SQL Multiple Minus vs Multiple Join Performance - sql

Question

I hope someone can explain the performance of joining multiple tables vs. using MINUS to eliminate records. I looked at a few other stack overflow questions but didn't see what I was looking for.
I thought these two queries would produce the same output, and I have always heard "use joins, use joins!", particularly from stackoverflow posts, that they were expected to be faster...
This is the first query I ran which I thought would be much slower, but it takes only a matter of minutes to run...
select some_id
from table1
MINUS
select some_id
from table2
where table2.value = 'some_value'
MINUS
select some_id
from table3
where table3.value = 'some_value'
group by some_id
This is the second query which I thought would be faster, but it has been running for over 3 hours now (with no end in sight?)
select some_id
from table1
join table2 on table1.id=table2.id
join table3 on table1.id=table3.id
where table2.value = 'some_value'
or table3.value = 'some_value'
group by some_id
I should note all 3 tables have > 1 Million records, up to 15 Million records each.
EDIT:
Sorry - I meant to let you know I was avoiding the use of NOT EXISTS in this question as a response, as I really am curious about just these two scenarios.

Answer 1

Try this version:
select some_id
from table1
where not exists (select 1 from table2 t2 on t1.id = t2.id and t2.value = 'some_value') or
not exists (select 1 from table3 t3 on t1.id = t3.id and t3.value = 'some_value')
For best performance, you want indexes on table2(id, value) and table3(id, value).

Answer 2

Firstly make sure you have the indexes in place,
to see the plan, if it is making use of full table scan, the go ahead with the creating of indexes else it is going to take a long , long time.
if you have plsql developer, then paste the query in the in sql window and press F5 it would give you the explain plan .
or can do this also,
SCOTT#research 17-APR-15> EXPLAIN PLAN FOR
2 select empno
3 from emp
4 MINUS
5 select empno
6 from empp
7 where empp.empno = '7839'
8 MINUS
9 select empno
10 from emppp
11 where emppp.empno = '7902'
12 group by empno
13 ;
Explained.
SCOTT#research 17-APR-15> SET LINESIZE 130
SCOTT#research 17-APR-15> SET PAGESIZE 0
SCOTT#research 17-APR-15> SELECT *
2 FROM TABLE(DBMS_XPLAN.DISPLAY);
Plan hash value: 4222598102
---------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 14 | 82 | 10 (90)| 00:00:01 |
| 1 | MINUS | | | | | |
| 2 | MINUS | | | | | |
| 3 | SORT UNIQUE NOSORT | | 14 | 56 | 2 (50)| 00:00:01 |
| 4 | INDEX FULL SCAN | PK_EMP | 14 | 56 | 1 (0)| 00:00:01 |
| 5 | SORT UNIQUE NOSORT | | 1 | 13 | 4 (25)| 00:00:01 |
|* 6 | TABLE ACCESS FULL | EMPP | 1 | 13 | 3 (0)| 00:00:01 |
| 7 | SORT UNIQUE NOSORT | | 1 | 13 | 4 (25)| 00:00:01 |
| 8 | SORT GROUP BY NOSORT| | 1 | 13 | 4 (25)| 00:00:01 |
|* 9 | TABLE ACCESS FULL | EMPPP | 1 | 13 | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
6 - filter("EMPP"."EMPNO"=7839)
9 - filter("EMPPP"."EMPNO"=7902)
Note
-----
- dynamic sampling used for this statement (level=2)
26 rows selected.
Execution Plan
----------------------------------------------------------
Plan hash value: 2137789089
---------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 8168 | 16336 | 29 (0)| 00:00:01 |
| 1 | COLLECTION ITERATOR PICKLER FETCH| DISPLAY | 8168 | 16336 | 29 (0)| 00:00:01 |
---------------------------------------------------------------------------------------------
or if you want to use autotrace then do,
set autotrace on explain
This is how it would look,
SCOTT#research 17-APR-15> select empno
2 from emp
3 MINUS
4 select empno
5 from empp
6 where empp.empno = '7839'
7 MINUS
8 select empno
9 from emppp
10 where emppp.empno = '7902'
11 group by empno
12 ;
EMPNO
----------
234
7499
7521
7566
7654
7698
7782
7788
7844
7876
7900
7934
12 rows selected.
Execution Plan
----------------------------------------------------------
Plan hash value: 4222598102
---------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 14 | 82 | 10 (90)| 00:00:01 |
| 1 | MINUS | | | | | |
| 2 | MINUS | | | | | |
| 3 | SORT UNIQUE NOSORT | | 14 | 56 | 2 (50)| 00:00:01 |
| 4 | INDEX FULL SCAN | PK_EMP | 14 | 56 | 1 (0)| 00:00:01 |
| 5 | SORT UNIQUE NOSORT | | 1 | 13 | 4 (25)| 00:00:01 |
|* 6 | TABLE ACCESS FULL | EMPP | 1 | 13 | 3 (0)| 00:00:01 |
| 7 | SORT UNIQUE NOSORT | | 1 | 13 | 4 (25)| 00:00:01 |
| 8 | SORT GROUP BY NOSORT| | 1 | 13 | 4 (25)| 00:00:01 |
|* 9 | TABLE ACCESS FULL | EMPPP | 1 | 13 | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
6 - filter("EMPP"."EMPNO"=7839)
9 - filter("EMPPP"."EMPNO"=7902)
Note
-----
- dynamic sampling used for this statement (level=2)
SCOTT#research 17-APR-15>
SCOTT#research 17-APR-15> select emp.empno
2 from emp
3 join empp on emp.empno=empp.empno
4 join emppp on emp.empno=emppp.empno
5 where empp.empno = '7839'
6 or emppp.empno = '7902'
7 group by emp.empno
8 ;
EMPNO
----------
7839
7902
Execution Plan
----------------------------------------------------------
Plan hash value: 1435156579
-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 30 | 8 (25)| 00:00:01 |
| 1 | HASH GROUP BY | | 1 | 30 | 8 (25)| 00:00:01 |
|* 2 | HASH JOIN | | 1 | 30 | 7 (15)| 00:00:01 |
| 3 | NESTED LOOPS | | 6 | 102 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL| EMPPP | 6 | 78 | 3 (0)| 00:00:01 |
|* 5 | INDEX UNIQUE SCAN| PK_EMP | 1 | 4 | 0 (0)| 00:00:01 |
| 6 | TABLE ACCESS FULL | EMPP | 10 | 130 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("EMP"."EMPNO"="EMPP"."EMPNO")
filter("EMPP"."EMPNO"=7839 OR "EMPPP"."EMPNO"=7902)
5 - access("EMP"."EMPNO"="EMPPP"."EMPNO")
Note
-----
- dynamic sampling used for this statement (level=2)