table
(u_id number(10),
u_no number(4),
s_id number(10),
s_no number(4),
o_id number(10),
o_no number(4),
constraint pk_test primary key(u_id, u_no));
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030301, 1, 1001, 1, 2001, 1);
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030302, 1, 1001, 1, 2001, 2);
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030303, 1, 1001, 1, 2001, 3);
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030304, 1, 1001, 1, 2001, 4);
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030305, 1, 1002, 1, 1001, 2);
insert into test_table(u_id, u_no, s_id, s_no, o_id, o_no)
values (2007030306, 1, 1002, 1, 1002, 1);
commit;
CREATE INDEX idx_test_s_id ON test_table(s_id, s_no);
set autotrace on
select s_id, s_no, o_id, o_no
from test_table
where s_id <> o_id
and s_no <> o_no
union all
select o_id, o_no, s_id, s_no
from test_table
where s_id <> o_id
and s_no <> o_no;
Execution Plan
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=6 Card=8 Bytes=416)
1 0 UNION-ALL
2 1 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=4 Bytes=208)
3 1 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=4 Bytes=208)
Statistics
223 recursive calls
2 db block gets
84 consistent gets
0 physical reads
0 redo size
701 bytes sent via SQL*Net to client
508 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
5 sorts (memory)
0 sorts (disk)
8 rows processed
-- i didnt understand why the above query is not using the index idx_test_s_id.
-- But still it is faster
select s_id, s_no, o_id, o_no
from test_table
where (u_id, u_no) in
(select u_id, u_no from test_table
minus
select u_id, u_no from test_table
where s_id = o_id
or s_no = o_no)
union all
select o_id, o_no, s_id, s_no
from test_table
where (u_id, u_no) in
(select u_id, u_no from test_table
minus
select u_id, u_no from test_table
where s_id = o_id
or s_no = o_no);
Execution Plan
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=16 Card=2 Bytes=156)
1 0 UNION-ALL
2 1 FILTER
3 2 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=6 Bytes=468)
4 2 MINUS
5 4 INDEX (UNIQUE SCAN) OF 'PK_TEST' (INDEX (UNIQUE)) (Cost=1 Card=1 Bytes=26)
6 4 TABLE ACCESS (BY INDEX ROWID) OF 'TEST_TABLE' (TABLE) (Cost=2 Card=1 Bytes=78)
7 6 INDEX (UNIQUE SCAN) OF 'PK_TEST' (INDEX (UNIQUE)) (Cost=1 Card=1)
8 1 FILTER
9 8 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=6 Bytes=468)
10 8 MINUS
11 10 INDEX (UNIQUE SCAN) OF 'PK_TEST' (INDEX (UNIQUE)) (Cost=1 Card=1 Bytes=26)
12 10 TABLE ACCESS (BY INDEX ROWID) OF 'TEST_TABLE' (TABLE) (Cost=2 Card=1 Bytes=78)
13 12 INDEX (UNIQUE SCAN) OF 'PK_TEST' (INDEX (UNIQUE)) (Cost=1 Card=1)
Statistics
53 recursive calls
8 db block gets
187 consistent gets
0 physical reads
0 redo size
701 bytes sent via SQL*Net to client
508 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
4 sorts (memory)
0 sorts (disk)
8 rows processed
-- The above query is using index PK_TEST. But still it has FULL SCAN to the
-- table two times it has the more cost.
1st query --> SELECT STATEMENT Optimizer=ALL_ROWS (Cost=6 Card=8 Bytes=416)
2nd query --> SELECT STATEMENT Optimizer=ALL_ROWS (Cost=16 Card=2 Bytes=156)
My queries are:
1) performance wise which query is better?
2) how do i make the 1st query to use an index
3) is there any other method to get the same result by using any index
Appreciate your immediate help.
Best regards
Muthu

Answer 1

Hi William
Nice...it works.. I have added "o_id" and "o_no" are in part of the index
and now the query uses the index
Execution Plan
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=2 Card=8 Bytes=416)
1 0 UNION-ALL
2 1 INDEX (FULL SCAN) OF 'IDX_TEST_S_ID' (INDEX) (Cost=1 Card=4 Bytes=208)
3 1 INDEX (FULL SCAN) OF 'IDX_TEST_S_ID' (INDEX) (Cost=1 Card=4 Bytes=208)
Statistics
7 recursive calls
0 db block gets
21 consistent gets
0 physical reads
0 redo size
701 bytes sent via SQL*Net to client
507 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
8 rows processed
But my questions are:
1) In a where clause, if "<>" condition is used, then, whether the system will use the index. Because I have observed in several situations even though the column in where clause is indexed, since the where condition is "like" or "is null/is not null"
then the index is not used. Same as like this, i assumed, if we use <> then indexes will not be used. Is it true?
2) Now, after adding "o_id" and "o_no" columns to the index, the Execution plan is:
Execution Plan
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=2 Card=8 Bytes=416)
1 0 UNION-ALL
2 1 INDEX (FULL SCAN) OF 'IDX_TEST_S_ID' (INDEX) (Cost=1 Card=4 Bytes=208)
3 1 INDEX (FULL SCAN) OF 'IDX_TEST_S_ID' (INDEX) (Cost=1 Card=4 Bytes=208)
Before it was :
Execution Plan
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=6 Card=8 Bytes=416)
1 0 UNION-ALL
2 1 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=4 Bytes=208)
3 1 TABLE ACCESS (FULL) OF 'TEST_TABLE' (TABLE) (Cost=3 Card=4 Bytes=208)
Difference only in Cost (reduced), not in Card, Bytes.
Can you explain, how can i decide which makes the performace better (Cost / Card / Bytes). Full Scan / Range Scan?
On statistics also:
Before:
Statistics
52 recursive calls
0 db block gets
43 consistent gets
0 physical reads
0 redo size
701 bytes sent via SQL*Net to client
507 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
8 rows processed
After:
Statistics
7 recursive calls
0 db block gets
21 consistent gets
0 physical reads
0 redo size
701 bytes sent via SQL*Net to client
507 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
8 rows processed
Difference in recursive calls & consistent gets.
Which one shows the query with better performance?
Please explain..
Regards
Muthu