Join in SQL Server when some query doesn't have results - sql-server

Question

I have a table with stock product movements (MOVSTOCKS) in two warehouses (CodAlm). For simplify the question, I will focus on a single product (with CodArt = C5):
CodArt
DescArt
CodAlm
UnidadesStock
EntranStock
SalenStock
FecDoc
TipDoc
C5
Palet
1
16
16
0
2021-12-31
IN
C5
Palet
2
0
0
0
2021-12-31
IN
C5
Palet
1
3
0
3
2022-01-11
SL
C5
Palet
1
4
0
4
2022-01-20
SL
C5
Palet
1
7
7
0
2022-02-01
EN
C5
Palet
1
6
0
6
2022-02-14
SL
C5
Palet
1
9
9
10
2022-05-01
IN
C5
Palet
2
1
1
0
2022-05-01
IN
C5
Palet
1
2
0
2
2022-06-10
SL
I need to get the stock on a certain day. For this, is necessary obtain stock quantity of the last inventory (TipDoc = IN) and add it purchases quantities (TipDoc = EN) and subtract the sales (TipDoc = SL).
I tried this query:
SELECT MV.CODART, MV.DESCART, MV.CODALM, SC.UNIDADESSTOCK + SUM(MV.ENTRANSTOCK) - SUM(MV.SALENSTOCK) as STOCK
FROM MOVSTOCKS MV
JOIN ( SELECT MV1.CODART, MV1.CODALM, MV2.FECDOC, MV1.UNIDADESSTOCK
FROM MOVSTOCKS MV1
JOIN ( SELECT CODART, CODALM, MAX(FECDOC) FECDOC
FROM MOVSTOCKS
WHERE TIPDOC = 'IN'
GROUP BY CODART, CODALM) MV2
ON MV1.CODART = MV2.CODART AND MV1.CODALM = MV2.CODALM AND MV1.FECDOC = MV2.FECDOC
WHERE MV1.TIPDOC = 'IN' ) SC
ON MV.CODART = SC.CODART AND MV.CODALM = SC.CODALM AND MV.FECDOC > SC.FECDOC
WHERE MV.CODART = 'C5' and MV.FECDOC <= '2022-06-01'
GROUP BY MV.CODART, MV.DESCART, MV.CODALM, SC.UNIDADESSTOCK
ORDER BY MV.CODART, MV.CODALM
With above data example and the query I expected to get following results:
CodArt
DescArt
CodAlm
Stock
C5
Palet
1
9
C5
Palet
2
1
The problem is that after the last inventory (2022-05-01) there have been no movements and then the join query get 0 rows because the filter MV.FECDOC <= '2022-06-01' in the WHERE doesn't get rows. I could modify the 'ON' condition in the join to MV.FECDOC >= SC.FECDOC and then get at least the inventory row, but I shouldn't do that because on inventory day there might be other previous movements that I shouldn't get for stock calculation.
Moreover, I will have the same problem if I want to get stock in a date for a product without inventory movements, because subquery 'SC' won't get rows.
Any help, please?

Answer 1

If you know that the inventory records exist, but the other transaction records might or might not exist afterwards, you will need to select your inventory records first and then LEFT JOIN the other transaction records.
If we only had to select one inventory record, we could start with a SELECT TOP 1 * ... ORDER BY FECDOC DESC with appropriate additional conditions. However, to get the latest inventory record for each of multiple warehouses, it get a bit more complicated. Two approaches that come to mind.
One would be to first select distinct warehouse codes (or perhaps distinct warehouse and product codes) and then CROSS APPLY a subselect to retrieve the latest inventory record for each.
The other approach would be to use the ROW_NUMBER() window function to number the inventory records by descending date (partitioned by warehouse codes and product codes) and then exclude all but row number = 1.
In either case, the next step would be to LEFT JOIN the transaction records, apply a GROUP BY, and SUM() up the results. Since SUM() returns NULL if there are no elements to sum, we need to use the ISNULL() function to assign a default value of zero.
The following two similar queries demonstrate the above approaches.
DECLARE #CODART VARCHAR(10) = 'C5'
DECLARE #AsOfDate DATETIME2 = '2022-06-01'
-- Method 1: SELECT DISTINCT followed by a cross apply
SELECT
SC.CODART, SC.DESCART, SC.CODALM,
SC.UNIDADESSTOCK + ISNULL(SUM(MV2.ENTRANSTOCK), 0) - ISNULL(SUM(MV2.SALENSTOCK), 0) as STOCK
FROM (
SELECT DISTINCT MV.CODART, MV.CODALM
FROM #MOVSTOCKS MV
WHERE MV.CODART = #CODART
) A
CROSS APPLY (
-- Most recent inventory for each selected CODART, CODALM
SELECT TOP 1 MV1.*
FROM #MOVSTOCKS MV1
WHERE MV1.CODART = A.CODART
AND MV1.CODALM = A.CODALM
AND MV1.TIPDOC = 'IN'
AND MV1.FECDOC <= #AsOfDate
ORDER BY MV1.FECDOC DESC
) SC
LEFT JOIN #MOVSTOCKS MV2
ON MV2.CODART = SC.CODART
AND MV2.CODALM = SC.CODALM
AND MV2.TIPDOC IN ('EN', 'SL')
AND MV2.FECDOC > SC.FECDOC
AND MV2.FECDOC <= #AsOfDate
GROUP BY SC.CODART, SC.DESCART, SC.CODALM, SC.UNIDADESSTOCK
ORDER BY SC.CODART, SC.CODALM
-- Method 2: Using the ROW_NUMBER() window function
SELECT
SC.CODART, SC.DESCART, SC.CODALM,
SC.UNIDADESSTOCK + ISNULL(SUM(MV2.ENTRANSTOCK), 0) - ISNULL(SUM(MV2.SALENSTOCK), 0) as STOCK
FROM (
-- Most recent inventory at or before #AsOfDate will have Recency = 1
SELECT MV1.*,
ROW_NUMBER() OVER(PARTITION BY CODART, CODALM ORDER BY FECDOC DESC) AS Recency
FROM #MOVSTOCKS MV1
WHERE MV1.FECDOC <= #AsOfDate
AND MV1.TIPDOC = 'IN'
) SC
LEFT JOIN #MOVSTOCKS MV2
ON MV2.CODART = SC.CODART
AND MV2.CODALM = SC.CODALM
AND MV2.FECDOC > SC.FECDOC
AND MV2.FECDOC <= #AsOfDate
AND MV2.TIPDOC IN ('EN', 'SL')
WHERE SC.Recency = 1
AND SC.CODART = #CODART
GROUP BY SC.CODART, SC.DESCART, SC.CODALM, SC.UNIDADESSTOCK
ORDER BY SC.CODART, SC.CODALM
Both queries above produce the desired result. See this db<>fiddle for a working demo.
*** UPDATE *** To handle cases where we might have transactions (added stock and/or sales) but no prior reference inventory, we will need to assume an initial inventory of zero and include all transactions since the beginning-of-time (start or data). This also requires making the initial SELECT DISTINCT the primary source for product and warehouse information. The CROSS APPLY becomes an OUTER APPLY (similar to a LEFT JOIN) and we need to make other adjustments to handle a possible null inventory record. That includes adjusting the date range for the transaction join.
The updated query would be something like:
SELECT
A.CODART, A.DESCART, A.CODALM,
ISNULL(SC.UNIDADESSTOCK, 0) + ISNULL(SUM(MV2.ENTRANSTOCK), 0) - ISNULL(SUM(MV2.SALENSTOCK), 0) as STOCK
, SC.UNIDADESSTOCK AS PriorInventory
, SUM(MV2.ENTRANSTOCK) AS NewStock
, SUM(MV2.SALENSTOCK) as Sales
FROM (
SELECT DISTINCT MV.CODART, MV.DESCART, MV.CODALM
FROM #MOVSTOCKS MV
WHERE MV.CODART = #CODART
) A
OUTER APPLY (
-- Most recent inventory for each selected CODART, CODALM combination
SELECT TOP 1 MV1.*
FROM #MOVSTOCKS MV1
WHERE MV1.CODART = A.CODART
AND MV1.CODALM = A.CODALM
AND MV1.TIPDOC = 'IN'
AND MV1.FECDOC <= #AsOfDate
ORDER BY MV1.FECDOC DESC
) SC
LEFT JOIN #MOVSTOCKS MV2
-- Transactions since inventory was last recorded
ON MV2.CODART = A.CODART
AND MV2.CODALM = A.CODALM
AND MV2.TIPDOC IN ('EN', 'SL') -- Alternately <> 'IN' or omitted entirely.
AND MV2.FECDOC > ISNULL(SC.FECDOC, '1900-01-01')
AND MV2.FECDOC <= #AsOfDate
GROUP BY A.CODART, A.DESCART, A.CODALM, SC.UNIDADESSTOCK
ORDER BY A.CODART, A.CODALM
See this db<>fiddle for an updated demo. For this demo, I added a new stock record for warehouse 3 and set the #AsOfDate to '2022-12-31'. I also added details to the result showing the separate initial inventory, new stock, and sales that feed the final STOCK calculation.
CODART
DESCART
CODALM
STOCK
PriorInventory
NewStock
Sales
C5
Palet
1
7
9
0
2
C5
Palet
2
1
1
null
null
C5
Palet
3
6
null
6
0
The above includes cases having inventory with transactions, inventory but no transactions, and transactions with no initial inventory.
I abandoned the ROW_NUMBER() alternate approach.