sql get average salary in self reference table - sql

Question

I've seen several questions/answers on how to recursively query a self-referencing table, but I am struggling to apply the answers I've found to aggregate up to each parent, grandparent, etc. regardless of where the item sits in the hierarchy.
Need to get an average salary for each department including hierarchy.
It means department should include the average salary of each sub-department and so on.
I've got nex db schema:
CREATE TABLE Employee
(
Id INT NOT NULL ,
Name VARCHAR(200) NOT NULL ,
Department_Id INT NOT NULL ,
PRIMARY KEY ( Id )
);
CREATE TABLE Department
(
Id INT NOT NULL ,
DepartmentName VARCHAR(200) NOT NULL ,
Parent_Id INT ,
PRIMARY KEY ( Id )
);
CREATE TABLE Salary
(
Id INT NOT NULL ,
Date DATETIME NOT NULL ,
Amount INT NOT NULL ,
Employee_Id INT NOT NULL ,
PRIMARY KEY ( Id )
);
I've tried something like that but it includes only 1st level of a hierarchy.
SELECT d.Id ,
d.DepartmentName ,
( SELECT AVG(s.Amount)
FROM dbo.Department dd
LEFT JOIN dbo.Department sdd ON dd.Id = sdd.Parent_Id
JOIN dbo.Employee e ON e.Department_Id = sdd.Id
OR e.Department_Id = dd.Id
JOIN dbo.Salary s ON s.Employee_Id = e.Id
WHERE dd.Id = d.Id
) AS avg_dep_salary
FROM dbo.Department d
WHERE d.Parent_Id IS NULL;
How can get an average salary of all levels?
EDIT: Added some inserts
INSERT INTO Employee
( Id, Name, Department_Id )
VALUES ( 1, 'Peter', 1 ),
( 2, 'Alex', 1 ),
( 3, 'Sam', 2 ),
( 4, 'James', 2 ),
( 5, 'Anna', 3 ),
( 6, 'Susan', 3 ),
( 7, 'Abby', 4 ),
( 8, 'Endy', 4 );
INSERT INTO Department
( Id, DepartmentName, Parent_Id )
VALUES ( 1, 'IT', NULL ),
( 2, 'HR', NULL),
( 3, 'SubIT', 1 ),
( 4, 'SubSubIT', 3 );
INSERT INTO Salary
( Id, Date, Amount, Employee_Id )
VALUES ( 1, '2013-01-09 16:03:50.003', 3000, 1 ),
( 2, '2013-01-11 16:03:50.003', 5000, 2 ),
( 3, '2013-01-09 16:03:50.003', 2000, 3 ),
( 4, '2013-01-11 16:03:50.003', 1000, 4 ),
( 5, '2013-01-09 16:03:50.003', 4000, 5 ),
( 6, '2013-01-11 16:03:50.003', 6000, 6 ),
( 7, '2013-01-09 16:03:50.003', 7000, 7 ),
( 8, '2013-01-13 16:03:50.003', 9000, 8 );
Expected result is:
Department | Average_Salary
__________________________________
IT | ( X1 + X2 + X3 ) / 3
HR | ( Y1 ) / 1
SubIT | ( X2 + X3 ) / 2
SubSubIT | ( X3 ) / 1
Where:
X1 - Average salary of IT department
X2 - Average salary of SubIT department
X3 - Average salary of SubSubIT department
Y1 - Average
salary of HR department

