来自 LINQ 语句的高效 SQL 与连接和聚合？

Question

我正在尝试做一个相对基本的 SQL 语句来连接和分组一些表并使用聚合函数进行汇总。我会用 SQL 写成这样：

select
  p.LocationID,
  NumReadings = count(*),
  MinDate = min(t.[DateTime]),
  MaxDate = max(t.[DateTime])
from Station p inner join Data pd on p.LocationID = pd.ReadingLocationID
    inner join ApplicationDateTime t on t.ApplicationDateTimeID = pd.DateTimeID
group by p.LocationID

当我在 EF4 中使用下面的 Linq 语句时，它会创建一些可怕的 SQL（见最底部）。有没有更好的方法来做到这一点？显式地执行连接而不是使用 EF 导航属性会使情况变得更糟。

我不关心美学，但看看查询执行，执行格式不佳的 SQL 需要 3-4 倍的时间。

from s in Station
select new DataSummary
{
   ReadingLocationID = s.ReadingLocationID,
   StationIdentifier = s.StationIdentifier,
   NumReadings = s.Data.Count(),
   MinDateLoaded = s.Data.Min(d => d.ApplicationDateTime.DateTime),
   MaxDateLoaded = s.Data.Max(d => d.ApplicationDateTime.DateTime)
};

这里是 SQL（注意：这里还有一些额外的复杂性，例如表示为另一个连接的继承关系，但这只会导致另一个嵌套级别）。

SELECT 
  [Project3].[LocationTypeID] AS [LocationTypeID], 
  [Project3].[ReadingLocationID] AS [ReadingLocationID], 
  [Project3].[LocationIdentifier] AS [LocationIdentifier], 
  [Project3].[C1] AS [C1], 
  CAST( [Project3].[C2] AS datetime2) AS [C2], 
  CAST( [Project3].[C3] AS datetime2) AS [C3]
FROM ( SELECT 
        [Project2].[ReadingLocationID] AS [ReadingLocationID], 
        [Project2].[LocationTypeID] AS [LocationTypeID], 
        [Project2].[LocationIdentifier] AS [LocationIdentifier], 
        [Project2].[C1] AS [C1], 
        [Project2].[C2] AS [C2], 
        (SELECT 
           MAX([Extent7].[DateTime]) AS [A1]
           FROM  [dbo].[Data] AS [Extent6]
           INNER JOIN [dbo].[ApplicationDateTime] AS [Extent7] ON [Extent6].[DateTimeID] = [Extent7].[ApplicationDateTimeID]
           WHERE [Project2].[ReadingLocationID] = [Extent6].[ReadingLocationID]) AS [C3]
            FROM ( SELECT 
            [Project1].[ReadingLocationID] AS [ReadingLocationID], 
            [Project1].[LocationTypeID] AS [LocationTypeID], 
            [Project1].[LocationIdentifier] AS [LocationIdentifier], 
            [Project1].[C1] AS [C1], 
            (SELECT 
               MIN([Extent5].[DateTime]) AS [A1]
               FROM  [dbo].[Data] AS [Extent4]
               INNER JOIN [dbo].[ApplicationDateTime] AS [Extent5] ON [Extent4].[DateTimeID] = [Extent5].[ApplicationDateTimeID]
               WHERE [Project1].[ReadingLocationID] = [Extent4].[ReadingLocationID]) AS [C2]
               FROM ( SELECT 
                  [Extent1].[ReadingLocationID] AS [ReadingLocationID], 
                  [Extent1].[LocationTypeID] AS [LocationTypeID], 
                  [Extent1].[LocationIdentifier] AS [LocationIdentifier], 
                  (SELECT 
                        COUNT(1) AS [A1]
                        FROM [dbo].[Data] AS [Extent3]
                        WHERE [Extent1].[ReadingLocationID] = [Extent3].[ReadingLocationID]) AS [C1]
                   FROM  [dbo].[ReadingLocation] AS [Extent1]
                   INNER JOIN [dbo].[Station] AS [Extent2] ON [Extent1].[ReadingLocationID] = [Extent2].[LocationID]
                   WHERE ([Extent1].[LocationTypeID] =  CAST( '1' AS int)) AND ([Extent2].[LineID] = 'ACBB3FDF-116C-4E8E-AA80-B925E4922AC8')
                   )  AS [Project1]
               )  AS [Project2]
)

帮助！谢谢。

Answer 1

使用不同的 ORM 怎么样？特别是像 PetaPoco 或 Massive 这样的 MicroOrm 可以让您在 SQL 中编写查询并返回 .NET 对象。

两者都是 Nuget 包：PetaPoco、Massive，因此您可以轻松安装它们。

如果您习惯编写 SQL 并且更愿意控制您的查询，那么它们可能是可行的候选者。

Answer 2

LINQ 可以做很多事情，但它绝对不是魔法。 尝试像编写 SQL 一样编写 LINQ。定义您的联接。

这就是你的 LINQ 语句的样子：

Station
 .Join(Data, s => new { s.LocationID }, d => new { LocationID = d.ReadingLocationID }, (s,d) => new { s.ID, s.LocationID, d.DateTimeID })
 .Join(ApplicationDateTime, j1 => new { j1.DateTimeID }, t => new { DateTimeID = t.ApplicationDateTimeID }, (j1,t) => { j.ID, j.LocationID, t.DateTime })
 .Group(j2 => new { j2.ID, j2.LocationID })
 .Select(g => new DataSummary
   {
      StationIdentifier = g.Key.ID,
      ReadingLocationID = g.Key.LocationID,
      NumReadings = g.Count(),
      MinDateLoaded = g.Min(j2 => j2.DateTime),
      MaxDateLoaded = g.Max(j2 => j2.DateTime)
   });

注意：上面提供的代码未经测试！

请记住，在 LINQ 中创建连接时，您在连接中使用的匿名对象的属性应该具有相同的名称和类型。这花了我一些时间让我的第一次加入工作。 示例：假设 Data 表上的 DateTimeID 列可以为 NULL，而 ApplicationDateTime 上的 ApplicationDateTimeID 列不能为 NULL。然后，您应该将该连接更改为：

.Join(ApplicationDateTime, j1 => new { j1.DateTimeID }, t => new { DateTimeID = (int?)t.ApplicationDateTimeID }, (j1,t) => { j.ID, j.LocationID, t.DateTime })