The line between the Customers and Orders tables is on the CustomerID columns because the CustomerID is the common link between these two tables. Similarly, the line between the Orders and Refunds tables is on the OrderID columns because the OrderID is the common link between these two tables. In other words, the Orders table is related to the Customers table by custome r. There must be a customer for an order to exist. The Refunds table is related to the Orders table by the order. There must be an order before a refund is issued. Note that the Refunds table is not directly related to the Customers table. How- ever, by joining all three tables together, we will be able to determine which customer a given refund was for. Let’s now examine the contents of each table. The Customers table has these values: CustomerID FirstName LastName 1 William Smith 2 Natalie Lopez 3 Brenda Harper 4 Adam Petrie The Orders table has this data: OrderID CustomerID OrderDate OrderAmount 1 1 2009-09-01 10.00 2 2 2009-09-02 12.50 3 2 2009-10-03 18.00 4 3 2009-09-15 20.00 The Refunds table contains this data: RefundID OrderID RefundDate RefundAmount 1 1 2009-09-02 5.00 2 3 2009-10-12 18.00 The Outer Join 121 Notice that three out of the four customers have placed orders. Likewise, only two refunds have been issued for the four orders placed. Left Joins Let’s now create a SELECT statement that joins all three tables together, using a LEFT JOIN: SELECT Customers.FirstName AS 'First Name', Customers.LastName AS 'Last Name', Orders.OrderDate AS 'Order Date', Orders.OrderAmount AS 'Order Amt', Refunds.RefundDate AS 'Refund Date', Refunds.RefundAmount AS 'Refund Amt' FROM Customers LEFT JOIN Orders ON Customers.CustomerID ¼ Orders.CustomerID LEFT JOIN Refunds ON Orders.OrderID ¼ Refunds.OrderID ORDER BY Customers.CustomerID, Orders.OrderID, RefundID The resulting data looks like: First Name Last Name Order Date Order Amt Refund Date Refund Amt William Smith 2009-09-01 10.00 2009-09-02 5.00 Natalie Lopez 2009-09-02 12.50 NULL NULL Natalie Lopez 2009-10-03 18.00 2009-10-12 18.00 Brenda Harper 2009-09-15 20.00 NULL NULL Adam Petrie NULL NULL NULL NULL DATABASE DIFFERENCES: Oracle Unlike SQL Server and MySQL, Oracle typically displays dates in a DD-MMM-YY format. For example, the date 2009-09-02 in the previous table will display as 02-SEP-09 in Oracle. However, no matter which database you’re using, the exact format in which dates are displayed will vary, depending on how your database was set up. Chapter 12 ■ Combining Tables with an Outer Join122 Before analyzing the previous SELECT statement, notice that there are two interesting aspects of the data you can see. First, Adam Petrie has no data shown other than his name. The reason for the lack of data is that there are no rows in the Orders table associated with that customer. The power of the outer join comes from the fact that you can see some data for Adam Petrie, even if he has no orders. If we had specified an INNER JOIN rather than a LEFT JOIN, you would see no rows at all for Adam. Similarly, you can see no refund data for either the 9/2/2009 order from Natalie Lopez or the order from Brenda Harper—because there are no rows in the Re- funds table associated with those orders. If we had specified an INNER JOIN rather than a LEFT JOIN, you would see no rows for those two orders. Let’s now look at the SELECT statement itself. The first few lines that specify the columns are nothing that you haven’t seen before. Notice that rather than using table aliases, we’re listing all the columns with their full names, with the table as a prefix. The first table listed is the Customers table. This table is shown after the FROM keyword. The second table shown is the Orders table, which appears after the first LEFT JOIN keyword. The subsequent ON clause specifies how the Orders table is linked to the Customers table. The third table shown is the Refunds table, which appears after the second LEFT JOIN keyword. The subsequent ON clause specifies how the Refunds table is joined to the Orders table. It is critical to realize that the order in which tables are listed in reference to the LEFT JOIN keyword is significant. When specifying a LEFT JOIN, the table to the left of LEFT JOIN is always the primary table. The table to the right of the LEFT JOIN is the secondary table. When joining between the primary and secondary tables, we want all rows in the primary table, even if there are no