How to join tables together based on subselects? - sql

Question

I'm stuck to figure out how to write a query. Basically I've three tables (Orders, Products, Orders_Products) which I want to join together and apply some filtering.
Orders table:
ORDER_ID CUSTOMER_ID
1 1
2 2
Products table:
PRODUCT_ID PRODUCT_NAME PRODUCT_TITLE
1 'P1' 'T1'
2 'P1' 'T2'
3 'P2' 'T3'
4 'P2' 'T4'
5 'P2' 'T5'
6 'P3' 'T6'
Orders_Products table:
ORDER_ID PRODUCT_ID
1 1
1 3
2 1
2 3
2 6
For example I want to get all Orders which consists (exactly) of the products P1/T1 and P2/T3. I tried something like this, but that doesn't work:
SELECT * FROM Orders
LEFT JOIN Orders_Products ON Orders_Products.ORDER_ID = Orders.ORDER_ID
LEFT JOIN Products ON Orders_Products.PRODUCT_ID = Products.PRODUCT_ID
WHERE EXISTS (SELECT * FROM Product WHERE PRODUCT_NAME = 'P1' AND PRODUCT_TITLE = 'T1')
AND EXISTS (SELECT * FROM Product WHERE PRODUCT_NAME = 'P2' AND PRODUCT_TITLE = 'T3');
EDIT: To clarify what I really have to achieve. The user should be able to search for orders matching the given products. The user enters one or more product name / product title combinations and gets all the orders which have exactly this products associated. What I get (from a web application) are only the name/title combinations and I have to use those in a query to get the ORDER_ID.

Answer 1

SELECT OrderID, COUNT(*) AS ProductsCount
FROM Orders_Products
WHERE (PRODUCT_ID = 1 OR PRODUCT_ID = 3)
GROUP BY OrderID
HAVING COUNT(*) = 2
EDIT: Please ignore the above statement. See if the following works.
SELECT OrderID,
SUM(CASE PRODUCT_ID WHEN 1 THEN 1 WHEN 3 THEN 1 ELSE 3 END)
AS ProductsCount
FROM Orders_Products
GROUP BY OrderID
HAVING SUM(CASE PRODUCT_ID WHEN 1 THEN 1 WHEN 3 THEN 1 ELSE 3 END) = 2
I guess this should get you Orders which has only these 2 products.

Answer 2

You probably cannot write simple queries in MySQL to achieve this. But ANSI SQL supports table value constructor which simplifies this type of query.
This basic query returns the full list of orders (5 rows):
SELECT * FROM Products
JOIN Orders_Products ON Orders_Products.PRODUCT_ID = Products.PRODUCT_ID
JOIN Orders ON Orders_Products.ORDER_ID = Orders.ORDER_ID
This query with table value constructor returns the orders that you need:
SELECT * FROM Products
JOIN Orders_Products ON Orders_Products.PRODUCT_ID = Products.PRODUCT_ID
JOIN Orders ON Orders_Products.ORDER_ID = Orders.ORDER_ID
LEFT JOIN (VALUES('P1', 'T1'), ('P2', 'T3')) V(P_NAME, P_TITLE) ON PRODUCT_NAME = P_NAME AND PRODUCT_TITLE=P_TITLE
This query groups the above to returns the ORDER_ID where there is no order outside the required list (eliminating the orders that have rows containing null):
SELECT ORDER_ID FROM Products
JOIN Orders_Products ON Orders_Products.PRODUCT_ID = Products.PRODUCT_ID
JOIN Orders ON Orders_Products.ORDER_ID = Orders.ORDER_ID
LEFT JOIN (VALUES('P1', 'T1'), ('P2', 'T3')) V(P_NAME, P_TITLE) ON PRODUCT_NAME = P_NAME AND PRODUCT_TITLE=P_TITLE
GROUP BY ORDER_ID HAVING COUNT(*) = 2
Among open source databases, HSQLDB is one that supports table value constructor and other user friendly features of ANSI SQL:2008