2

我正在运行此代码来获取数据：

var docs = session.Query<Content>().Where(x =>  x.Attachments.Count > 0).Skip(pageNumber * Pagesize).Take(count).ToList();

它应该获取所有带有附件的文件。我添加了分页来提升它，而不是一步获取所有文档（大约有 10k 文档匹配，全部 600k）。

NHibernate 执行的查询：

exec sp_executesql N'SELECT TOP (@p0) CastleId12_, Version12_, Abstract12_, Publishe4_12_, Title12_, Body12_, Brand12_, Source12_, SourceCo9_12_, IdInSource12_, Documen11_12_, HTML12_, Subscri13_12_, FileLoc14_12_, OtherMe15_12_, Companies12_, Keywords12_, Subscri18_12_, Author12_, Documen20_12_, SourceF21_12_, SourceB22_12_, UpdateDate12_, SourceU24_12_, Content25_12_, Interna26_12_, Workben27_12_, Checksum12_, Field29_12_ FROM (select content0_.CastleId as CastleId12_, content0_.Version as Version12_, content0_.Abstract as Abstract12_, content0_.PublishedDate as Publishe4_12_, content0_.Title as Title12_, content0_.Body as Body12_, content0_.Brand as Brand12_, content0_.Source as Source12_, content0_.SourceContentId as SourceCo9_12_, content0_.IdInSource as IdInSource12_, content0_.DocumentType as Documen11_12_, content0_.HTML as HTML12_, content0_.Subscriptions as Subscri13_12_, content0_.FileLocation as FileLoc14_12_, content0_.OtherMetadata as OtherMe15_12_, content0_.Companies as Companies12_, content0_.Keywords as Keywords12_, content0_.SubscriptionUpdateDate as Subscri18_12_, content0_.Author as Author12_, content0_.DocumentStatus as Documen20_12_, content0_.SourceFileExtension as SourceF21_12_, content0_.SourceBaseName as SourceB22_12_, content0_.UpdateDate as UpdateDate12_, content0_.SourceUpdateDate as SourceU24_12_, content0_.ContentUpdateDate as Content25_12_, content0_.InternalDocumentType as Interna26_12_, content0_.WorkbenchList as Workben27_12_, content0_.Checksum as Checksum12_, content0_.Field_id as Field29_12_, ROW_NUMBER() OVER(ORDER BY CURRENT_TIMESTAMP) as __hibernate_sort_row from [Content] content0_ where (select cast(count(*) as INT) from ContentAttachments attachment1_ where content0_.CastleId=attachment1_.DocumentId)>@p1) as query WHERE query.__hibernate_sort_row > @p2 ORDER BY query.__hibernate_sort_row',N'@p0 int,@p1 int,@p2 int',@p0=50,@p1=0,@p2=0

如果我以这种方式运行它，在 SSMS 中执行需要 1.5 分钟，当然它会在我的应用程序中引发超时异常。

如果我将其更改为：

SELECT TOP (50) CastleId12_, Version12_, Abstract12_, Publishe4_12_, Title12_, Body12_, Brand12_, Source12_, SourceCo9_12_, IdInSource12_, Documen11_12_, HTML12_, Subscri13_12_, FileLoc14_12_, OtherMe15_12_, Companies12_, Keywords12_, Subscri18_12_, Author12_, Documen20_12_, SourceF21_12_, SourceB22_12_, UpdateDate12_, SourceU24_12_, Content25_12_, Interna26_12_, Workben27_12_, Checksum12_, Field29_12_ FROM (select content0_.CastleId as CastleId12_, content0_.Version as Version12_, content0_.Abstract as Abstract12_, content0_.PublishedDate as Publishe4_12_, content0_.Title as Title12_, content0_.Body as Body12_, content0_.Brand as Brand12_, content0_.Source as Source12_, content0_.SourceContentId as SourceCo9_12_, content0_.IdInSource as IdInSource12_, content0_.DocumentType as Documen11_12_, content0_.HTML as HTML12_, content0_.Subscriptions as Subscri13_12_, content0_.FileLocation as FileLoc14_12_, content0_.OtherMetadata as OtherMe15_12_, content0_.Companies as Companies12_, content0_.Keywords as Keywords12_, content0_.SubscriptionUpdateDate as Subscri18_12_, content0_.Author as Author12_, content0_.DocumentStatus as Documen20_12_, content0_.SourceFileExtension as SourceF21_12_, content0_.SourceBaseName as SourceB22_12_, content0_.UpdateDate as UpdateDate12_, content0_.SourceUpdateDate as SourceU24_12_, content0_.ContentUpdateDate as Content25_12_, content0_.InternalDocumentType as Interna26_12_, content0_.WorkbenchList as Workben27_12_, content0_.Checksum as Checksum12_, content0_.Field_id as Field29_12_, ROW_NUMBER() OVER(ORDER BY CURRENT_TIMESTAMP) as __hibernate_sort_row from [Content] content0_ where (select cast(count(*) as INT) from ContentAttachments attachment1_ where content0_.CastleId=attachment1_.DocumentId)>0) as query WHERE query.__hibernate_sort_row > 0 ORDER BY query.__hibernate_sort_row

（唯一的变化是我不在 sp_executesql 中运行它，而是内联参数）执行大约需要 2 秒。

有谁知道如何修改我的 Nhibernate 查询以使其工作得更快？我尝试更改页面大小和页面，但没有任何改变。我读过，sp_executesql但我发现的唯一一件事是关于varchar论点：https ://stackoverflow.com/a/4540108/1714342

为什么 SSMS 执行速度如此之慢？

4

2 回答 2

1

您可以尝试：

  • 在您的 NHibernate 查询中设置一个显式OrderBy（比如在主键上）（我猜默认生成ORDER BY CURRENT_TIMESTAMP的对优化器没有帮助）

  • 代替

      .Where(x =>  x.Attachments.Count > 0)

    和

      .Where(x =>  x.Attachments.Any())

    （我猜这会导致生成不同的 SQL 查询）

于 2013-08-14T13:47:09.750 回答
0

它应该获取所有带有附件的文件。

我应该建议不要加载所有文档，而只需加载文档的基本信息并将其显示给用户。当用户实际要求您显示数据内容时，加载该特定文档。

于 2013-08-14T12:02:20.187 回答