LEFT outer join returns returns duplicate rows even after adding distinct - sql

Question

Restricting the price change table to select a particular item
select * from price
where item = '13'
results of the query above
item Date_Changed New Old START_DATE end_DATE
13 01/11/2018 00:00 5.61 4.88 01/11/2018 00:00 30/11/2018 00:00
13 30/11/2018 00:00 2.84 5.61 01/11/2018 00:00 17/12/2018 00:00
13 17/12/2018 00:00 2.39 2.84 30/11/2018 00:00 17/12/2018 00:00
sales table
Date Item Qty Amount
05/07/2018 00:00 13 3 14.64
05/07/2018 00:00 13 3 14.64
04/07/2018 00:00 13 3 14.64
02/07/2018 00:00 13 1 4.88
02/07/2018 00:00 13 6 29.28
06/07/2018 00:00 13 7 34.16
03/07/2018 00:00 13 4 19.52
12/07/2018 00:00 13 2 9.76
10/08/2018 00:00 13 1 4.88
Sample code
SELECT distinct a.[Inv]
, (CASE
WHEN a.Date <= b.START_DATE THEN (b.Old * a.Qty)
WHEN a.Date between b.START_DATE and b.dt_end_DATE THEN (b.New * a.Qty)
ELSE 0
END) as calc_amount
,(a.[amount] - (CASE
WHEN a.Date <= b.START_DATE THEN (b.Old * a.Qty)
WHEN a.Date between b.START_DATE and b.end_DATE THEN (b.New * a.Qty)
ELSE 0
END)) as variance
[sales] a
left outer join price b
on a.[Item] = b.item
where b.item = '13'
The script then returns 27 rows instead of 9 rows. can someone assist on how i can improve my script to be more accurate

Answer 1

Use outer apply. I assume you want the most recent start date from price, so this looks like:
select s.*,
(case when s.date <= p.start_date
then p.Old * s.Qty
else p.New * s.Qty
end) as calc_amount
from sales s outer apply
(select top (1) p.*
from prices p
where p.item = s.item and
p.start_date <= s.date
order by p.date desc
) p

Answer 2

I am not sure if this is what you are looking for. but perhaps you can skip the join?
I created 2 sample data with the columns that you might need.
DECLARE #price table (item varchar(2),date_start date, new_price numeric(9,2))
Insert into #price (item , date_start,new_price)
values
( '13', '20190101', '1.00'),
( '13', '20190102', '1.01'),
( '13', '20190103', '1.02')
DECLARE #sales table (item varchar(2),date_sales date,qty int)
Insert into #sales (item , date_sales,qty)
values
( '13', '20190101', '5'),
( '13', '20190101', '2'),
( '13', '20190102', '5'),
( '13', '20190102', '2'),
( '13', '20190103', '5'),
( '13', '20190103', '2')
declare #item as varchar(2) = '13'
SELECT (select top (1) new_price from #price b where a.date_sales>=b.date_start and b.item = #item order by b.date_start desc ) * a.qty as 'new_price* qty'
from #sales a
where a.item = #item
I have not tested this in a table with a huge data set, so I also can't vouch for the speed of this query. I believe it would be better to have some kind of other ID to join the table

Answer 3

Your question is not clear... You added sample data, but I doubt this is correct...
Your price table is open to erronous data. It would be better to store just the price and a validFrom-date. In this case you can pick the price on a give date easily. Your format is open to overlapping periodes and there is no good reason to store the former price once again. That's why I ignore all fields you should not use...
Try this. I've changed the dates in a way to simulate validity periodes.
A mock-up scenario (please to this the next time for us):
CREATE TABLE priceMock(item INT, Date_Changed DATE, New DECIMAL(10,4), Old DECIMAL(10,4), [START_DATE] DATE, end_DATE DATE);
SET DATEFORMAT dmy;
INSERT INTO priceMock VALUES
(13,'01/11/2018 00:00',5.61,4.88,'01/07/2018 00:00','06/07/2018 00:00')
,(13,'30/11/2018 00:00',2.84,5.61,'07/07/2018 00:00','10/07/2018 00:00')
,(13,'17/12/2018 00:00',2.39,2.84,'11/07/2018 00:00','15/08/2018 00:00');
GO
CREATE TABLE salesMock ([Date] DATE, Item INT, Qty INT, Amount DECIMAL(10,4));
SET DATEFORMAT dmy;
INSERT INTO salesMock VALUES
('05/07/2018 00:00',13,3,14.64)
,('05/07/2018 00:00',13,3,14.64)
,('04/07/2018 00:00',13,3,14.64)
,('02/07/2018 00:00',13,1,4.88 )
,('02/07/2018 00:00',13,6,29.28)
,('06/07/2018 00:00',13,7,34.16)
,('03/07/2018 00:00',13,4,19.52)
,('10/07/2018 00:00',13,2,9.76 )
,('10/08/2018 00:00',13,1,4.88 );
GO
I'd add an inline-table-valued-function to get exactly one single line back.
CREATE FUNCTION dbo.GetPriceForItemOnDate(#item INT,#ValidOn DATE)
RETURNS TABLE
AS
RETURN
SELECT TOP 1 *
FROM priceMock
WHERE item=#item
AND [START_DATE] <= #ValidOn
ORDER BY [START_DATE] DESC
GO
--This query will combine your sales data with the price valid on the given date
SELECT s.[Date]
,s.Item
,s.Qty
,p.New AS CurrentPrice
,s.Qty * p.New AS ComputedAmount
FROM salesMock s
OUTER APPLY dbo.GetPriceForItemOnDate(s.item,s.[Date]) p
GO
--Clean up (carefull with real data)
DROP FUNCTION dbo.GetPriceForItemOnDate;
DROP TABLE priceMock;
DROP TABLE salesMock;
The idea in short:
The function will first filter to price lines for the given item. The second filter will cut the list and return just the prices for the given date and before the given date. As we sort this by the date in descending order we will get the latest price on top. By using TOP 1 we return just the one single line we want.
General remark: I use a validFrom-approach here. But you can turn this to the opposite and use a validTo-approach. The idea is the same.