CODEWITHSUNDEEP
sqlsql-serverstring-aggregation
ThebulletKin
ThebulletKin

Reputation: 25

Table column aggregation results in the same value duplicating multiple times in one cell

I have a set of tables aimed at representing a profile storage system. I have created a view that should display all the important parts of this profile, that being things like the name, email, number of followers and the list of activities they are a part of.

However, I've noticed that the values in the activities cell are duplicating. The same thing appears multiple times when each activity name should only appear once. In the below table the Nulls at the bottom are intentional.

Weight Height Activities TotalFollowing TotalFollowers
68 170 Horse Riding, Horse Riding, Horse Riding, Horse Riding, Horse Riding, Horse Riding, Horse Riding, Horse Riding, Bird Watching, Bird Watching, Bird Watching, Bird Watching, Bird Watching, Bird Watching, Bird Watching, Bird Watching, Hunting, Hunting, Hunting, Hunting, Hunting, Hunting, Hunting, Hunting 4 2
63 179 Horse Riding, Horse Riding, Horse Riding, Hiking, Hiking, Hiking, Swimming, Swimming, Swimming, Fishing, Fishing, Fishing 1 3
72 130 NULL 3 1
NULL NULL NULL 1 2

There is a user that is identified by it's MemberID, a link table called User_Activity which only contains MemberID and ActivityID, the latter of which points to the activity table that holds the name for an activity under a certain ID.

ActivityID ActivityName
1 Horse Riding
2 Hiking
3 Bird Watching
4 Backpacking
5 Swimming
6 Fishing
7 Hunting

And the user activity tables holds these.

MemberID ActivityID
1 1
1 3
1 7
2 1
2 2
2 5
2 6

The SQL for this view is as follows.
I assume the issue has come from the fact that the user follows table and the references I use there and up pointing back to the user, so it counts their activities again and adds them to the column depending on how many followers they have, since these two example records repeat a different number of times.

CREATE VIEW [View_Profile_Information] AS
SELECT U.MemberID, 
U.FirstName, 
U.LastName, 
L.Town, 
L.County, 
L.Country, 
U.Email, 
U.AboutMe, 
U.DistanceUnit, 
U.ActivityTimePreference, 
U.Weight,        
U.Height, 
STRING_AGG(A.ActivityName, ', ') AS Activities, 
COUNT(DISTINCT UF.Follows_MemberID) AS TotalFollowing, 
COUNT(DISTINCT UF2.This_MemberID) AS TotalFollowers
FROM CW1.[User] U

LEFT JOIN CW1.[Location] L
ON U.LocationID = L.LocationID

LEFT JOIN CW1.[User_Activity] UA
ON U.MemberID = UA.MemberID    
              
LEFT JOIN CW1.[Activity] A
ON UA.ActivityID = A.ActivityID  

LEFT JOIN CW1.[User_Follows] UF 
ON U.MemberID = UF.This_MemberID

LEFT JOIN CW1.[User_Follows] UF2
ON U.MemberID = UF2.Follows_MemberID

GROUP BY U.MemberID, U.FirstName, LastName, Town, County, Country, Email, AboutMe, DistanceUnit, ActivityTimePreference, Weight, 
       Height

Upvotes: 0

Views: 96

Answers (1)

T N
T N

Reputation: 9907

What you have is a query that contains multiple independent one-to-many joins. This is near guaranteed to cause data duplication in the results. Using DISTINCT or GROUP BY to try to resolve the duplicates is rarely the correct solution for unexpected duplicate data.

What you need to do is to moved each relationship branch into separate subqueries. In this case those subqueries can either be included directly in the select list or wrapped up in a CROSS APPLY.

Something like:

CREATE VIEW [View_Profile_Information] AS
SELECT
    U.MemberID, U.FirstName, U.LastName, 
    L.Town, L.County, L.Country, 
    U.Email, U.AboutMe, U.DistanceUnit,
    U.ActivityTimePreference, U.Weight, U.Height,
    (
        SELECT STRING_AGG(A.ActivityName, ', ')
        FROM CW1.[User_Activity] UA
        JOIN CW1.[Activity] A
           ON UA.ActivityID = A.ActivityID
        WHERE UA.MemberID = U.MemberID
    ) AS Activities,
    (
        SELECT COUNT(*)
        FROM CW1.[User_Follows] UF 
        WHERE UF.This_MemberID = U.MemberID
    ) AS TotalFollowing, 
    (
        SELECT COUNT(*)
        FROM CW1.[User_Follows] UF2
        WHERE UF2.Follows_MemberID = U.MemberID
    ) AS TotalFollowers
FROM CW1.[User] U
LEFT JOIN CW1.[Location] L
    ON U.LocationID = L.LocationID

You can also use CROSS APPLY to calculate the aggregated values separately down in the FROM section of the query, so that the final select list is cleaner and more readable.

CREATE VIEW View_Profile_Information AS
SELECT
    U.MemberID, U.FirstName, U.LastName, 
    L.Town, L.County, L.Country, 
    U.Email, U.AboutMe, U.DistanceUnit,
    U.ActivityTimePreference, U.Weight, U.Height,
    A.Activities, TFG.TotalFollowing, TFW.TotalFollowers
FROM CW1.User U
LEFT JOIN CW1.Location L
    ON U.LocationID = L.LocationID
CROSS APPLY (
    SELECT STRING_AGG(A.ActivityName, ', ') AS Activities
    FROM CW1.User_Activity UA
    JOIN CW1.Activity A
       ON UA.ActivityID = A.ActivityID
    WHERE UA.MemberID = U.MemberID
) AS A
CROSS APPLY (
    SELECT COUNT(*) AS TotalFollowing
    FROM CW1.User_Follows UF 
    WHERE UF.This_MemberID = U.MemberID
) AS TFG
CROSS APPLY (
    SELECT COUNT(*) AS TotalFollowers
    FROM CW1.User_Follows UF2
    WHERE UF2.Follows_MemberID = U.MemberID
) AS TFW

Note that I have removed the DISTINCTs and GROUP BY, and COUNT(*) is now just counting rows.

Upvotes: 1

Related Questions