sql-server-2005 - 如何在sql server 2005中获得n个最高工资

Question

我在谷歌搜索中找到了以下解决方案，但我无法理解。谁能解释这个查询是如何处理的。

SELECT *
FROM   employee_test e1
WHERE  3 = (SELECT DISTINCT count(*)
            FROM   employee_test e2
            WHERE  e2.emp_sal > e1.emp_sal) 

Answer 1

It's a very inefficient correlated query to return the employee with the 4th highest salary.

SELECT *
FROM   employee_test e1
WHERE  3 = (SELECT DISTINCT count(*)
            FROM   employee_test e2
            WHERE  e2.emp_sal > e1.emp_sal);

Assuming salaries (emp_sal) are unique, the subquery counts how many employees have a higher salary than the employee in 'e1'. If the count is 3, then this is the employee with the 4th highest salary.

If there is a tie for 4th (e.g. through 6th or 10th), all the ties will be shown in the result. If there is a tie that involves the 4th and one higher slot, e.g. 2nd to 4th or 3rd and 4th, then there will be no results, since NO value of emp_sal has EXACTLY 3 records with higher emp_sal values.

The better form would be

select top(1) with ties * 
from (
    select *, rn=rank() over (order by emp_sal desc)
    from employee_test
    ) x
where rn <= 4
order by rn desc;

This shows the outright 4th highest salary (and ties if any), or if there is a tie across persons 3rd through 6th (covering spot #4), it will show 3rd through 6th.