Re-write oracle query to avoid multiple table scan - sql

Question

I hope everyone is fine and learning more. I need some advice on select statement and on its fine tuning. I am using Oracle 11gR2. Please find below table and data scripts.
create table employee (emp_id number, emp_name varchar2(50), manager_id number);
create table department (dept_id number, dept_name varchar2(50), emp_name varchar2(50), manager_level varchar2(20));
create table manager_lookup (manager_level_id number, manager_level varchar2(20));
insert into employee values (1, 'EmpA',3);
insert into employee values (2, 'EmpB',1);
insert into employee values (3, 'EmpC',1);
insert into employee values (4, 'EmpD',2);
insert into employee values (5, 'EmpE',1);
insert into employee values (6, 'EmpF',3);
insert into department values (1, 'DeptA','EmpD','Level3');
insert into department values (2, 'DeptB','EmpC','Level2');
insert into department values (3, 'DeptC','EmpA','Level1');
insert into department values (4, 'DeptD','EmpF','Level1');
insert into department values (5, 'DeptD','EmpA','Level3');
insert into department values (6, 'DeptA',NULL,'Level3');
insert into manager_lookup values (1, 'Level1');
insert into manager_lookup values (2, 'Level2');
insert into manager_lookup values (3, 'Level3');
commit;
Below query is returning me dept_id by passing some emp_name. I need those dept_id where manager_level is same as emp_name passed but don't need to have that same emp_name in result data set.
SELECT b.dept_id
FROM (SELECT DISTINCT manager_level
FROM department dpt
WHERE emp_name = 'EmpA'
and emp_name is not null) a,
department b
WHERE a.manager_level = b.manager_level
AND NVL (b.emp_name, 'ABC') <> 'EmpA';
Above query is returning me data set as below:
dept_id
--------
1
4
6
I want same result set but need to rewrite above query in a way to avoid department table scan two times. This is just sample query but in real time scanning big table two times is giving performance issues. I want to re-write this query in a way to perform better and avoid same table scan two times.
Can you please help on providing your wonderful suggestions or solutions? I will really appreciate all responses.
Thank you for going over the question.

Answer 1

If you want your query to be more efficient, then use indexes:
create index idx_department_name_level on department(emp_name, manager_level)
and
create index idx_department_name_level on department(manager_level, emp_name, dept_id)

Answer 2

also, you have a redundant null check which may be avoiding indexes...
SELECT b.dept_id
FROM (SELECT manager_level
FROM department dpt
WHERE emp_name = 'EmpA') a,
department b
WHERE a.manager_level = b.manager_level
AND NVL (b.emp_name, 'ABC') <> 'EmpA';
post your explain plan for more help

Answer 3

This should work:
SELECT a.*
FROM
(
SELECT d.*,
SUM(CASE WHEN emp_name = 'EmpA' THEN 1 ELSE 0 END)
OVER (PARTITION BY manager_level) AS hits
FROM department d
) a
WHERE hits > 0
AND NVL(emp_name, 'Dummy') <> 'EmpA'
ORDER BY dept_id
;
The query does the following:
Calculate how many times EmpA appears in a given manager_level
Keep all records with a manager_level that has at least one occurrence of EmpA in it
Excluding the EmpA records themselves
SQL Fiddle of the query in action: http://sqlfiddle.com/#!4/a9e03/7
You can verify that the execution plan contains only one full table scan.