Oracle: How to use left outer join to get all entries from left table and satisfying the condition in Where clause - sql

Question

I have the tables below.
Client:
ID | clientName
--------------
1 A1
2 A2
3 A3
Order:
OrdID clientID status_cd
------------------------
100 1 DONE
101 1 SENT
102 3 SENT
Status:
status_cd status_category
DONE COMPL
SENT INPROG
I have to write a query to get all the clients and count of order against all of them, whether the client_id exists in Order table or not and has the orders with "COMPL" as status category.
In this case, I am using the query below but it's filtering out the clients which has no orders. I want to get all clients such that the expected result is as below.
Query:
select c.ID, count(distinct o.OrdID)
from client c, order o, status s
where c.ID=o.client_id(+)
and o.status_cd=s.status_cd where s.status_category='COMPL'
group by c.ID
Expected result:
C.ID count(distinct o.OrdID)
----------------------------
1 1
2 0
3 0
Can someone please help me with this? I know, in this case, left outer join is behaving like inner join when I am using where clause, but is there any other way to achieve the results above?

Answer 1

This can be dealt with a lot easier when using an explicit join operator:
select c.ID, count(distinct s.status_cd)
from client c
left join orders o on o.clientid = c.id
left join status s on s.status_cd = o.status_cd and s.status_category='COMPL'
group by c.ID;
The above assumes that orders.status_cd is defined as not null
Another option is to move the join between orders and status in a derived table:
select c.ID, count(distinct o.ordid)
from client c
left join (
select o.ordid
from orders o
join status s on s.status_cd = o.status_cd
where s.status_category='COMPL'
) o on o.clientid = c.id
group by c.ID;
The above "states" more clearly (at least in my eyes) that only orders within that status category are of interest compared to the first solution

Answer 2

As usual, there are lots of ways to express this requirement.
Try ANSI join people will hate me an vote down this answer ;) :
select c.ID, count(distinct o.OrdID)
from client c, order o, status s
where c.ID = o.client_id(+)
and o.status_cd = s.status_cd
and s.status_category='COMPL'
group by c.ID
;
or
select c.ID
, nvl((select count(distinct o.OrdID)
from order o, status s
where c.ID = o.client_id
and o.status_cd = s.status_cd
and s.status_category='COMPL'
), 0) as order_count
from client c
group by c.ID
;
or
with ord as
(select client_id, count(distinct o.OrdID) cnt
from order o, status s
where 1=1
and o.status_cd = s.status_cd
and s.status_category='COMPL'
group by client_id
)
select c.ID
, nvl((select cnt from ord o where c.ID = o.client_id ), 0) as order_count
from client c
group by c.ID
;
or
...

Answer 3

The second WHERE should be an AND.
Other than that, you need the plus sign, (+), marking left outer join, in the second join condition as well. It is not enough to left-outer-join the first two tables.
Something like
select c.ID, count(distinct o.OrdID)
from client c, order o, status s
where c.ID=o.client_id(+)
and o.status_cd=s.status_cd(+) AND s.status_category='COMPL'
-- ^^^ ^^^ (not WHERE)
group by c.ID
Of course, it would be much better if you used proper (SQL Standard) join syntax.