Select columns without group by in aggregate - sql

Question

I'm using SQL Server 2014. I have a structure like this:
Id BIGINT,
ItemName NVARCHAR(4000),
RecordDate DATETIME2,
Supplier NVARCHAR(450),
Quantity DECIMAL(18, 2),
ItemUnit NVARCHAR(2000),
EntityUnit NVARCHAR(2000),
ItemSize DECIMAL(18, 2),
PackageSize DECIMAL(18, 2),
FamilyCode NVARCHAR(20),
Family NVARCHAR(500),
CategoryCode NVARCHAR(20),
Category NVARCHAR(500),
SubCategoryCode NVARCHAR(20),
SubCategory NVARCHAR(500),
ItemGroupCode NVARCHAR(20),
ItemGroup NVARCHAR(500),
PurchaseValue DECIMAL(18, 2),
UnitPurchaseValue DECIMAL(18, 2),
PackagePurchaseValue DECIMAL(18, 2),
FacilityCode NVARCHAR(450),
CurrencyCode NVARCHAR(5)
I'd like to select distinct ItemNames from BatchRecords table paired with the max Id among the items with the same ItemName as well as Supplier, Quantity and other values of the item with the max Id for each ItemName. So far, I came up with the following SP, definitely it doesn't work yet as GROUP BY throws an error.
I could probably use a subquery, but then how do I satisfy the condition with max Ids for each unique ItemName? Also, any input to the stored procedure quality/obvious bottleneck is highly appreciated as it has to be somewhat quick.
CREATE PROCEDURE dbo.GetRecordsPageFlat
(#BatchIds dbo.GenericIntArray READONLY,
#FileRequestId INT,
#PageSize INT,
#PageCount INT,
#LastId BIGINT,
#NameMaskValue NVARCHAR(128) = NULL,
#NameMaskType INT = NULL,
#FamilyCodeMaskValue NVARCHAR(128),
#CategoryCodeMaskValue NVARCHAR(128),
#SubCategoryCodeMaskValue NVARCHAR(128)
)
AS
SET NOCOUNT ON;
DECLARE #Temp dbo.RecordImportStructure
DECLARE #ErrorCode INT
DECLARE #Step NVARCHAR(200)
DECLARE #Rows INT
--OUTPUT ##ROWCOUNT
--OUTPUT INSERTED.Id
INSERT INTO #Temp (
Id,
ItemName,
Supplier,
Quantity,
ItemUnit,
EntityUnit,
ItemSize,
PackageSize,
PurchaseValue,
UnitPurchaseValue,
PackagePurchaseValue,
CurrencyCode
)
SELECT
BR.Id,
BR.ItemName,
BR.Supplier,
BR.Quantity,
BR.ItemUnit,
BR.EntityUnit,
BR.ItemSize,
BR.PackageSize,
BR.ItemGroup,
BR.UnitPurchaseValue,
BR.PackagePurchaseValue,
C.IsoCode
FROM
dbo.BatchRecords BR
LEFT OUTER JOIN
dbo.FacilityInstances F ON F.Id = BR.FacilityInstanceId
LEFT OUTER JOIN
dbo.Currencies C ON C.Id = BR.CurrencyId
--OPTION(RECOMPILE)
WHERE
BR.DataBatchId IN (SELECT * FROM #BatchIds)
AND BR.Id > #LastId
AND (#FamilyCodeMaskValue IS NULL OR BR.FamilyCode = #FamilyCodeMaskValue)
AND (#CategoryCodeMaskValue IS NULL OR BR.CategoryCode = #CategoryCodeMaskValue)
AND (#SubCategoryCodeMaskValue IS NULL OR BR.SubCategoryCode = #SubCategoryCodeMaskValue)
AND (#NameMaskType IS NULL AND #NameMaskValue IS NULL
OR ((#NameMaskType = 1 AND BR.ItemName LIKE #NameMaskValue + '%')
OR (#NameMaskType = 2 AND BR.ItemName LIKE '%' + #NameMaskValue)
OR (#NameMaskType = 3 AND BR.ItemName LIKE '%' + #NameMaskValue + '%')
))
GROUP BY
BR.ItemName
ORDER BY
BR.Id
OFFSET #PageCount * #PageSize ROWS
FETCH NEXT #PageSize ROWS ONLY;
UPDATE dbo.BatchActionRequests
SET PageNumber = #PageCount+1,
LatestItemId = (SELECT MAX(Id) FROM #Temp)
WHERE Id = #FileRequestId

