■ All corresponding columns in each SELECT columnlist must have the same, or compatible, datatypes. With reference to these rules, notice that both SELECT statements in the search have three columns. Each of the three columns has data in the same order and with the same datatype. When using the UNION, you should use column aliases to give the same column name to all corresponding columns. In our example, the first column of the first SELECT has an original name of OrderDate. The first column of the second SELECT has an original name of ReturnDate. To ensure that the first column in the final result has the desired name, both OrderDate and ReturnDate are given a column alias of Date. This also allows the column to be referenced in an ORDER BY columnlist. Also notice that the second column of each SELECT utilizes literal values. We created a calculated column named Type, which has a value of either Order or Return. This allows us to tell which table each row comes from. Finally, notice that the ORDER BY clause applies to the final results of both queries combined together. This is how it should be, since there would be no point to applying a sort to the indi vidual queries. At this point, it is useful to step back and talk about why it was necessary to employ the UNION operator rather than simply join the Orders and Returns tables together in a single SELECT statement. Since both tables have a CustomerID column, why didn’t we simply join the two tables together on this column? The problem with this possibility is tha t the two tables are really only indirectly related to each. Customers can place orders and customers can initiate returns, but there is no direct connection between orders and returns. Additionally, even if there were a direct connection between the two tables, a join would not accomplish what is desired. With a proper join, related information can be placed together on the same row. In this case, however, we are interested in showing orders and returns on separate rows. The UNION operator must be used to display data in this manner. In essence, the UNION allows us to retrieve unrelated or partially related data in a single statement. Chapter 15 ■ Set Logic156 Distinct and Non-Distinct Unions There are actuall y two variations of the UNION operator: UNION and UNION ALL . There is only a slight difference between the two. The UNION operator eliminates all duplicate rows. The UNION ALL operator specifies that all rows are to be included, even if they are duplicates. The UNION operator eliminates duplicates in a manner similar to the DISTINCT keyword previously seen. Whereas DISTINCT applies to a single SELECT, the UNION eliminates duplicates in all SELECT statements combined together via the UNION. In the previous example with the Orders and Returns tables, there was no pos- sibility of duplication, so it didn’t matter which was used. Here’s an example that illustrates the difference. Let’s say that you are only interested in the dates on which any orders or returns were issued. You don’t want to see multiple rows for the same date. The following statement accomplishes this task: SELECT OrderDate AS 'Date' FROM Orders UNION SELECT ReturnDate AS 'Date' FROM Returns Order by Date The resulting data is: Date 2009-10-13 2009-10-23 2009-12-05 2009-12-07 2009-12-15 2009-12-28 Notice that there is only one occurrence of 2009-12-28. Even though there is one row with 2009-12-28 in the Orders table and one row with 2009-12-28 in the Returns table, the UNION operato r ensures that the 2009-12-28 date is only lis- ted once. Distinct and Non-Distinct Unions 157 Let’s change the statement, adding a DISTINCT to each individual SELECT, but also specifying UNION ALL rather than UNION, as follows: SELECT DISTINCT OrderDate AS 'Date' FROM Orders UNION ALL SELECT DISTINCT ReturnDate AS 'Date' FROM Returns ORDER BY Date The outpu t is now: Date 2009-10-13 2009-10-23 2009-12-05 2009-12-07 2009-12-15 2009-12-28 2009-12-28 The DISTINCT ensures that each order date or return date is only listed once. Even though there are two orders from 2009-10-13, that date is only shown one time. However, the UNION ALL allows duplicates between the Orders SELECT and the Returns SELECT. So you can see that 2009-12-28 is listed twice, once from the Orders table and once from the Returns table. Intersecting Queries The UNION and UNION ALL operators return data that is in either of the sets specified in the two SELECT statements being combined. This is like using an OR operator to combine data from two logical sets. SQL provides an operator called INTERSECT, which only pulls data that is in both of the two sets being looked at. The INTERSECT is analogous to the AND operator and handles the second scenario stated at the start of the chapter: ■ Data that is in both SET A and SET B Chapter 15 ■ Set Logic158 DATABASE DIFFERENCES: MySQL MySQL doesn’t support the INTERSECT operator. Using the same Orders and Returns tables, let’s say that you want to see dates on which there are both orders and returns. A statement that accomplishes this is: SELECT OrderDate AS 'Date' FROM Orders INTERSECT SELECT ReturnDate AS 'Date' FROM Returns ORDER BY Date The result is: Date 2009-12-28 Only one row is shown because this is the only date that appears in both the Orders and Returns tables. There is one additional variation on the intersect operation, which is provided by the EXCEPT operator. Whereas the INTERSECT returns data that is in both sets, the EXCEPT returns data that is in one set but not the other and handles the third and fourth scenarios stated at the start of the chapter: ■ Data that is in SET A, but not in SET B ■ Data that is in SET B, but not in SET A The general format of the EXCEPT is: SelectStatementOne EXCEPT SelectStatementTwo ORDER BY columnlist Intersecting Queries 159 This statement will show data that is in SelectStatementOne but not in Select- StatementTwo. Here’s an example: SELECT OrderDate AS 'Date' FROM Orders EXCEPT SELECT ReturnDate AS 'Date' FROM Returns ORDER BY Date The result is: Date 2009-10-13 2009-12-05 2009-12-15 This data shows dates on which orders were placed, but on which no refunds were issued. Notice that 2009-12-28 does not appear, since a refund was issued on that date. DATABASE DIFFERENCES: MySQL and Oracle MySQL doesn’t support the EXCEPT operator. The equivalent of the EXCEPT operator in Oracle is MINUS. Looking Ahead In this chapter, we’ve seen different ways to combine multiple sets of SELECT statements into a single statement. The most commonly used operator is the UNION, which allows you to combine data that is in either of two different sets. The UNION is analogous to the OR operator. The UNION ALL is a variant of the UNION that allows duplicate rows to be shown. Similarly, the INTERCEPT operator allows that data to be presented if it is in both of the two sets of data being combined. The INTERCEPT is analogous to the AND operator. Finally, the EXCEPT operator allows for selection of data that is in one set but not in another. Chapter 15 ■ Set Logic160 . first SELECT has an original name of OrderDate. The first column of the second SELECT has an original name of ReturnDate. To ensure that the first column in the final result has the desired name, both OrderDate. only pulls data that is in both of the two sets being looked at. The INTERSECT is analogous to the AND operator and handles the second scenario stated at the start of the chapter: ■ Data that is. The most commonly used operator is the UNION, which allows you to combine data that is in either of two different sets. The UNION is analogous to the OR operator. The UNION ALL is a variant of the UNION