oracle 11g pivot query optimization - multiple rows to single row - sql

Question

I have below tables
user table
USER_ID USER_NAME
1 smith
2 clark
3 scott
4 chris
5 john
property table
P_ID PROPERTY
1 first_name
2 last_name
3 age
4 skill
user_property table
PV_ID USER_ID P_ID VALUE
1 1 1 Smith
2 1 2 A
3 1 3 34
4 1 4 Java
5 1 4 DB
6 2 1 Clark
7 2 2 B
8 2 3 39
9 2 4 Java
10 2 4 net
11 2 4 linux
12 3 1 Scott
13 3 2 C
14 3 3 31
I want to write a query which will fetch data from all above tables as below:(Skill will be first skill for that user if available otherwise null)
USER_ID USER_NAME FIRST_NAME LAST_NAME SKILL
1 smith Smith A Java
2 clark Clark B Java
3 scott Scott C null
I have tried like below but getting performance issue:
SELECT
u.user_id,
u.user_name,
MAX(DECODE(p.property, 'first_name', text_value)) firstName,
MAX(DECODE(p.property, 'last_name', text_value)) lastName,
MAX(DECODE(p.property, 'age', text_value)) age,
MAX(DECODE(p.property, 'skill', text_value)) skill
FROM user u,
property p,
user_property up,
WHERE u.user_id = up.user_id
AND p.p_id = up.p_id
GROUP BY u.user_id,
u.user_name;
How could i write this as optimized query for oracle 11g.

Answer 1

The performance of your query depends on the size of the table and on the indexes you have on these tables. It most cases, it is best practice to have an index on each primary and foreign key. The index on the primary key is a must anyway. The index on the foreign key speeds up joins and prevents table locks when you delete rows.
An alternative to your query would be to use more joins instead of subselects and to use the WITH clause to simplify it:
with t as (
select u.user_id, u.user_name, up.p_id, up.value
from user_property up
join user u on u.user_id = up.user_id
)
select u.user_id, u.user_name,
t_first_name.value first_name,
t_last_name.value last_name,
(select min(value) from t where t.user_id = u.user_id and t.p_id = 4) skill
from user u
left join t t_first_name on t_first_name.user_id = u.user_id and t_first_name.p_id = 1
left join t t_last_name on t_last_name.user_id = u.user_id and t_last_name.p_id = 2;
BTW: It's a data model which isn't well suited for SQL. I hope these user properties are the exception and the rest of the database has a cleaner design.

Answer 2

I have tried below query however getting Cartesian product.
with t as (
select u.user_id, u.user_name, up.p_id, up.value
from user_property up
join user u on u.user_id = up.user_id
where u.user_name = 'smith'
)
select u.user_id, u.user_name,
t_first_name.value first_name,
t_last_name.value last_name,
(select min(value) from t where t.user_id = u.user_id and t.p_id = 4) skill
from user u
left join t t_first_name on t_first_name.user_id = u.user_id and t_first_name.p_id = 1
left join t t_last_name on t_last_name.user_id = u.user_id and t_last_name.p_id = 2;
if i execute below query i get 5 as mentioned in my above example (As there are 5 rows in my user_property table for user_id 1 in example)
select count(u.user_id)
from user_property up
join user u on u.user_id = up.user_id
where u.user_name = 'smith'
Hence if i execute below query i get count as 3 as there are 3 rows in my use table example
with t as (
select u.user_id, u.user_name, up.p_id, up.value
from user_property up
join user u on u.user_id = up.user_id
where u.user_name = 'smith'
)
select count(u.user_id)
from user u
left join t t_first_name on t_first_name.user_id = u.user_id and t_first_name.p_id = 1
left join t t_last_name on t_last_name.user_id = u.user_id and t_last_name.p_id = 2;