## SQL Topic: Subqueries

You might have noticed that even with a complete query, there are many questions that we can’t answer about our data without additional post, or pre, processing. In these cases, you can either make multiple queries and process the data yourself, or you can build a more complex query using SQL subqueries.

#### Example: General subquery

Lets say your company has a list of all Sales Associates, with data on the revenue that each Associate brings in, and their individual salary. Times are tight, and you now want to find out which of your Associates are costing the company more than the average revenue brought per Associate.

First, you would need to calculate the average revenue all the Associates are generating:

 1 2  SELECT AVG(revenue_generated) FROM sales_associates;

And then using that result, we can then compare the costs of each of the Associates against that value. To use it as a subquery, we can just write it straight into the WHERE clause of the query:

 1 2 3 4 5  SELECT * FROM sales_associates WHERE salary > (SELECT AVG(revenue_generated) FROM sales_associates);

As the constraint is executed, each Associate’s salary will be tested against the value queried from the inner subquery.

A subquery can be referenced anywhere a normal table can be referenced. Inside a FROM clause, you can JOIN subqueries with other tables, inside a WHERE or HAVING constraint, you can test expressions against the results of the subquery, and even in expressions in the SELECT clause, which allow you to return data directly from the subquery. They are generally executed in the same logical order as the part of the query that they appear in, as described in the last lesson.

Because subqueries can be nested, each subquery must be fully enclosed in parentheses in order to establish proper hierarchy. Subqueries can otherwise reference any tables in the database, and make use of the constructs of a normal query (though some implementations don’t allow subqueries to use LIMIT or OFFSET).

### Correlated subqueries

A more powerful type of subquery is the correlated subquery in which the inner query references, and is dependent on, a column or alias from the outer query. Unlike the subqueries above, each of these inner queries need to be run for each of the rows in the outer query, since the inner query is dependent on the current outer query row.

#### Example: Correlated subquery

Instead of the list of just Sales Associates above, imagine if you have a general list of Employees, their departments (engineering, sales, etc.), revenue, and salary. This time, you are now looking across the company to find the employees who perform worse than average in their department.

For each employee, you would need to calculate their cost relative to the average revenue generated by all people in their department. To take the average for the department, the subquery will need to know what department each employee is in:

 1 2 3 4 5 6  SELECT * FROM employees WHERE salary > (SELECT AVG(revenue_generated) FROM employees AS dept_employees WHERE dept_employees.department = employees.department);

These kinds of complex queries can be powerful, but also difficult to read and understand, so you should take care using them. If possible, try and give meaningful aliases to the temporary values and tables. In addition, correlated subqueries can be difficult to optimize, so performance characteristics may vary across different databases.

### Existence tests

When we introduced WHERE constraints in Lesson 2: Queries with constraints, the IN operator was used to test whether the column value in the current row existed in a fixed list of values. In complex queries, this can be extended using subqueries to test whether a column value exists in a dynamic list of values.

 1 2 3 4 5  -- Select query with subquery constraint SELECT *, … FROM mytable WHERE column IN/NOT IN (SELECT another_column FROM another_table);

When doing this, notice that the inner subquery must select for a column value or expression to produce a list that the outer column value can be tested against. This type of constraint is powerful when the constraints are based on current data.

## SQL Topic: Unions, Intersections & Exceptions

When working with multiple tables, the UNION and UNION ALL operator allows you to append the results of one query to another assuming that they have the same column count, order and data type. If you use the UNION without the ALL, duplicate rows between the tables will be removed from the result.

 1 2 3 4 5 6 7 8  -- Select query with set operators SELECT column, another_column FROM mytable UNION / UNION ALL / INTERSECT / EXCEPT SELECT other_column, yet_another_column FROM another_table ORDER BY column DESC LIMIT n;

In the order of operations as defined in Lesson 12: Order of execution, the UNION happens before the ORDER BY and LIMIT. It’s not common to use UNIONs, but if you have data in different tables that can’t be joined and processed, it can be an alternative to making multiple queries on the database.

Similar to the UNION, the INTERSECT operator will ensure that only rows that are identical in both result sets are returned, and the EXCEPT operator will ensure that only rows in the first result set that aren’t in the second are returned. This means that the EXCEPT operator is query order-sensitive, like the LEFT JOIN and RIGHT JOIN.

UNION 类似，INTERSECT 会返回多个结果间共有的行，EXCEPT 则只会返回第一个结果中有而第二个结果中没有的行。所以 EXCEPT 是查询顺序敏感的操作，和 LEFT JOINRIGHT JOIN 一样。

Both INTERSECT and EXCEPT also discard duplicate rows after their respective operations, though some databases also support INTERSECT ALL and EXCEPT ALL to allow duplicates to be retained and returned.

INTERSECTEXCEPT 也都会去掉结果中的重复行，但有的数据库支持通过 INTERSECT ALLEXCEPT ALL 保留重复行。

THE END