实体框架，高效的NavigationProperty.OfType查询
本文关键字：NavigationProperty OfType 查询 高效 框架 实体 | 更新日期: 2023-09-27 17:58:02
我在EF4中使用每表类型（TPT）继承构建高效查询时遇到问题。
我有一个名为集的实体，每个集可以有多个事件。有几种不同类型的事件，它们都源自一个名为Event的基本实体。我想过滤所有不包含特定类型事件的剧集事件有一个导航属性，它是所有事件的集合（即基本事件类型的集合）
我试过：
from episode in context.EpisodeSet
where episode.Events.OfType<DerivedEvent>().Count() == 0
select episode
和
from episode in context.EpisodeSet
where episode.Events.Where(p => p is DerivedEvent).Count() == 0
select episode
这两者都会生成一个典型的长SQL扩展，用于查询每个Event类型的表。
难道不应该有一种方法在LINQ中表达这个查询，只涉及结果SQL中Episode和DerivedEvent表之间的联接吗？
编辑：作为对ProfessorX的响应，这里是生成的SQL（基本上只是所有事件表的典型的大规模联合）
SELECT
[Extent1].[Id] AS [Id],
[Extent1].[TypeId] AS [TypeId],
[Extent1].[PatientId] AS [PatientId],
[Extent1].[CentreId] AS [CentreId],
[Extent1].[CreatedOn] AS [CreatedOn],
[Extent1].[UpdatedOn] AS [UpdatedOn],
[Extent1].[CreatedBy] AS [CreatedBy],
[Extent1].[UpdatedBy] AS [UpdatedBy]
FROM [dbo].[Episode] AS [Extent1]
WHERE EXISTS (SELECT
1 AS [C1]
FROM [dbo].[Event] AS [Extent2]
LEFT OUTER JOIN (SELECT
[Extent3].[Id] AS [Id],
cast(1 as bit) AS [C1]
FROM [dbo].[InvasiveDischargableEvent] AS [Extent3] ) AS [Project1] ON [Extent2].[Id] = [Project1].[Id]
LEFT OUTER JOIN (SELECT
[UnionAll4].[C1] AS [C1],
[UnionAll4].[C2] AS [C2],
[UnionAll4].[C3] AS [C3],
[UnionAll4].[C4] AS [C4],
[UnionAll4].[C5] AS [C5],
[UnionAll4].[C6] AS [C6],
[UnionAll4].[C7] AS [C7]
FROM (SELECT
[UnionAll3].[C1] AS [C1],
[UnionAll3].[C2] AS [C2],
[UnionAll3].[C3] AS [C3],
[UnionAll3].[C4] AS [C4],
[UnionAll3].[C5] AS [C5],
[UnionAll3].[C6] AS [C6],
[UnionAll3].[C7] AS [C7]
FROM (SELECT
[UnionAll2].[C1] AS [C1],
[UnionAll2].[C2] AS [C2],
[UnionAll2].[C3] AS [C3],
[UnionAll2].[C4] AS [C4],
[UnionAll2].[C5] AS [C5],
[UnionAll2].[C6] AS [C6],
[UnionAll2].[C7] AS [C7]
FROM (SELECT
[UnionAll1].[Id] AS [C1],
[UnionAll1].[C1] AS [C2],
[UnionAll1].[C2] AS [C3],
[UnionAll1].[C3] AS [C4],
[UnionAll1].[C4] AS [C5],
[UnionAll1].[C5] AS [C6],
[UnionAll1].[C6] AS [C7]
FROM (SELECT
[Extent4].[Id] AS [Id],
cast(0 as bit) AS [C1],
cast(1 as bit) AS [C2],
cast(0 as bit) AS [C3],
cast(0 as bit) AS [C4],
cast(0 as bit) AS [C5],
cast(0 as bit) AS [C6]
FROM [dbo].[InvasivePSQ10Event] AS [Extent4]
UNION ALL
SELECT
[Extent5].[Id] AS [Id],
cast(0 as bit) AS [C1],
cast(0 as bit) AS [C2],
cast(0 as bit) AS [C3],
cast(0 as bit) AS [C4],
cast(0 as bit) AS [C5],
cast(1 as bit) AS [C6]
FROM [dbo].[InvasivePostTreatmentEvent] AS [Extent5]) AS [UnionAll1]
UNION ALL
SELECT
[Extent6].[Id] AS [Id],
cast(0 as bit) AS [C1],
cast(0 as bit) AS [C2],
