ensure same UPCs in every assortment-group with SQL - sql

Question

I'm having a table where Assortments are groups containing multiple products (UPC is the productId). I want to verify that the same UPCs are in each Assortment, but I don't know how to do this.
Here is some test data - we're on SQL Server 2014:
--"correct" as each assortment contains the same products
DECLARE #input TABLE (Assortment VARCHAR(100) NOT NULL, UPC VARCHAR(20) NOT NULL);
INSERT INTO #input (Assortment,UPC)
VALUES ('A','1'),('A','2'),('B','1'),('B','2');
--"not correct" as _not_ each assortment contains the same products
DECLARE #input TABLE (Assortment VARCHAR(100) NOT NULL, UPC VARCHAR(20) NOT NULL);
INSERT INTO #input (Assortment,UPC)
VALUES ('A','1'),('A','2'),('B','1'),('B','3');
My first idea was to count the number of products for each Assortment and check if the count is the same for each Assortment and do the same thing the other way round - checking that the number of Assortments a UPC is in is always the same.
But this does not work in a case like this:
DECLARE #input TABLE (Assortment VARCHAR(100) NOT NULL, UPC VARCHAR(20) NOT NULL);
INSERT INTO #input (Assortment,UPC)
VALUES ('A','1'),('A','2'),('B','3'),('B','4');
My next idea was to group by UPC and use a CLR-function to concat all Assortments a UPC is in and compare those for differences, but unfortunately it ignores sort order and by that does not work.
Any suggestions?

Answer 1

This can be solved by using grouping with FOR XML PATH as suggested in the comment from AaronBertrand.
Here is the code I'm using now:
DECLARE #input TABLE (Assortment VARCHAR(100) NOT NULL, UPC VARCHAR(20) NOT NULL);
INSERT INTO #input (Assortment,UPC)
VALUES ('A','1'),('A','2'),('B','1'),('B','2');
DECLARE #grouped TABLE (UPC VARCHAR(100) NOT NULL, Assortments VARCHAR(MAX) NOT NULL)
INSERT INTO #grouped (UPC,Assortments)
SELECT UPC, Assortments = STUFF(
(SELECT N', ' + i2.Assortment
FROM #input AS i2
WHERE i2.UPC = i.upc
ORDER BY
i2.Assortment
FOR XML PATH(N''),
TYPE).value(N'./text()[1]', N'nvarchar(max)'), 1, 2, N'')
FROM #input AS i
GROUP BY
UPC;
--if this returns something, it is "not correct"
SELECT g.UPC,
g.Assortments
FROM #grouped AS g
WHERE EXISTS (
SELECT g2.Assortments
FROM #grouped AS g2
WHERE g2.UPC <> g.UPC AND
g2.Assortments <> g.Assortments
)
Edit:
After thinking about it again cross joining DISTINCT Assortments WITH DISTINCT UPCs to get all combinations that should be there and then doing a LEFT OUTER JOIN to #input should also work and avoids the string concatenation.
DECLARE #input TABLE (Assortment VARCHAR(100) NOT NULL, UPC VARCHAR(20) NOT NULL);
INSERT INTO #input (Assortment,UPC)
VALUES ('A','1'),('A','2'),('B','3'),('B','4');
;WITH distinct_upcs AS (
SELECT DISTINCT
i.UPC
FROM #input AS i
),
distinct_assortments AS (
SELECT DISTINCT
i.Assortment
FROM #input AS i
),
crossed AS (
SELECT du.UPC,
da.Assortment
FROM distinct_assortments AS da CROSS JOIN
distinct_upcs AS du
)
--if this returns something it is "not correct"
SELECT c.UPC,
c.Assortment
FROM crossed AS c LEFT OUTER JOIN
#input AS i ON
c.UPC = i.UPC AND
c.Assortment = i.Assortment
WHERE i.UPC IS NULL