matches with any rows in the secondary table. In the first specified LEFT JOIN, the Customers table is on the left and the Orders table is on the right of the LEFT JOIN, which signifies that Customers is primary and Orders is secondary. In other words, we want to see all selected data from the Customers table, even if there isn’t a corresponding match in the secondary table for that row. Similarly, in the second LEFT JOIN, the Orders table is to the left and the Refunds table is to the right. That means that we are specifying Orders as primary Left Joins 123 and Refunds as secondary in this join. We want all orders, even if there are no matching refunds for that order. Finally, we included an ORDER BY clause. Notice that the fields specified in the ORDER BY are not selected in the original columnlist. Testing for NULL Values In the previous SELECT, we had one customer with no orders and two orders with no associated refunds. Unlike the INNER JOIN, the LEFT JOIN allows these rows with missing values to appear. To test your understanding of the LEFT JOIN, let’s now ask how we would list only those orders for which no refund was issued. The solution involves adding a WHERE clause that tests for NULL values, as follows: SELECT Customers.FirstName AS 'First Name', Customers.LastName AS 'Last Name', Orders.OrderDate AS 'Order Date', Orders.OrderAmount AS 'Order Amt' FROM Customers LEFT JOIN Orders ON Customers.CustomerID ¼ Orders.CustomerID LEFT JOIN Refunds ON Orders.OrderID ¼ Refunds.OrderID WHERE Orders.OrderID IS NOT NULL AND Refunds.RefundID IS NULL ORDER BY Customers.CustomerID, Orders.OrderID The resulting data is: First Name Last Name Order Date Order Amt Natalie Lopez 2009-09-02 12.50 Brenda Harper 2009-09-15 20.00 The WHERE clause first tests Orders.OrderID to make sure that it isn’t NULL. Doing so ensures that you don’t see customers who never placed an order. The second portion of the WHERE clause tests Refunds.RefundID to make sure that it is NULL. This ensures that you only see orders that don’t match a refund. Chapter 12 ■ Combining Tables with an Outer Join124 Right Joins The previous SELECT statements utilized the LEFT JOIN keyword. The good news about right joins is that they are identical in concept to the left join. The only difference between left and right joins is the order in which the two tables in the join are listed. In left joins, the primary table is listed to the left of the LEFT JOIN keyword. The secondary table, which may or may not contain matching rows, is listed to the right of the LEFT JOIN keyword. In right joins, the primary table is listed to the right of the RIGHT JOIN keyword. The secondary table is listed to the left of the RIGHT JOIN keyword. That’s the only difference. The FROM clause of the previous SELECT statement was: FROM Customers LEFT JOIN Orders ON Customers.CustomerID ¼ Orders.CustomerID LEFT JOIN Refunds ON Orders.OrderID ¼ Refunds.OrderID If we desire to restate this using right joins, we could change it to: FROM Refunds RIGHT JOIN Orders ON Orders.OrderID ¼ Refunds.OrderID RIGHT JOIN Customers ON Customers.CustomerID ¼ Orders.CustomerID The point to note is that it’s only the order in which tables are listed before and after the RIGHT JOIN that matters. The order in which columns are listed after the ON keyword has no significance. Basically, this means that it’s completely unnecessary to ever use the RIGHT JOIN keyword. Anything that can be specified with a RIGHT JOIN can be specified with a LEFT JOIN. Our suggestion is to stick with the LEFT JOIN, since we intuitively tend to think in terms of listing more important, or primary, tables first. Table Order in Outer Joins We previously noted that the order in which tables were specified in an inner join was not material. The same is not true of outer joins, since the order that tables Table Order in Outer Joins 125 . is listed to the right of the LEFT JOIN keyword. In right joins, the primary table is listed to the right of the RIGHT JOIN keyword. The secondary table is listed to the left of the RIGHT JOIN. table, even if there are no matches with any rows in the secondary table. In the first specified LEFT JOIN, the Customers table is on the left and the Orders table is on the right of the LEFT JOIN,. left of LEFT JOIN is always the primary table. The table to the right of the LEFT JOIN is the secondary table. When joining between the primary and secondary tables, we want all rows in the primary