LEFT JOIN 不会返回所有值答案

【问题标题】：LEFT JOIN doesn't return all valuesLEFT JOIN 不会返回所有值
【发布时间】：2020-12-17 10:44:25
【问题描述】：

我有以下表格：

create table Cars
(
  CarID int,
  CarType varchar(50)
);


create table Maintenances
(
  CarID int,
  MaintenanceDate date,
  MaintenanceCost money
);

create table Repairs
(
  CarID int,
  RepairDate date,
  RepairCost money
);

我试过这个：

SELECT C.CarType,
SUM(MaintenanceCost),
SUM(RepairCost)
FROM Cars AS C 
LEFT JOIN Maintenances AS M ON M.CarID=C.CarID
LEFT JOIN Repairs AS R ON R.CarID=C.CarID
WHERE M.MaintenanceDate BETWEEN '2020-12-01' AND '2020-12-31' AND
R.RepairDate BETWEEN '2020-12-01' AND '2020-12-31'
GROUP BY C.CarType

但它不会返回所有 CarType 的值。看到这个小提琴：Fiddle

【问题讨论】：

您的WHERE 正在将您的外连接变成内连接

标签： sql sql-server datetime left-join where-clause

【解决方案1】：

将LEFT JOINed 表上的条件从WHERE 子句移动到连接的ON 子句 - 否则，它们将成为强制性的，并且任一表中的不匹配行都将被过滤掉。

所以：

SELECT C.CarType, SUM(M.MaintenanceCost), SUM(R.RepairCost)
FROM Cars AS C 
LEFT JOIN Maintenances M 
    ON  M.CarID = C.CarID 
    AND M.MaintenanceDate BETWEEN '20201201' AND '20201231'
LEFT JOIN Repairs R 
    ON  R.CarID = C.CarID 
    AND R.RepairDate BETWEEN '20201201' AND '20201231'
GROUP BY C.CarType

旁注：

我在 SUM() 中的列前面加上它们所属的表（我不得不猜测）：这是一种很好的做法，并且使查询不言自明，明确
所有版本的 SQL Server 都支持YYYYMMDD 作为日期文字，无论区域设置如何，因此它可能是比YYYY-MM-DD 更好的选择

【讨论】：

【解决方案2】：

这是因为您按维护和维修表的日期进行过滤。只有 CarID 3 在两个表格中都有您给定日期范围的条目。

如果您希望结果包含您的日期范围内的所有汽车类型，您可以尝试此查询。

    SELECT C.CarType,
    ISNULL(SUM(MaintenanceCost),0) AS MaintenanceCost,
    ISNULL(SUM(RepairCost),0) AS RepairCost
     Cars AS C 
     JOIN Maintenances AS M 
       M.CarID = C.CarID
       M.MaintenanceDate BETWEEN '2020-12-01' AND '2020-12-31'
     JOIN Repairs AS R 
       R.CarID = C.CarID
       R.RepairDate BETWEEN '2020-12-01' AND '2020-12-31'
     BY C.CarType

【讨论】：

【解决方案3】：

如果您希望SUM()s 正确，我建议使用相关子查询编写此逻辑。问题是每个Carid 都会得到一个笛卡尔积——笛卡尔积意味着总和计算正确：

如果CarType 在Cars 中是唯一的，您可以使用：

SELECT C.CarType,
       (SELECT SUM(M.MaintenanceCost)
        FROM Maintenances M 
        WHERE M.CarID = C.CarID AND
              M.MaintenanceDate BETWEEN '20201201' AND '20201231' 
       ),
       (SELECT SUM(R.RepairCost)
        FROM Repairs R
        WHERE R.CarID = R.CarID AND
              R.RepairDate BETWEEN '20201201' AND '20201231' 
       )
FROM Cars C ;

否则，这看起来有点复杂，因为您在每个子查询中都需要CarType：

SELECT C.CarType,
       (SELECT SUM(M.MaintenanceCost)
        FROM Maintenances M JOIN
             Cars C
             ON M.CarID = C.CarID
        WHERE C.CarType = CT.CarType AND
              M.MaintenanceDate BETWEEN '20201201' AND '20201231' 
       ),
       (SELECT SUM(R.RepairCost)
        FROM Repairs R JOIN
             Cars C
             ON R.CarID = C.CarID
        WHERE C.CarType = CT.CarType AND
              R.RepairDate BETWEEN '20201201' AND '20201231' 
       )
FROM (SELECT DISTINCT CarType FROM Cars C) CT

【讨论】：