Answer 1

It looks like a top-n-per-group problem.
There are two common approaches to it: using ROW_NUMBER and CROSS APPLY. Here is the ROW_NUMBER variant. See Get top 1 row of each group for details.
WITH
CTE
AS
(
SELECT
BR.Id,
BR.ItemName,
BR.Supplier,
BR.Quantity,
BR.ItemUnit,
BR.EntityUnit,
BR.ItemSize,
BR.PackageSize,
-- BR.ItemGroup,???
BR.UnitPurchaseValue,
BR.PackagePurchaseValue,
C.IsoCode AS CurrencyCode,
ROW_NUMBER() OVER (PARTITION BY BR.ItemName ORDER BY BR.Id DESC) AS rn
FROM
dbo.BatchRecords BR
LEFT OUTER JOIN dbo.FacilityInstances F ON F.Id = BR.FacilityInstanceId
LEFT OUTER JOIN dbo.Currencies C ON C.Id = BR.CurrencyId
WHERE
BR.DataBatchId IN (SELECT * FROM #BatchIds)
AND BR.Id > #LastId
AND (#FamilyCodeMaskValue IS NULL OR BR.FamilyCode = #FamilyCodeMaskValue)
AND (#CategoryCodeMaskValue IS NULL OR BR.CategoryCode = #CategoryCodeMaskValue)
AND (#SubCategoryCodeMaskValue IS NULL OR BR.SubCategoryCode = #SubCategoryCodeMaskValue)
AND (#NameMaskType IS NULL AND #NameMaskValue IS NULL
OR ((#NameMaskType = 1 AND BR.ItemName LIKE #NameMaskValue + '%')
OR (#NameMaskType = 2 AND BR.ItemName LIKE '%' + #NameMaskValue)
OR (#NameMaskType = 3 AND BR.ItemName LIKE '%' + #NameMaskValue + '%')
))
)
INSERT INTO #Temp (
Id,
ItemName,
Supplier,
Quantity,
ItemUnit,
EntityUnit,
ItemSize,
PackageSize,
-- PurchaseValue,???
UnitPurchaseValue,
PackagePurchaseValue,
CurrencyCode
)
SELECT
Id,
ItemName,
Supplier,
Quantity,
ItemUnit,
EntityUnit,
ItemSize,
PackageSize,
-- PurchaseValue,???
UnitPurchaseValue,
PackagePurchaseValue,
CurrencyCode
FROM CTE
WHERE rn = 1
ORDER BY
Id
OFFSET #PageCount * #PageSize ROWS
FETCH NEXT #PageSize ROWS ONLY
OPTION(RECOMPILE);
For each ItemName the query will pick the row with the largest Id.

Answer 2

;WITH CTC
AS
(
SELECT MAX(BR.ID) AS Id, BR.ItemName
FROM dbo.BatchRecords BR
LEFT OUTER JOIN dbo.FacilityInstances F ON F.Id = BR.FacilityInstanceId
WHERE BR.DataBatchId IN (SELECT * FROM #BatchIds)
AND BR.Id > #LastId
AND (#FamilyCodeMaskValue IS NULL OR BR.FamilyCode = #FamilyCodeMaskValue)
AND (#CategoryCodeMaskValue IS NULL OR BR.CategoryCode = #CategoryCodeMaskValue)
AND (#SubCategoryCodeMaskValue IS NULL OR BR.SubCategoryCode = #SubCategoryCodeMaskValue)
AND (#NameMaskType IS NULL AND #NameMaskValue IS NULL
OR ((#NameMaskType = 1 AND BR.ItemName LIKE #NameMaskValue + '%')
OR (#NameMaskType = 2 AND BR.ItemName LIKE '%' + #NameMaskValue)
OR (#NameMaskType = 3 AND BR.ItemName LIKE '%' + #NameMaskValue + '%')
))
GROUP BY
BR.ItemName
)
INSERT INTO #Temp (
Id,
ItemName,
Supplier,
Quantity,
ItemUnit,
EntityUnit,
ItemSize,
PackageSize,
PurchaseValue,
UnitPurchaseValue,
PackagePurchaseValue,
CurrencyCode
)
SELECT BR.Id,
BR.ItemName,
BR.Supplier,
BR.Quantity,
BR.ItemUnit,
BR.EntityUnit,
BR.ItemSize,
BR.PackageSize,
BR.ItemGroup,
BR.UnitPurchaseValue,
BR.PackagePurchaseValue,
C.IsoCode
FROM CTC t
JOIN dbo.BatchRecords BR ON t.Id = BR.Id
LEFT OUTER JOIN dbo.Currencies C ON C.Id = BR.CurrencyId
ORDER BY BR.Id
OFFSET #PageCount * #PageSize ROWS
FETCH NEXT #PageSize ROWS ONLY;