SQL Server Row_number in ORDER BY CASE - sql

Question

EDITED THE WHOLE TOPIC.
I need to create a view that sort article per type.
If I only have the type : *VALUE -> I need to show this line only.
If I have the type : *VALUE & 2 -> Still showing row accordingly to *VALUE type only.
If I only have the type : 2 -> Showing this one.
I already did somethink like this :
VALUE* is a value that should come from an another table with a Join.
SELECT Id_item ,Name_item , Type_item , Id_type_item FROM ITEM
WHERE Name_item = 'Gillette' AND (Id_Type_item = VALUE* OR Id_Type_item ='10')
ORDER BY CASE
WHEN row_number() OVER(ORDER BY Id_item DESC , Id_Type_Item DESC) <= 1 THEN 0
ELSE 1
END;
But it does that in the case where we've got both row for the types(*VALUE & 10):
Id_item / Name_item / Type_item / Id_Type_Item
1 Gillette 45 30 (*VALUE)
1 Gillette 2 10
So I think that the order by on the Over() could be useful to always sort by *VALUE (which are in reality another column from another table)
I always want to select 1 row of data only ! :)

Answer 1

I'm guessing, that what you want is the "first" row returned from each SELECT? There's no need to use a separate SELECT statement for each variable on the same table, you can use a window function to do so. I believe this is what you might be after.
WITH CTE AS(
--The following assumes table A and B have the same DDL (which begs the question, why are they different tables?)
SELECT *,
ROW_NUMBER() OVER (PARTITION BY var
ORDER BY (SELECT NULL)) AS RN --Replace SELECT(NULL) with your actual ordering criteria
FROM A
WHERE var IN (1,2)
UNION --ALL(?)
SELECT *
ROW_NUMBER() OVER (PARTITION BY var
ORDER BY (SELECT NULL)) AS RN --Replace SELECT(NULL) with your actual ordering criteria
FROM B
WHERE var IN (3))
SELECT *
FROM CTE
WHERE RN = 1;

Answer 2

Here is a possible solution. In this case ROW_NUMBER, RANK and DENSE_RANK would all work. However, ROW_COUNT is not a valid window function in sql server.
DECLARE #A TABLE(ID INT, Value INT)
DECLARE #B TABLE(ID INT,Value INT)
INSERT INTO #A VALUES (1,1),(2,1),(3,2),(4,3),(5,2),(6,1),(7,3)
INSERT INTO #B VALUES (1,1),(2,1),(3,1),(4,2),(5,3),(6,2),(7,1),(8,3)
;WITH D AS
(
SELECT ID,Value FROM #A WHERE Value IN(1,2)
UNION ALL
SELECT ID,Value FROM #B WHERE Value IN (3)
)
SELECT * FROM
(
SELECT
ID, Value,
ValueRankInSet = DENSE_RANK() OVER(PARTITION BY VALUE ORDER BY ID) -- <-- If you do not have an ID field you can subst ID with NEWID() as order is not important
FROM D
)AS X
WHERE ValueRankInSet = 1

Answer 3

Assign the priority within your Select(s) and then order by it in the row_number:
with cte as
(
SELECT *,
row_number()
over (-- partition by ???
order by prio) as Position
FROM
(
SELECT 1 as prio, * FROM A WHERE var = 1
UNION -- probably a more efficient UNION ALL
SELECT 2 as prio, * FROM A WHERE var = 2
UNION -- probably a more efficient UNION ALL
SELECT 3 as prio, * FROM B WHERE var = 3
)
)
select *
from cte
WHERE Position = 1