Return one SQL row per product with price and latest date

Question

I am not a SQL guy, I have used it in the past and rarely have an issue that cant be solved by google... however this time I need to ask the Community.

I have a database

Answer 1

You can use a CTE and the ranking function PARTITION BY:

WITH CTE AS
(
    select t.ProdNo, t.TransactionDate as 'LastPurchaseDate', t.Price,
           rn = row_number() over (partition by ProdNo order by TransactionDate desc)
    from Transactions t
)
SELECT ProdNo, LastPurchaseDate, Price  FROM CTE WHERE RN = 1

Answer 2

Try this. by using row_number().

             select * from 
             (
             select 
             T.ProdNo, 
             T.TransactionDate as 'LastPurchaseDate', 
             T.Price,           
             row_number() over (partition by ProdNo order by TransactionDate desc) as rnk
             from Transactions T
             )a
             where rnk='1'

Answer 3

SELECT A.ProdNo ,Price ,A.TransactionDate,PurchasedBy 
FROM #Transactions 
JOIN 
(
SELECT ProdNo,MAX(TransactionDate) TransactionDate
FROM #Transactions 
GROUP BY ProdNo
)A ON A.TransactionDate = #Trans.TransactionDate  

Answer 4

Pls Try this --If one ProdNo has two same max TransactionDate, will return both two rows

select ProdNo,Price,TransactionDate from Transactions t 
where not exists (select 1 from Transactions where ProdNo=t.ProdNo and TransactionDate>t.TransactionDate)

Answer 5

You're on the right track. What if you use:

select t.ProdNo, t.TransactionDate as 'LastPurchaseDate', t.Price
from Transactions t
inner join (
     select ProdNo, max(TransactionDate) as 'LastPurchaseDate'
     from Transactions
     WHERE Price > 0
     group by ProdNo
 ) tm on t.ProdNo = tm.ProdNo and t.TransactionDate= tm.LastPurchaseDate

Note the change in join conditions.

What happened in your query: LastPurchaseDate = tm.LastPurchaseDate. There is only one column called LastPurchaseDate, so it's equating it with itself, which is always true. So you're left with t.ProdNo = tm.ProdNo, since t.ProdNo is not unique, you get multiple records returned for each t.ProdNo.

Answer 6

One method uses the ROW_NUMBER windowed function.

My query uses a common table expression (CTE) to provide sample data. When using this technique you always need a CTE, or subquery. This is because values generated in the SELECT clause are not available to the WHERE clause. This is a consequence of something called the Logical Processing Order. In other words; SQL Server generates the row numbers after it has filtered the data. CTEs/Subqueries provide you with a second WHERE clause, that is actioned after the row numbers have been generated**.

-- Returning the most recent record from a transaction table.
WITH SampleDate AS
    (
        SELECT
            ROW_NUMBER() OVER (PARTITION BY ProdNo ORDER BY TransactionDate DESC) AS Rn,
            *
        FROM
            (
                VALUES                      
                    ('3STRFLEX', 13.02, '20162911', 'AWC '),
                    ('3STRFLEX', 15.02, '20162011', 'DWC '),
                    ('3STRFLEX', 15.02, '20160101', 'AWC '),
                    ('AFTV2'   , 35.49, '20162708', 'AWC '),
                    ('AFTV2'   , 29.99, '20160106', 'DWC '),
                    ('AFTV2'   , 29.99, '20160205', 'AWC ')
            ) AS x(ProdNo, Price, TransactionDate, PurchasedBy)
    )
SELECT
    *
FROM
    SampleDate
WHERE
    Rn = 1
;

** Actually this isn't entirely true. It is called the logical order for a reason. SQL Sever can/will exeucte your queries any way it sees fit. But however your query is physically executed it will respect the logical order.