Answer 1

Sample data
I've added few rows with a wider tree structure.
DECLARE #Employee TABLE
(
Id INT NOT NULL ,
Name VARCHAR(200) NOT NULL ,
Department_Id INT NOT NULL ,
PRIMARY KEY ( Id )
);
DECLARE #Department TABLE
(
Id INT NOT NULL ,
DepartmentName VARCHAR(200) NOT NULL ,
Parent_Id INT ,
PRIMARY KEY ( Id )
);
DECLARE #Salary TABLE
(
Id INT NOT NULL ,
Date DATETIME NOT NULL ,
Amount INT NOT NULL ,
Employee_Id INT NOT NULL ,
PRIMARY KEY ( Id )
);
INSERT INTO #Employee
( Id, Name, Department_Id )
VALUES
( 1, 'Peter', 1 ),
( 2, 'Alex', 1 ),
( 3, 'Sam', 2 ),
( 4, 'James', 2 ),
( 5, 'Anna', 3 ),
( 6, 'Susan', 3 ),
( 7, 'Abby', 4 ),
( 8, 'Endy', 4 ),
(10, 'e_A', 10),
(11, 'e_AB', 11),
(12, 'e_AC', 12),
(13, 'e_AD', 13),
(14, 'e_ACE', 14),
(15, 'e_ACF', 15),
(16, 'e_ACG', 16);
INSERT INTO #Department
( Id, DepartmentName, Parent_Id )
VALUES
( 1, 'IT', NULL ),
( 2, 'HR', NULL),
( 3, 'SubIT', 1 ),
( 4, 'SubSubIT', 3 ),
(10, 'A', NULL ),
(11, 'AB', 10),
(12, 'AC', 10),
(13, 'AD', 10),
(14, 'ACE', 12),
(15, 'ACF', 12),
(16, 'ACG', 12);
INSERT INTO #Salary
( Id, Date, Amount, Employee_Id )
VALUES
( 1, '2013-01-09 16:03:50.003', 3000, 1 ),
( 2, '2013-01-11 16:03:50.003', 5000, 2 ),
( 3, '2013-01-09 16:03:50.003', 2000, 3 ),
( 4, '2013-01-11 16:03:50.003', 1000, 4 ),
( 5, '2013-01-09 16:03:50.003', 4000, 5 ),
( 6, '2013-01-11 16:03:50.003', 6000, 6 ),
( 7, '2013-01-09 16:03:50.003', 7000, 7 ),
( 8, '2013-01-13 16:03:50.003', 9000, 8 ),
(10, '2013-01-13 16:03:50', 100, 10),
(11, '2013-01-13 16:03:50', 100, 11),
(12, '2013-01-13 16:03:50', 100, 12),
(13, '2013-01-13 16:03:50', 100, 13),
(14, '2013-01-13 16:03:50', 100, 14),
(15, '2013-01-13 16:03:50', 100, 15),
(16, '2013-01-13 16:03:50', 100, 16);
Query
WITH
CTE_Departments
AS
(
SELECT
D.Id
,D.Parent_Id
,D.DepartmentName
,SUM(Amount) AS DepartmentAmount
,COUNT(*) AS DepartmentCount
FROM
#Department AS D
INNER JOIN #Employee AS E ON E.Department_Id = D.Id
INNER JOIN #Salary AS S ON S.Employee_Id = E.Id
GROUP BY
D.Id
,D.Parent_Id
,D.DepartmentName
)
,CTE_Recursive
AS
(
SELECT
CTE_Departments.Id AS OriginalID
,CTE_Departments.DepartmentName AS OriginalName
,CTE_Departments.Id
,CTE_Departments.Parent_Id
,CTE_Departments.DepartmentName
,CTE_Departments.DepartmentAmount
,CTE_Departments.DepartmentCount
,1 AS Lvl
FROM CTE_Departments
UNION ALL
SELECT
CTE_Recursive.OriginalID
,CTE_Recursive.OriginalName
,CTE_Departments.Id
,CTE_Departments.Parent_Id
,CTE_Departments.DepartmentName
,CTE_Departments.DepartmentAmount
,CTE_Departments.DepartmentCount
,CTE_Recursive.Lvl + 1 AS Lvl
FROM
CTE_Departments
INNER JOIN CTE_Recursive ON CTE_Recursive.Id = CTE_Departments.Parent_Id
)
SELECT
OriginalID
,OriginalName
,SUM(DepartmentAmount) AS SumAmount
,SUM(DepartmentCount) AS SumCount
,SUM(DepartmentAmount) / SUM(DepartmentCount) AS AvgAmount
FROM CTE_Recursive
GROUP BY
OriginalID
,OriginalName
ORDER BY OriginalID
;
Result
+------------+--------------+-----------+----------+-----------+
| OriginalID | OriginalName | SumAmount | SumCount | AvgAmount |
+------------+--------------+-----------+----------+-----------+
| 1 | IT | 34000 | 6 | 5666 |
| 2 | HR | 3000 | 2 | 1500 |
| 3 | SubIT | 26000 | 4 | 6500 |
| 4 | SubSubIT | 16000 | 2 | 8000 |
| 10 | A | 700 | 7 | 100 |
| 11 | AB | 100 | 1 | 100 |
| 12 | AC | 400 | 4 | 100 |
| 13 | AD | 100 | 1 | 100 |
| 14 | ACE | 100 | 1 | 100 |
| 15 | ACF | 100 | 1 | 100 |
| 16 | ACG | 100 | 1 | 100 |
+------------+--------------+-----------+----------+-----------+
Run the query step-by-step, CTE-by-CTE to understand how it works.
CTE_Departments gives total amount and number of people for each department.
CTE_Recursive recursively generates child rows for each department, while keeping the OriginalID - the ID of the department where the recursion started.
Final query simply groups everything by this OriginalID.

