Count all subordinates who directly or indirectly reports to managers

Question

I got a problem with one task.

I need to count all subordinates (distinct) who directly or indirectly reports to a specific manager

I have an Employee table like this:

EMPLOYEE_ID Int,
MANAGER_ID Int,
EMPLOYEE_NAME varchar(200)

Example:

            Alex(1)
    --------------------
    Jhon(2)          Kevin(3)
------------------------------
Mike(4) Amanda(5)  Tom(6) Jery(7)

I can count only employee who directly reports to a manager:

SELECT 
    MANAGER_ID
   ,COUNT(MANAGER_ID) as SubCount
FROM [dbo].[EMPLOYEE]
GROUP BY MANAGER_ID

But in result I have something like this:

Manager_ID | SubCount
----------------------
1          | 2
2          | 2
3          | 2
----------------------

Instead:

Manager_ID | SubCount
----------------------
1          | 6
2          | 2
3          | 2
----------------------

I will be happy for any suggestions or idea how to do this.

Answer 1

declare @t table(EMPLOYEE_ID Int,
MANAGER_ID Int,
EMPLOYEE_NAME varchar(200))
insert @t values(1,null,'Alex'),(2,1,'Jhon'),(3,1,'Kevin'),
(4,2,'Mike'),(5,2,'Amanda'),(6,3,'Tom'),(7,3,'Jerry')

;with a as
(
select EMPLOYEE_ID boss,EMPLOYEE_ID from @t t
  where exists (select 1 from @t where t.EMPLOYEE_ID = MANAGER_ID)
union all
select a.boss, t.EMPLOYEE_ID
from @t t join a on t.MANAGER_ID = a.EMPLOYEE_ID
)
--subtracting 1 because it is also counting the manager
select boss, count(*)-1 SubCount from a group by boss
option (maxrecursion 20)

Answer 2

You need to use a recursive CTE for this:

with managers as (
      select employee_id, manager_id, 1 as level
      from employees e
      union all
      select e.employee_id, m.manager_id, e.level+1
      from managers m join
           employees e
           on m.manager_id = e.employee_id
     )
select m.manager_id, count(*)
from managers m
group by m.manager_id;

The recursive CTE creates all employee/manager pairs. The final query does the aggregation and counting.

EDIT:

It sounds like there are cycles in the employee/manager relationship. I think the following fixes this (I don't have SQL Server available right now to test it):

with managers as (
      select employee_id, manager_id, 1 as level
      from employees e
      union all
      select e.employee_id, m.manager_id, e.level+1
      from managers m join
           employees e
           on m.manager_id = e.employee_id
      where e.employee_id not in (select employee_id from managers)
     )
select m.manager_id, count(*)
from managers m
group by m.manager_id;