Understanding Multi-Row Subqueries in Oracle 10g - oracle

Question

I executed a SQL statement and come across a mess. I am not able to understand how this output is coming.
My employee table is: Emp_Id is primary key and dept_no is a foreign key to some other table.
EMP_ID EMP_NAME DEPT_NO MGR_NAME MGR_NO
---------- -------------------- ---------- ---------- -----------
111 Anish 121 Tanuj 1123
112 Aman 122 Jasmeet 1234
1123 Tanuj 122 Vipul 122
1234 Jasmeet 122 Anish 111
122 Vipul 123 Aman 112
100 Chetan 123 Anoop 666
101 Antal Aman
1011 Anjali 126
1111 Angelina 127
My dep1 table is:
DEPT_ID DEPT_NAME
---------- -------------
121 CSE
122 ECE
123 MEC
And the two tables are not linked at all.
The SQL Query is:
SQL> select emp_name
from employee
where dept_no IN (select dept_no from dep1 where dept_name='MEC');
And the output is:
EMP_NAME
--------------------
Anish
Aman
Tanuj
Jasmeet
Vipul
Chetan
Anjali
Angelina
8 rows selected.
And if I change the where condition to dept_name='me' it returns no rows.
Can someone explain why the execution is not generating an error since dept_no is not the column of dep1 table. And how the output is being generated.

Answer 1

if you run this query:
select emp_name
from employee
where dept_no IN (select t.dept_no from dep1 t where dept_name='MEC');
you will see the error there for in your query dept_no comes from employee table (not from dep1 table)and when dept_no is null ,no result will be come back from it and if you change your dept_name to something which is not in dep1 table it is clear that your dep1 table returns nothing and then dept_no cant be in nothing.

Answer 2

As from your query,
..where dept_no IN (select dept_no ...); -- it is similar as using EXISTS
The condition EXISTS is being done here: ( oracle won't return error for EXISTS clause).
CREATE TABLE my_test(ID INT);
CREATE TABLE my_new_test ( new_ID INT);
EXPLAIN PLAN FOR
select * from my_test where id in( select id from my_new_test);
select * from table(dbms_xplan.display);
-----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 13 | 4 (0)| 00:00:01 |
|* 1 | FILTER | | | | | |
| 2 | TABLE ACCESS FULL | MY_TEST | 1 | 13 | 2 (0)| 00:00:01 |
|* 3 | FILTER | | | | | |
| 4 | TABLE ACCESS FULL| MY_NEW_TEST | 1 | | 2 (0)| 00:00:01 |
-----------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter( EXISTS (SELECT 0 FROM "MY_NEW_TEST" "MY_NEW_TEST" WHERE
:B1=:B2))
3 - filter(:B1=:B2)
Note
-----
- dynamic sampling used for this statement (level=2)
If you execute the plan for the valid columns , here ( new_id):
then normal access is done:
1 - ACCESS("ID"="NEW_ID")
And will cause and error for the following:
EXPLAIN PLAN FOR
select * from my_test where id in( select some_thing from my_new_test);
SQL Error: ORA-00904: "SOME_THING": invalid identifier

Answer 3

Let me try to answer it.
Oracle uses optimizer to decide the explain plan. Oracle may rewrite your query as it likes and thinks which one is better. And in and exists are interchangeable and performance depend on different things. (exists ends in full table scan and in uses index).
Let me take your case. Below is the explain plan for your query
Plan hash value: 3333342911
----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 225 | 6 (0)| 00:00:01 |
|* 1 | FILTER | | | | | |
| 2 | TABLE ACCESS FULL | EMPLOYEE | 9 | 225 | 3 (0)| 00:00:01 |
|* 3 | FILTER | | | | | |
|* 4 | TABLE ACCESS FULL| DEPARTMENT | 1 | 12 | 3 (0)| 00:00:01 |
----------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter( EXISTS (SELECT 0 FROM "DEPARTMENT" "DEPARTMENT" WHERE
:B1=:B2 AND "DEPT_NAME"='MEC'))
3 - filter(:B1=:B2)
4 - filter("DEPT_NAME"='MEC')
Note
-----
- dynamic sampling used for this statement (level=2)
This explain plan clearly shows that the query is re-written to use exists and it is equivalent to
select emp_name from employee where exists (select 0 from department where dept_name = 'MEC' and dept_no = dept_no);
The above query is a valid query and you are getting the right results.
Bind variables are nothing but dept_no (joining column).
Refer this IN vs EXISTS in oracle link to know more about in and exists.
Your explain plan is completely different if you use the correct column name. Below is the query and explain plan
Query:
select emp_name from employee where dept_no IN (select dept_id from department where dept_name='MEC');
Explain Plan:
Plan hash value: 3817251802
---------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2 | 100 | 7 (15)| 00:00:01 |
|* 1 | HASH JOIN SEMI | | 2 | 100 | 7 (15)| 00:00:01 |
| 2 | TABLE ACCESS FULL| EMPLOYEE | 9 | 225 | 3 (0)| 00:00:01 |
|* 3 | TABLE ACCESS FULL| DEPARTMENT | 1 | 25 | 3 (0)| 00:00:01 |
---------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("DEPT_NO"="DEPT_ID")
3 - filter("DEPT_NAME"='MEC')
Note
-----
- dynamic sampling used for this statement (level=2)
Oracle thinks it is better to use the filter and hash join to get the required details.
This behaviour depends on the oracle query parser and optimizer.

Answer 4

How about a join?
select a.emp_name
from employee a
join dep1 b
on a.dept_no = b.dept_id
where b.dept_name = 'MEC'