Answer 2

Here is one way
with avg_per_dep as (
select
[Month] = eomonth(s.date), d.Id, d.DepartmentName
, avgDep = avg(s.Amount * 1.0)
from
Salary s
join Employee e on s.Employee_Id = e.Id
join Department d on e.Department_Id = d.Id
group by d.Id, d.DepartmentName, eomonth(s.date)
)
, rcte as (
select
i = Id, Id
, list = cast(',' + cast(Id as varchar(10)) + ',' as varchar(max))
, step = 1
from
Department
union all
select
a.i, b.Id, cast(a.list + cast(b.Id as varchar(10)) + ',' as varchar(max))
, step + 1
from
rcte a
join Department b on a.Id = b.Parent_Id
)
select
d.DepartmentName, c.[Month]
, Average_Salary = avg(c.avgDep)
from
(
select
top 1 with ties i, list
from
rcte
order by row_number() over (partition by i order by step desc)
) t
join avg_per_dep c on t.list like '%,' + cast(c.Id as varchar(10)) + ',%'
join Department d on t.i = d.Id
group by t.i, d.DepartmentName, c.[Month]
Output
DepartmentName [Month] Average_Salary
---------------------------------------------
IT 2013-01-31 5666.666666
HR 2013-01-31 1500.000000
SubIT 2013-01-31 6500.000000
SubSubIT 2013-01-31 8000.000000
Idea:
Calculate average salary per department
Get a list of departments with all childs with recursive CTE.
Join two table and calculate avg with childs

Answer 3

You could also use below query to get the expected result
WITH Department_Path
AS (SELECT Id, CAST(CONCAT('#', Id, '#') AS VARCHAR(255)) AS Path
FROM Department
WHERE Parent_Id IS NULL
UNION ALL
SELECT Child.Id, CAST(CONCAT(Parent.Path, Child.Id, '#') AS VARCHAR(255)) AS Path
FROM Department Child
INNER JOIN Department_Path Parent
ON Parent.Id = Child.Parent_Id)
SELECT Department.Id,
Department.DepartmentName,
AVG(Salary.Amount) As Average_Salary,
COUNT(Employee.Id) AS Employee_Count
FROM Department
INNER JOIN Department_Path
ON CHARINDEX(CONCAT('#', Department.Id, '#'), Department_Path.Path) > 0
INNER JOIN Employee
ON Employee.Department_Id = Department_Path.Id
INNER JOIN Salary
ON Salary.Employee_Id = Employee.Id
GROUP BY Department.Id,
Department.DepartmentName;
The idea is that each employee is belong to a list of hierarchy departments. For each department, we could retrieve all employee who belong to it and then calculate the average salary.