Question

我正在使用LINQPad来学习LINQ，而且我遇到了绊脚石。

目标是获取网络ID，网络名称和每个站点的列表。

这是我原来的SQL：

SELECT n.iStationId AS NetworkID, n.sPrettyName AS NetworkName, COUNT(s.iStationID) AS StationCount
FROM T_StationInfo AS s, T_StationInfo as n
WHERE s.iNetworkId = n.iStationId
GROUP BY n.sPrettyName, n.iStationId
ORDER BY COUNT(s.iStationID) DESC

这是我的LINQ：

from s in T_stationInfo
from n in T_stationInfo
where s.INetworkID == n.IStationID
group s by s.INetworkID into stations
orderby stations.Count(x => x.INetworkID == stations.Key) descending
select new {
    NetworkId = stations.Key,
    NetworkName = T_stationInfo.Single(x => x.IStationID == stations.Key).SPrettyName,
    StationCount = stations.Count(x => x.INetworkID == stations.Key)
};

LINQ的执行时间要长5倍。我正在查看linq语句生成的SQL，它会将t_stationInfo表拉入7次。

我相信这是因为我滥用LINQ但我看不到在哪里或如何。

什么LINQ语句会创建等效的SQL，或者至少是SQL表现不佳的SQL？

一对夫妇注意到：

无法更改表/数据库的结构。
这个问题更多的是学习使用LINQ而不是获取ID，名称和计数列表。
我很感激！：）

- 编辑 -

只是为了澄清结构：表中的每一行都是具有各种信息（名称，联系人等）并且可以拥有父级的实体。那些父母也在桌旁。在这种情况下，父母不能有父母。它们的父字段为NULL或0。

因此，要获取工作站的父级名称（在表中称为网络），我将工作站信息表拉两次，并将父ID（网络ID）加入实体ID（工作站ID），以便单行我有电台的信息和父母的信息。因此，同桌的两个起源。

这有意义吗？

- EDIT2 -

这是原始LINQ查询生成的sql：

SELECT [t2].[iNetworkID] AS [NetworkId], (
    SELECT [t5].[sPrettyName]
    FROM [t_stationInfo] AS [t5]
    WHERE (CONVERT(Decimal(29,4),[t5].[iStationID])) = [t2].[iNetworkID]
    ) AS [NetworkName], (
    SELECT COUNT(*)
    FROM [t_stationInfo] AS [t6], [t_stationInfo] AS [t7]
    WHERE ([t6].[iNetworkID] = [t2].[iNetworkID]) AND ([t2].[iNetworkID] = [t6].[iNetworkID]) AND ([t6].[iNetworkID] = (CONVERT(Decimal(29,4),[t7].[iStationID])))
    ) AS [StationCount]
FROM (
    SELECT [t0].[iNetworkID]
    FROM [t_stationInfo] AS [t0], [t_stationInfo] AS [t1]
    WHERE [t0].[iNetworkID] = (CONVERT(Decimal(29,4),[t1].[iStationID]))
    GROUP BY [t0].[iNetworkID]
    ) AS [t2]
ORDER BY (
    SELECT COUNT(*)
    FROM [t_stationInfo] AS [t3], [t_stationInfo] AS [t4]
    WHERE ([t3].[iNetworkID] = [t2].[iNetworkID]) AND ([t2].[iNetworkID] = [t3].[iNetworkID]) AND ([t3].[iNetworkID] = (CONVERT(Decimal(29,4),[t4].[iStationID])))
    ) DESC

Answer 1

如果有的话，我不会对你的表现产生多大的影响。但是当我查看你的查询时，我看到一个函数被声明了两次：

stations.Count(s => s.INetworkID == stations.Key)

使用let子句是否可以提高性能？

from station in T_stationInfo
from network in T_stationInfo
where station.INetworkID == network.IStationID
group station by station.INetworkID into stations
let stationCount = stations.Count(x => x.INetworkID == stations.Key)
orderby stationCount descending
select new
{
    NetworkId = stations.Key,
    NetworkName = T_stationInfo.First(x => x.IStationID == stations.Key).sPrettyName,
    StationCount = stationCount
};

我觉得应该有更好的方法来分配NetworkName属性，但我不确定。

哦，抱歉重命名变量。我将s更改为station，将n更改为network，以帮助我更好地遵循它。

Answer 2

感谢Alexander Taran的评论，我重新调查了组语法，最后了解了如何按多个字段进行分组。这让我想到了这个：

from s in T_stationInfo
from n in T_stationInfo
where s.INetworkID == n.IStationID
group s by new { s.INetworkID, n.SPrettyName } into stations
orderby stations.Count() descending
select new {
    NetworkId = stations.Key.INetworkID,
    NetworkName = stations.Key.SPrettyName,
    StationCount = stations.Count()
};

生成以下SQL：

SELECT [t2].[iNetworkID] AS [NetworkId], [t2].[sPrettyName] AS [NetworkName], [t2].[value2] AS [StationCount]
FROM (
    SELECT COUNT(*) AS [value], COUNT(*) AS [value2], [t0].[iNetworkID], [t1].[sPrettyName]
    FROM [t_stationInfo] AS [t0], [t_stationInfo] AS [t1]
    WHERE [t0].[iNetworkID] = (CONVERT(Decimal(29,4),[t1].[iStationID]))
    GROUP BY [t0].[iNetworkID], [t1].[sPrettyName]
    ) AS [t2]
ORDER BY [t2].[value] DESC

外部选择看起来像简单的重命名和重新排序。如果我将ORDER BY移动到内部选择并剥离外部选择，它将以与我手工制作的SQL相同的速度运行。

作为奖励，新查询似乎更容易理解。

- 编辑 -

根据Justin Rusbatch的回答，我将Stations.Count（）拉出一个let，它确实有一个小的性能增加（大约10％）。它还在某种程度上清理了渲染的sql。

截至目前，这是我所取得的最好成绩：

from station in T_stationInfo 
from network in T_stationInfo 
where station.INetworkID == network.IStationID 
group station by new { station.INetworkID, network.SPrettyName } into stations 
let stationCount = stations.Count()
orderby stationCount descending 
select new 
{ 
    NetworkId = stations.Key.INetworkID, 
    NetworkName = stations.Key.SPrettyName, 
    StationCount = stationCount
};

这会创建以下SQL：

SELECT [t2].[iNetworkID] AS [NetworkId], [t2].[sPrettyName] AS [NetworkName], [t2].[value] AS [StationCount]
FROM (
    SELECT COUNT(*) AS [value], [t0].[iNetworkID], [t1].[sPrettyName]
    FROM [t_stationInfo] AS [t0], [t_stationInfo] AS [t1]
    WHERE [t0].[iNetworkID] = (CONVERT(Decimal(29,4),[t1].[iStationID]))
    GROUP BY [t0].[iNetworkID], [t1].[sPrettyName]
    ) AS [t2]
ORDER BY [t2].[value] DESC

与我开始的地方相比，这是一个巨大的改进。

我的LINQ效率很低

2 个答案: