Yes, you are correct. In the sample above, CTE would be the best approach.

But just an FYI, we can also just add a where clause in my query and have the same result as yours…

select * from OrganizationUnits where OrganizationUnitId = 3
union all
select s1.*
from OrganizationUnits s1
inner join OrganizationUnits s2 on s1.ParentOrganizationUnitId = s2.OrganizationUnitId
WHERE
s2.OrganizationUnitId = 3 OR s2.ParentOrganizationUnitId = 3

Thanks.

1 | World | NULL | 1
2 | USA | 1 | 1
3 | China | 1 | 1
4 | Fujian | 3 | 1
5 | Longyan | 4 | 1
6 | Virgina | 2 | 1
7 | Falls Church | 6 | 1
8 | California | 2 | 1
9 | Los Angeles | 8 | 1

And lets say we wanted to get all children for “World”. That should return everything, right?

Mine as expected returns all rows if I specify 1 as who I want to get the children for. Yours also returns all because it unions all. If I wanted to get JUST the children for China, and ran yours like this:

select * from OrganizationUnits where OrganizationUnitId = 3
union all
select s1.*
from OrganizationUnits s1
inner join OrganizationUnits s2 on s1.ParentOrganizationUnitId = s2.OrganizationUnitId

It would still return everything, including USA which is not a child of China. However using a CTE will properly return

3 | China | 1 | 1
4 | Fujian | 3 | 1
5 | Longyan | 4 | 1

Thank you.

select * from OrganizationUnits where OrganizationUnitId = 1
union all
select s1.*
from OrganizationUnits s1
inner join OrganizationUnits s2 on s1.ParentOrganizationUnitId = s2.OrganizationUnitId

http://dgoyani.blogspot.com/2005/09/effective-way-to-expand-data-tree-in.html