SQL Repeated Outer Join Improving Efficiency - sql-server

Question

Probelm Statement:
Write a query to return 2015 sales information for each supplier. We would like to include all suppliers in the result set, regardless of whether their products were sold in 2015.
Sales are determined using Sales.Orders and Sales.OrderLines as in the previous two questions. However, since we are asking for this information from the perspective of the supplier, you also need to use the tables Warehouse.StockItems and Purchasing.Suppliers.
The columns required in the result set are:
SupplierID - As it appears in the table Purchasing.Suppliers.
SupplierName - As it appears in the table Purchasing.Suppliers.
OrderCount - The number of orders placed on products for each supplier.
Sales - The subtotal from the orders placed, calculated from Quantity and UnitPrice of the table Sales.OrderLines.
The results should be sorted such that the supplier with the highest sales is at the top. If two suppliers have the same sales, next use the order count with the highest count at the top. If two suppliers have the same sales and order count, use the supplier name in ascending order as the final tie breaker. This will ensure a deterministic result.
I am using the WorldWideImporters Microsoft sample database tables. I am trying to return the 2015 sales information for each supplier in Purchasing.Suppliers. I am returning the OrderCount and the Sum of the 2015 sales in respective columns. I am having trouble with joins here since I have to connect Suppliers to the Warehouse.StockItems and then connect these items to specific OrderLines which have a field for StockItemID.
The problem is that usually I would join orders to orderlines, so that I could filter only orders and thus orderlines in 2015. However, with the table structure that I have specified, it seems I have to connect OrderLines to Orders.
So what I did was to join those Orders back with OrderLines to provide the result I am used to. Here is my attempt at a solution:
<pre>
SELECT S.SupplierID
,S.SupplierName
,COUNT(DISTINCT O.OrderID) AS OrderCount
,ISNULL(SUM(OLP.Quantity * OLP.UnitPrice), 0.00) AS Sales
FROM Purchasing.Suppliers AS S
LEFT OUTER JOIN Warehouse.StockItems AS W ON S.SupplierID = W.SupplierID
LEFT OUTER JOIN Sales.OrderLines AS OL ON W.StockItemID = OL.StockItemID
LEFT OUTER JOIN Sales.Orders AS O ON OL.OrderID = O.OrderID
AND O.OrderDate BETWEEN '2015-01-01' AND '2015-12-31'
LEFT OUTER JOIN Sales.OrderLines AS OLP ON O.OrderID = OLP.OrderID
GROUP BY S.SupplierID
,S.SupplierName
ORDER BY Sales DESC
,OrderCount
,SupplierName;
</pre>
Edit:
Results:
Look to have every supplier as expected even ones that had no sales or orders. I am not sure if the calculated sales is correct though and I am not sure how to verify. Didn't know if anyone saw a flaw in my query.
I have no idea if this is correct or the most efficient way to solve this problem. I do have constraints that I can only use joins, no subqueries, unions, etc.
Any help in understanding would be appreciated. Thank you.

Answer 1

To benchmark the orders, without regard to suppliers or order lines:
/* query 1 */
SELECT
COUNT(*) AS ordercount
FROM Sales.Orders AS o
WHERE o.OrderDate >= '20150101' AND o.OrderDate < '20160101'
Then to benchmark the orders lines, without regard to suppliers:
/* query 2 */
SELECT
COUNT(DISTINCT o.OrderID) AS ordercount
, SUM(olp.Quantity * olp.UnitPrice) AS sales
FROM Sales.Orders AS o
INNER JOIN Sales.OrderLines AS olp ON olP.OrderID = o.OrderID
WHERE o.OrderDate >= '20150101' AND o.OrderDate < '20160101'
Now start introducing more joins, and if the values alter then the most recent join is at fault:
/* query 3 */
SELECT
COUNT(DISTINCT o.OrderID) AS ordercount
, SUM(olp.Quantity * olp.UnitPrice) AS sales
FROM Sales.Orders AS o
INNER JOIN Sales.OrderLines AS olp ON olP.OrderID = o.OrderID
INNER JOIN Warehouse.StockItems AS w ON w.StockItemID = olp.StockItemID
WHERE o.OrderDate >= '20150101' AND o.OrderDate < '20160101'
and then:
/* query 4 */
SELECT
COUNT(DISTINCT o.OrderID) AS ordercount
, SUM(olp.Quantity * olp.UnitPrice) AS sales
FROM Sales.Orders AS o
INNER JOIN Sales.OrderLines AS olp ON olP.OrderID = o.OrderID
INNER JOIN Warehouse.StockItems AS w ON w.StockItemID = olp.StockItemID
INNER JOIN Purchasing.Suppliers s ON s.SupplierID = w.SupplierID
WHERE o.OrderDate >= '20150101' AND o.OrderDate < '20160101'
You certainly don't need to join order lines twice, regarding the left joins it depends on what it is you are trying to achieve, e.g.:
Only suppliers with orders in the date range:
/* query 5 */
SELECT
s.SupplierID
, s.SupplierName
, COUNT(DISTINCT o.OrderID) AS ordercount
, ISNULL(SUM(olp.Quantity * olp.UnitPrice), 0.00) AS sales
FROM Purchasing.Suppliers s
INNER JOIN Warehouse.StockItems AS w ON s.SupplierID = w.SupplierID
INNER JOIN Sales.OrderLines AS olp ON w.StockItemID = olp.StockItemID
INNER JOIN Sales.Orders AS o ON olP.OrderID = o.OrderID
WHERE o.OrderDate >= '20150101'
AND o.OrderDate < '20160101' -- note: this is "the next" day
GROUP BY
s.SupplierID
, s.SupplierName
ORDER BY
sales DESC
, ordercount
, SupplierName;
All suppliers with stock references:
/* query 6 */
SELECT
s.SupplierID
, s.SupplierName
, COUNT(DISTINCT o.OrderID) AS ordercount
, ISNULL(SUM(olp.Quantity * olp.UnitPrice), 0.00) AS sales
FROM Purchasing.Suppliers s
INNER JOIN Warehouse.StockItems AS w ON s.SupplierID = w.SupplierID
LEFT JOIN Sales.OrderLines AS olp ON w.StockItemID = olp.StockItemID
LEFT JOIN Sales.Orders AS o ON olP.OrderID = o.OrderID
AND o.OrderDate >= '20150101'
AND o.OrderDate < '20160101' -- note: this is "the next" day
GROUP BY
s.SupplierID
, s.SupplierName
ORDER BY
sales DESC
, ordercount
, SupplierName;
Every supplier:
/* query 7 */
SELECT
s.SupplierID
, s.SupplierName
, COUNT(DISTINCT o.OrderID) AS ordercount
, ISNULL(SUM(olp.Quantity * olp.UnitPrice), 0.00) AS sales
FROM Purchasing.Suppliers s
LEFT JOIN Warehouse.StockItems AS w ON s.SupplierID = w.SupplierID
LEFT JOIN Sales.OrderLines AS olp ON w.StockItemID = olp.StockItemID
LEFT JOIN Sales.Orders AS o ON olP.OrderID = o.OrderID
AND o.OrderDate >= '20150101'
AND o.OrderDate < '20160101' -- note: this is "the next" day
GROUP BY
s.SupplierID
, s.SupplierName
ORDER BY
sales DESC
, ordercount
, SupplierName;
Please be very cautious about using between for date ranges, the most reliable way to define a date range is to use >= and < as shown above, this way it does not matter what the time precision of the data is. Also YYYYMMDD is the safest date literal format in TSQL.