cast(1 as bit) AS [C3],
cast(0 as bit) AS [C4],
cast(0 as bit) AS [C5],
cast(0 as bit) AS [C6]
FROM [dbo].[InvasiveTreatmentEvent] AS [Extent6]) AS [UnionAll2]
UNION ALL
SELECT
[Extent7].[Id] AS [Id],
cast(0 as bit) AS [C1],
cast(0 as bit) AS [C2],
cast(0 as bit) AS [C3],
cast(0 as bit) AS [C4],
cast(1 as bit) AS [C5],
cast(0 as bit) AS [C6]
FROM [dbo].[InvasiveConsultationEvent] AS [Extent7]) AS [UnionAll3]
UNION ALL
SELECT
[Extent8].[Id] AS [Id],
cast(1 as bit) AS [C1],
cast(0 as bit) AS [C2],
cast(0 as bit) AS [C3],
cast(0 as bit) AS [C4],
cast(0 as bit) AS [C5],
cast(0 as bit) AS [C6]
FROM [dbo].[InvasiveMOXFQEvent] AS [Extent8]) AS [UnionAll4]
UNION ALL
SELECT
[Extent9].[Id] AS [Id],
cast(0 as bit) AS [C1],
cast(0 as bit) AS [C2],
cast(0 as bit) AS [C3],
cast(1 as bit) AS [C4],
cast(0 as bit) AS [C5],
cast(0 as bit) AS [C6]
FROM [dbo].[InvasiveReferralEvent] AS [Extent9]) AS [UnionAll5] ON [Extent2].[Id] = [UnionAll5].[C1]
WHERE ([Extent1].[Id] = [Extent2].[EpisodeId]) AND (CASE WHEN (( NOT (([UnionAll5].[C2] = 1) AND ([UnionAll5].[C2] IS NOT NULL))) AND ( NOT (([UnionAll5].[C3] = 1) AND ([UnionAll5].[C3] IS NOT NULL))) AND ( NOT (([UnionAll5].[C4] = 1) AND ([UnionAll5].[C4] IS NOT NULL))) AND ( NOT (([Project1].[C1] = 1) AND ([Project1].[C1] IS NOT NULL))) AND ( NOT (([UnionAll5].[C5] = 1) AND ([UnionAll5].[C5] IS NOT NULL)))) THEN '2X' WHEN (([UnionAll5].[C5] = 1) AND ([UnionAll5].[C5] IS NOT NULL)) THEN '2X0X' WHEN (([Project1].[C1] = 1) AND ([Project1].[C1] IS NOT NULL) AND ( NOT (([UnionAll5].[C6] = 1) AND ([UnionAll5].[C6] IS NOT NULL))) AND ( NOT (([UnionAll5].[C7] = 1) AND ([UnionAll5].[C7] IS NOT NULL)))) THEN '2X1X' WHEN (([UnionAll5].[C4] = 1) AND ([UnionAll5].[C4] IS NOT NULL)) THEN '2X2X' WHEN (([UnionAll5].[C2] = 1) AND ([UnionAll5].[C2] IS NOT NULL)) THEN '2X3X' WHEN (([UnionAll5].[C6] = 1) AND ([UnionAll5].[C6] IS NOT NULL)) THEN '2X1X0X' WHEN (([UnionAll5].[C7] = 1) AND ([UnionAll5].[C7] IS NOT NULL)) THEN '2X1X1X' ELSE '2X4X' END LIKE '2X4X%')
)
经过多次挠头之后，我终于实现了这一点：
var episodes = (from episode in context.EpisodeSet
join e in context.EventSet.OfType<DerivedEvent>() on episode.Id equals e.EpisodeId into outer
from o in outer.DefaultIfEmpty()
where o == null
select episode)
因此，我没有尝试将OfType过滤器应用于导航属性，而是将其应用于ObjectSet并进行外部联接。OfType和"as"类型筛选似乎不适用于导航属性。
这会产生在DerivedEvent表中没有相应事件的事件，并且使用您手动编写的SQL。
LINQ遵循您自然使用SQL编写查询的方式。很容易被所有这些导航属性所吸引，这些导航属性导致了好看的LINQ但糟糕的SQL。
。Any（）比。计数（）根据我的预期，您的查询可以优化为：
context.EpisodeSet
.Where(e => e.Events.Any(p => p is DerivedEvent))
.Select(e => e);