Count Values associated with key in Sql Server - sql

Question

I have three tables
Table Category
CategoryID Category
1 Climate
2 Area
Table CategoryDetail
DetailID CategoryID Desc
1 1 Hot
2 1 Cold
3 2 Area1
Table CategoryDetailValues
PK AnotherFK CategoryDetailID
1 1 1
2 1 1
3 1 2
4 2 1
Here AnotherFK is foreign key referring to another table. In record 1 and 2 duplicate exists that's ok but AnotherFK 1 has reference of CategoryDetailID 1 and 2 which has categoryID of 1 which is not ok
So from above tables
this result is valid from above three table
PK AnotherFK CategoryID DetailID Desc
1 1 1 1 Hot
2 1 1 1 Hot
But below result is not valid
PK AnotherFK CategoryID DetailID Desc
2 1 1 1 Hot
3 1 1 2 Cold
I can not put same AnotherFK in two different DetailID which has same CategoryID. I could have eliminated this by introducing CategoryID in CategoryDetailValues table and creating unique constraint but I am not allowed to do so.
Now my aim is to find all those record in CategoryDetailValues table which has different DetailID that are associated with same CategoryID. So that I can delete them.
Trying to achieve this in SQL Server 2012.

Answer 1

If your goal is to highlight all AnotherFK cases that have the same CategoryID, but differenty DetailIDs, the following ought to do the trick (pseudo-code):
SELECT * FROM (SELECT AnotherFK, ROW_NUMBER() OVER
(ORDER BY AnotherFK, CategoryID) AS rn FROM #myTable) AS a
WHERE rn > 1
Sample code:
CREATE TABLE #myTable
(
AnotherFK int
, CategoryID int
, DetailID int
) ;
INSERT INTO #myTable (
AnotherFK
, CategoryID
, DetailID
)
VALUES (1, 1, 1)
, (1, 1, 2);
SELECT * FROM (SELECT AnotherFK, ROW_NUMBER() OVER (ORDER BY AnotherFK, CategoryID) AS rn FROM #myTable) AS a
WHERE rn > 1
DROP TABLE #myTable
If this is not what you are after, please elaborate

Answer 2

I think you could use something like this:
Script to create sample tables:
CREATE TABLE mytable(
PK INTEGER NOT NULL PRIMARY KEY
,AnotherFK INTEGER NOT NULL
,CategoryDetailID INTEGER NOT NULL
);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (1,1,1);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (2,1,1);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (3,1,2);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (4,2,1);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (5,1,3);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (6,1,3);
INSERT INTO mytable(PK,AnotherFK,CategoryDetailID) VALUES (7,1,3);
CREATE TABLE mytable2(
DetailID INTEGER NOT NULL
,CategoryID INTEGER NOT NULL
,Descr VARCHAR(5) NOT NULL
);
Query to show "suspect" record (I think you have to decide what records delete...):
SELECT * FROM (
SELECT * ,COUNT(*) OVER (PARTITION BY CategoryID, ANotherFK) AS X
, COUNT(*) OVER (PARTITION BY CategoryID, DetailID, ANotherFK) AS X1
FROM mytable A
INNER JOIN mytable2 B ON A.CategoryDetailID= B.DetailID
)C
WHERE X-X1 >0
Output:
+--+----+-----------+------------------+----------+------------+-------+---+----+
| | PK | AnotherFK | CategoryDetailID | DetailID | CategoryID | Descr | X | X1 |
+--+----+-----------+------------------+----------+------------+-------+---+----+
| | 1 | 1 | 1 | 1 | 1 | Hot | 3 | 2 |
| | 2 | 1 | 1 | 1 | 1 | Hot | 3 | 2 |
| | 3 | 1 | 2 | 2 | 1 | Cold | 3 | 1 |
+--+----+-----------+------------------+----------+------------+-------+---+----+

Answer 3

This query will look in Categorydetail for records with duplicate DetailID. Than join the tables to provide you the details. It's still up to you to decide which records should be deleted.
select *
from(
Select CategoryID
from CategoryDetail
group by CategoryID
having count(DetailID)>1)aggr
join CategoryDetail c on aggr.CategoryID = c.CategoryID
join CategoryDetailValues v on c.CategoryDetailID = v.CategoryDetailID

Answer 4

You want one value per AnotherFK and Category. So the third table should have a composite key:
CategoryDetailValues(AnotherFK, CategoryID, DetailID, HowMany)
with a unique constraint on AnotherFK, CategoryID and both building a foreign key to CategoryDetail(CategoryID, DetailID).
In order to clean up data first, you'd have to look for ambiguities:
select AnotherFK, CategoryID, DetailID
from
(
select
cdv.AnotherFK, cd.CategoryID, cdv.DetailID,
count(distinct cd.DetailID) over (partition by cdv.AnotherFK, cd.CategoryID) as cnt
from CategoryDetailValues cdv
join CategoryDetail cd on cd.DetailID = cdv.CategoryDetailID
)
where cnt > 1
order by AnotherFK, CategoryID, DetailID

Answer 5

You could try this solution that comes with following assumption: within CDV table, for every [AnotherFK] value (ex. 1) should be displayed only those rows with the minimum [CategoryDetailID] (ex. 1)
SELECT *
FROM (
SELECT cdv.PK, cdv.AnotherFK, cd.CategoryID, cd.[Desc],
Rnk = DENSE_RANK() OVER(PARTITION BY cdv.AnotherFK ORDER BY cdv.CategoryDetailID)
FROM dbo.CategoryDetailValues cdv
JOIN dbo.CategoryDetail cd ON cd.DetailID = cdv.DetailID
WHERE cdv.AnotherFK = 1
) x
WHERE x.Rnk = 1