正在用SQL服务器计算满足条件的各种分类百分比

Question

我有一个 table，其中包含有关城市间航班的信息，如下所示：

    origin_city dest_city   time    
    Dothan AL   Atlanta GA    171    
    Dothan AL   Atlanta GA    171    
    Dothan AL   Elsewhere AL    2    
    Dothan AL   Elsewhere AL    2    
    Dothan AL   Elsewhere AL    2    
    Boston MA   New York NY     5    
    Boston MA   City MA         1    
    New York NY Boston MA       5    
    New York NY Boston MA       5    
    New York NY Boston MA       5    
    New York NY Poughkipsie NY  2

我想针对每个始发城市查找飞行时间少于 3 小时的航班所占的百分比。所以结果会是这样的：

    Dothan AL    60
    Boston MA    50
    New York NY  25

我认为可行的代码如下所示：

     SELECT F.origin_city as origin_city,    
       ((SELECT COUNT(*) FROM Flights as F2
       WHERE F2.actual_time < 3) / (SELECT COUNT(*) FROM Flights as  F3)) * 100
     AS percentage
     FROM Flights as F
     GROUP BY F.origin_city
     ORDER BY percentage;
     GO

当我运行它时，我得到了一个原始城市列表和一个百分比列，正如预期的那样，但百分比始终为 0。我仍然对子查询很困惑（如您所见).

Answer 1

您的百分比是在整个 table 而不是按来源城市分组。尝试这样的事情：

 SELECT F.origin_city as origin_city,    
   (SUM(CASE WHEN F.actual_time < 3 THEN 1 ELSE 0 END) / COUNT(*) )  * 100 AS percentage
 FROM Flights as F
 GROUP BY F.origin_city
 ORDER BY percentage;
 GO

FWIW 当前子查询的问题是当前行与子查询中的数据之间没有连接。您可以将其重写为：

 SELECT F.origin_city as origin_city,    
 ((SELECT COUNT(*) FROM Flights as F2
 WHERE F2.origin_city = F.origin_city and F2.actual_time < 3) / (SELECT COUNT(*) FROM Flights as  F3 where F3.origin_city = F.origin_city)) * 100
 AS percentage
 FROM Flights as F
 GROUP BY F.origin_city
 ORDER BY percentage;
 GO

但是当您已经有足够的数据来进行计算时，就没有必要为每一行重新查询 table，如上所示。

Answer 2

我会使用 AVG() 作为 window 函数：

SELECT F.origin_city as origin_city, 
       AVG( CASE WHEN F2.actual_time < 3 THEN 100.0 ELSE 0 END) as percentage
FROM Flights F
GROUP BY F.origin_city
ORDER BY percentage;

这假定时间以小时为单位。根据 Google 地图，你可以在 68 小时内从多森步行到亚特兰大，所以 171 是可疑的。

正在用SQL服务器计算满足条件的各种分类百分比

Calculating Percentage of Various Categories that Meet Conditions with SQL Server

sql

sql-server

subquery

correlated-subquery