SQL to return distinct rows with a max value of column? Complicated by duplicates of every other column - duplicates

Question

I have 1 table, it contains 6 columns (DivisionId, ItemId, BuyerCode, BuyerName, ReceivedPounds, POIssueDate) and all 6 are not unique. My DISTINCT value I eventually want is a CONCAT of DivisionID_ItemId. I have taken a lot of attempts at this already, I want to return every unique row of DivisionId and ItemId and the BuyerCode with the most pounds for that DivisionId ItemId combo. My complication is I have multiple BuyerCodes for most itemid's and I have multiple itemid's for different divisionid's. I have gotten close, but my final result still gives me every buyer for every item. Here is my current query in SSMS. I'm a self taught hack so feel free, I tried to break out each subquery. T1 sums the pounds, FT is the MAX function, T2 is to join buyer to MAX div and item. Problem is I still get multiple occurrences of divisionId and ItemId in my result. I am lacking the way to make the result as distinct as the FT subquery.
SELECT FT.DivisionId,FT.ItemId, T2.BuyerCode, T2.BuyerName
FROM( SELECT DISTINCT T1.DivisionId,T1.ItemId, MAX(T1.SUMPOUNDS) as MaxPounds
FROM( SELECT DISTINCT TCO.DivisionId, TCO.ItemId, TCO.BuyerCode, TCO.BuyerName, SUM(TCO.ReceivedPounds) as SUMPOUNDS
FROM [GTDev].[dbo].[TCO_FinalData_ABC] TCO
WHERE PO_Issue_Date > (GETDATE()-90) AND ReceivedPounds > 0
GROUP BY TCO.DivisionId, TCO.ItemId, TCO.BuyerCode, TCO.BuyerName) as T1
GROUP BY T1.DivisionId,T1.ItemId) as FT
LEFT JOIN ( SELECT DISTINCT T2.DivisionId, T2.ItemId, T2.BuyerCode, T2.BuyerName, SUM(T2.ReceivedPounds) as SUMPOUNDS
FROM [GTDev].[dbo].[TCO_FinalData_ABC] T2
WHERE PO_Issue_Date > (GETDATE()-90) AND ReceivedPounds > 0
GROUP BY T2.DivisionId, T2.ItemId, T2.BuyerCode, T2.BuyerName) as T2
ON FT.DivisionId = T2.DivisionId AND FT.ItemId = T2.ItemId AND FT.MaxPounds = T2.SUMPOUNDS
GROUP BY FT.DivisionId,FT.ItemId, T2.BuyerCode, T2.BuyerName
TL;DR - Need to SUM pounds for every unique DivId, ItemId, Buyer and then select the MAX pounds of every DivId and ItemId and return the Buyer with the most pounds for each DivId ItemId combo.
Thank you in advance! Feel free to tell me what is wrong and not rewrite the code unless you want to. Not looking for someone to do my work, just advice to get over this issue!

Answer 1

I would create an id with the DivId and ItemId, like:
select DivId||'_'||ItemId as indicator, {other columns} from {your tables and conditions}
Then I would run a query on top of it, like:
select sum(pounds), indicator, buyer from ({the previous query})
group by indicator,buyer, then select the max, and rank it (this is the trick that you are missing)
Look at this whole sequence, I think this is what you need: (I broke this down in several pieces, you can simplify it or do it like this)
select * from (
select distinct indicator, max(ReceivedPoundsSum), BuyerCode,RANK() OVER ( PARTITION BY indicator ORDER BY max(ReceivedPoundsSum) DESC ) AS "Rank" from (
select sum(ReceivedPounds) ReceivedPoundsSum, indicator, BuyerCode,BuyerName from (
with data as(
select 1 DivisionId, 1 ItemId,1 BuyerCode,'user1' BuyerName,55 ReceivedPounds,sysdate POIssueDate from dual union
select 1 DivisionId, 1 ItemId,3 BuyerCode,'user3' BuyerName,15 ReceivedPounds,sysdate POIssueDate from dual union
select 2 DivisionId, 2 ItemId,1 BuyerCode,'user1' BuyerName,25 ReceivedPounds,sysdate POIssueDate from dual union
select 2 DivisionId, 2 ItemId,2 BuyerCode,'user2' BuyerName,35 ReceivedPounds,sysdate POIssueDate from dual union
select 2 DivisionId, 2 ItemId,3 BuyerCode,'user3' BuyerName,45 ReceivedPounds,sysdate POIssueDate from dual union
select 3 DivisionId, 3 ItemId,5 BuyerCode,'user5' BuyerName,55 ReceivedPounds,sysdate POIssueDate from dual union
select 1 DivisionId, 3 ItemId,3 BuyerCode,'user3' BuyerName,5 ReceivedPounds,sysdate POIssueDate from dual union
select 4 DivisionId, 4 ItemId,4 BuyerCode,'user4' BuyerName,1 ReceivedPounds,sysdate POIssueDate from dual union
select 5 DivisionId, 4 ItemId,4 BuyerCode,'user4' BuyerName,12 ReceivedPounds,sysdate POIssueDate from dual union
select 6 DivisionId, 5 ItemId,4 BuyerCode,'user4' BuyerName,13 ReceivedPounds,sysdate POIssueDate from dual union
select 6 DivisionId, 5 ItemId,1 BuyerCode,'user1' BuyerName,14 ReceivedPounds,sysdate POIssueDate from dual union
select 6 DivisionId, 6 ItemId,2 BuyerCode,'user2' BuyerName,16 ReceivedPounds,sysdate POIssueDate from dual union
select 5 DivisionId, 10 ItemId,1 BuyerCode,'user1' BuyerName,157 ReceivedPounds,sysdate POIssueDate from dual)
select DivisionId||'_'||ItemId as indicator,a.* from data a
) group by indicator, BuyerCode,BuyerName
) group by indicator, BuyerCode
) where "Rank"=1
The result will look like this:
Hope this helps!