简化 Entity Framework Core 查询

Question

我不明白如何简化这个请求

var dialogs = await dbContext.UsersDialogs
            .AsNoTracking()
            .Where(x => x.UserId == userId)
            .Select(x => new DialogModel
            {
                Id = x.DialogId,
                Login = x.Dialog.Name,
                Image = x.User.FacialImage,
                IsConfirm = x.Dialog.Messages.OrderBy(x => x.DateCreate).LastOrDefault().IsRead,
                DateTime = x.Dialog.Messages.OrderBy(x => x.DateCreate).LastOrDefault().DateCreate,
                LastMessage = x.Dialog.Messages.OrderBy(x => x.DateCreate).LastOrDefault().Content,
                LastUserId = x.Dialog.Messages.OrderBy(x => x.DateCreate).LastOrDefault().UserId
            })
            .ToListAsync();

它将它转换成这样的请求

SELECT [u].[DialogId] AS [Id], [d].[Name] AS [Login], [u0].[FacialImage] AS [Image], (
      SELECT TOP(1) [m].[IsRead]
      FROM [Messages] AS [m]
      WHERE [d].[Id] = [m].[DialogId]
      ORDER BY [m].[DateCreate] DESC) AS [IsConfirm], (
      SELECT TOP(1) [m0].[DateCreate] 
  FROM [Messages] AS [m0]
      WHERE [d].[Id] = [m0].[DialogId]
      ORDER BY [m0].[DateCreate] DESC) AS [DateTime], (
      SELECT TOP(1) [m1].[Content]
      FROM [Messages] AS [m1]
      WHERE [d].[Id] = [m1].[DialogId]
      ORDER BY [m1].[DateCreate] DESC) AS [LastMessage], (
      SELECT TOP(1) [m2].[UserId]
      FROM [Messages] AS [m2]
      WHERE [d].[Id] = [m2].[DialogId]
      ORDER BY [m2].[DateCreate] DESC) AS [LastUserId]
  FROM [UsersDialogs] AS [u]
  INNER JOIN [Dialogs] AS [d] ON [u].[DialogId] = [d].[Id]
  INNER JOIN [Users] AS [u0] ON [u].[UserId] = [u0].[Id]
  WHERE [u].[UserId] = @__userId_0

我以某种方式优化它？我不想使用 SQL 查询，因为 Linq 对我来说似乎更方便。

Answer 1

您已创建直接转换为相同 SQL 的查询。您可以通过附加Select 省略重复查询。也不需要AsNoTracking - EF Core 不添加自定义实体。

var dialogs = await dbContext.UsersDialogs
    .Where(x => x.UserId == userId)
    .Select(x => new 
    { 
        UserDialog = x, 
        LastMessage = x.Dialog.Messages.OrderByDescending(x => x.DateCreate).FirstOrDefault()
    })
    .Select(x => new DialogModel
    {
        Id = x.UserDialog.DialogId,
        Login = x.UserDialog.Dialog.Name,
        Image = x.UserDialog.User.FacialImage,
        IsConfirm = x.LastMessage.IsRead,
        DateTime = x.LastMessage.DateCreate,
        LastMessage = x.LastMessage.Content,
        LastUserId = x.LastMessage.UserId
    })
    .ToListAsync();

但上述查询的质量取决于当前 EF Core LINQ 翻译器的质量。因此，添加了另一个变体：

var query =  
    from ud in dbContext.UsersDialogs
    from lm in ud.Dialog.Messages.OrderByDescending(x => x.DateCreate)
       .Take(1).DefaultIfEmpty()
    select new DialogModel
    {
        Id = ud.DialogId,
        Login = ud.Dialog.Name,
        Image = ud.User.FacialImage,
        IsConfirm = lm.IsRead,
        DateTime = lm.DateCreate,
        LastMessage = lm.Content,
        LastUserId = lm.UserId
    };

var dialogs = await query.ToListAsync();

Answer 2

就LINQ查询而言，您可以通过使用查询语法和let来简化它：

from d in dbContext.UsersDialogs
where d.UserId == userId
let lastMessage = d.Dialog.Messages.OrderBy(d => d.DateCreate).LastOrDefault()
select new DialogModel
{
    Id = d.DialogId,
    Login = d.Dialog.Name,
    Image = d.User.FacialImage,
    IsConfirm = lastMessage.IsRead,
    DateTime = lastMessage.DateCreate,
    LastMessage = lastMessage.Content,
    LastUserId = lastMessage.UserId
}

但这并没有优化 SQL 查询。 EF 为 Messages 表中的每个字段一遍又一遍地生成相同的子查询。在 SQL Server 中，查询计划不会将这些子查询优化到一个分支中。

如果你真的想优化 SQL 查询，你必须这样做：

(
    from d in dbContext.UsersDialogs
    where d.UserId == userId
    select new
    {
        Id = d.DialogId,
        Login = d.Dialog.Name,
        Image = d.User.FacialImage,
        LastMessage = (from d.Dialog.Messages
                       orderby d.DateCreate
                       select new
                       {
                           IsConfirm = d.IsRead
                           d.DateCreate,
                           d.Content,
                           d.UserId
                       }).LastOrDefault()
    }
).AsEnumerable()
.Select(x => new DialogModel
{
    Id = x.DialogId,
    Login = x.Dialog.Name,
    Image = x.User.FacialImage,
    LastMessage.IsConfirm,
    LastMessage.DateCreate,
    LastMessage.Content,
    LastMessage.UserId
})

通过添加AsEnumerable，强制客户端评估查询的最后一部分，EF 生成一个带有一个子查询的查询，该子查询使用ROW_NUMBER() OVER 函数仅获取最后一条消息一次。但是，这样做当然很麻烦。但如果第一个查询受到相当大的性能影响，则可能需要这样做。