SQL reporting query - sql

Question

I have a database with following structure.
CREATE TABLE Party
(
PartyID INT IDENTITY
PRIMARY KEY ,
StatusID INT ,
Weigth INT ,
OldWeigth INT
);
GO
CREATE TABLE PartyLocation
(
PartyLocationID INT IDENTITY
PRIMARY KEY ,
PartyID INT FOREIGN KEY REFERENCES dbo.Party ( PartyID ) ,
LocationID INT ,
Distance INT
);
GO
CREATE TABLE PartyRole
(
PartyRoleID INT IDENTITY
PRIMARY KEY ,
PartyID INT FOREIGN KEY REFERENCES dbo.Party ( PartyID ) ,
RoleID INT
);
with some simple data.
INSERT INTO dbo.Party
( StatusID, Weigth, OldWeigth )
VALUES ( 1, -- StatusID - int
10, -- Age - int
20 -- OldAge - int
),
( 1, 15, 25 ),
( 2, 20, 30 );
INSERT INTO dbo.PartyLocation
( PartyID, LocationID, Distance )
VALUES ( 1, -- PartyID - int
1, -- LocationID - int
100 -- Distance - int
),
( 1, 2, 200 ),
( 1, 3, 300 ),
( 2, 1, 1000 ),
( 2, 2, 2000 ),
( 3, 1, 10000 );
INSERT INTO dbo.PartyRole
( PartyID, RoleID )
VALUES ( 1, -- PartyID - int
1 -- RoleID - int
),
( 1, 2 ),
( 1, 3 ),
( 2, 1 ),
( 2, 2 ),
( 3, 1 );
I want to query the following information
Return sum of Weigth of all parties that has roleID = 1 in PartyRole table
Return sum of OldWeigth of all parties that has statusID = 2
Return sum of distances of all parties that has locationID = 3
Return sum of distances of all parties that has roleID = 2
So the expected results are
FilteredWeigth FilteredOldWeigth FilteredDistance AnotherFilteredDistance
-------------- ----------------- ---------------- -----------------------
45 30 600 3600
Can we write a query that will query each table just once? If no what will be the most optimal way to query the data?

Answer 1

You can try this.
SELECT
FilteredWeigth = SUM(CASE WHEN RoleID = 1 AND RN_P = 1 THEN Weigth END) ,
FilteredOldWeigth = SUM(CASE WHEN StatusID = 2 AND RN_P = 1 THEN OldWeigth END),
FilteredDistance = SUM(CASE WHEN LocationID = 3 AND RN_L = 1 THEN Distance END),
AnotherFilteredDistance = SUM(CASE WHEN RoleID = 2 THEN Distance END)
FROM (
SELECT P.Weigth, P.StatusID, P.OldWeigth, PL.LocationID, PL.Distance, PR.RoleID,
RN_P = ROW_NUMBER() OVER (PARTITION BY P.PartyID ORDER BY PL.PartyLocationID),
RN_L = ROW_NUMBER() OVER (PARTITION BY PL.LocationID ORDER BY PR.PartyRoleID)
FROM Party P
INNER JOIN PartyLocation PL ON P.PartyID = PL.PartyID
INNER JOIN PartyRole PR ON P.PartyID = PR.PartyID
) AS T

Answer 2

the below gives
45 20 300 3600
the third column gives 300 which does not correspond to your expected result.
with q1
as
(
select sum(weigth) FilteredWeigth
from party join partyrole on party.partyid = partyrole.partyid
where partyrole.RoleID = '1'
),
q2 as
(
select sum(weigth) OldWeigth from party where StatusID = '2'
),
q3 as (
select sum(Distance) FilteredDistance
from party join PartyLocation on party.partyid = PartyLocation.partyid
where PartyLocation.locationID = '3'
),
q4 as
(
select sum(Distance) AnotherFilteredDistance
from party join partyrole on party.partyid = partyrole.partyid
join PartyLocation on party.partyid = PartyLocation.partyid
where partyrole.RoleID = '2'
)
select FilteredWeigth,OldWeigth,FilteredDistance,AnotherFilteredDistance
from q1,q2,q3,q4

Answer 3

When Using Individual Queries, you can achieve this using the following
Return sum of Weight of all parties that has roleID = 1 in PartyRole table
SELECT
SUM(Weight) FilteredWeigth
FROM dbo.Party P
WHERE EXISTS
(
SELECT
1
FROM dbo.PartyRole PR
WHERE PR. PartyID = P.PartyID
AND PR.RoleId = 1
)
Return sum of OldWeigth of all parties that has statusID = 2
SELECT
SUM(OldWeigth) FilteredOldWeigth
FROM dbo.Party P
WHERE EXISTS
(
SELECT
1
FROM dbo.PartyRole PR
WHERE PR. PartyID = P.PartyID
AND PR.RoleId = 2
)
Return sum of distances of all parties that has locationID = 3
SELECT
SUM(Distance) FilteredDistance
FROM dbo.PartyLocation
WHERE LocationID = 3
Return sum of distances of all parties that has roleID = 2
SELECT SUM(Distance) FROM PartyLocation PL
WHERE EXISTS
(
SELECT 1 FROM PartyRole PR
WHERE PR.PartyID = PL.PartyID
AND PR.Roleid = 2
)
If you want to get the result of all these in a single result set. then maybe you can try a pivot query. Like this
WITH CTE
AS
(
SELECT
'FilteredWeigth' ColNm,
SUM(Weigth) Val
FROM dbo.Party P
WHERE EXISTS
(
SELECT
1
FROM dbo.PartyRole PR
WHERE PR. PartyID = P.PartyID
AND PR.RoleId = 1
)
UNION
SELECT
'FilteredOldWeigth' ColNm,
SUM(OldWeigth) Val
FROM dbo.Party P
WHERE EXISTS
(
SELECT
1
FROM dbo.PartyRole PR
WHERE PR. PartyID = P.PartyID
AND PR.RoleId = 2
)
UNION
SELECT
'FilteredDistance' ColNm,
SUM(Distance) Val
FROM dbo.PartyLocation
WHERE LocationID = 3
UNION
SELECT
'AnotherFilteredDistance' ColNm,
SUM(Distance) Val FROM PartyLocation PL
WHERE EXISTS
(
SELECT 1 FROM PartyRole PR
WHERE PR.PartyID = PL.PartyID
AND PR.Roleid = 2
)
)
SELECT
*
FROM CTE
PIVOT
(
SUM(Val)
FOR ColNm IN
(
[FilteredWeigth],[FilteredOldWeigth],[FilteredDistance],[AnotherFilteredDistance]
)
)Pvt
The Result Will be

Answer 4

I could think of only three possible options:
Union query with four different select statements as answered by #ab-bennett
Join all tables then use select statements as answered by sarslan
Mix of 1 and 2, based on experiments
Coming to the question you asked:
Can we write a query that will query each table just once?
Assuming best performance is the goal, following could happen in each of the above cases:
All select statements would have their own where clause. This would perform best when where produces few rows compared to the count(*). Note that Joins are terrible for very large tables.
A join is made once, and the desired output is obtained from the same Joined table. This would perform optimal when where produces significant number of rows and the table is not too big to join.
You can mix JOIN / IN / EXISTS / WHERE to optimize your queries based on number of rows you are having in table. This approach could be used when your dataset cardinality might not vary a lot.