Resolving hierarchy in a database table - sql-server

Question

I have an org chart table which is modeled like this:
+-------------+------------+-----------------+
| Employee_ID | Manager_ID | Department_Name |
+-------------+------------+-----------------+
| 1 | 2 | Level1 |
| 2 | 3 | Level2 |
| 3 | | Level3 |
+-------------+------------+-----------------+
So, each employee refers to another row, in a chain which represents the org chart. With all employees, this model is used to represent the hierarchy.
However, for reporting purposes, we'd need to query a denormalized table, i.e. where the data is represented like this:
+-------------+--------+--------+--------+
| Employee_ID | ORG_1 | ORG_2 | ORG_3 |
+-------------+--------+--------+--------+
| 1 | Level1 | | |
| 2 | Level1 | Level2 | |
| 3 | Level1 | Level2 | Level3 |
+-------------+--------+--------+--------+
with an many ORG_x columns as needed to represent all levels that can be found. Then you can do simple groupings such as GROUP BY ORG_1, ORG_2, ORG_3. Note that one could reasonably assume the maximum number of levels.
So here's my question: since the database sits on SQL server, can I expect this to be feasible in Transact-SQL so that I could build a view?
Before I start learning T-SQL, I want to make sure I'm on the right track.
(BTW, if yes, I'd be interested in recommendations for a good tutorial!)
Thanks!
R.

Answer 1

I would use common table expressions with PIVOT:
DECLARE #T TABLE
(
Employee_ID int,
Manager_ID int,
Department_Name varchar(10)
);
INSERT #T VALUES
(1,2,'Level 1'),
(2,3,'Level 2'),
(3,NULL,'Level 3');
WITH C AS (
SELECT Employee_ID, Manager_ID, Department_Name
FROM #T
UNION ALL
SELECT T.Employee_ID, T.Manager_ID, C.Department_Name
FROM C
JOIN #T T ON C.Manager_ID=T.Employee_ID
), N AS (
SELECT ROW_NUMBER() OVER (PARTITION BY Employee_ID ORDER BY Department_Name) N, *
FROM C
)
SELECT Employee_ID, [1] ORG_1, [2] ORG_2, [3] ORG_3
FROM N
PIVOT (MAX(Department_Name) FOR N IN ([1],[2],[3])) P
ORDER BY Employee_ID
Result:
Employee_ID ORG_1 ORG_2 ORG_3
----------- ---------- ---------- ----------
1 Level 1 NULL NULL
2 Level 1 Level 2 NULL
3 Level 1 Level 2 Level 3
Note: If you have only 3 levels, you can also do simple 3 x JOIN

Answer 2

Yes the pattern you have here is known as an adjacency list. It is very common. The downside is that build your tree requires you to use recursion which can lead to performance problems on large sets. Another approach that is a lot faster is to use the Nested Sets model. It is a little less intuitive at first but once you understand the concept it is super easy.
No matter which model you use to store your data it is going to require a dynamic pivot or a dynamic crosstab to get it in the denormalized format you need.