ef-core 嵌套表聚合函数给出“NavigationExpandingExpressionVisitor failed”

Question

我有一个嵌套数据模型，我想从中获取按顶级属性分组的聚合数据。我的模型例如：

public class Scan {
    public long Id {get; set;}
    public int ProjectId { get; set; }
    public int ScanUserId { get; set; }
    public ICollection<ScanItem>? Items { get; set; }
}

public class ScanItem
{
    public long Id { get; set; }
    public long InventoryScanId { get; set; }
    public double Qty { get; set; }
}

我想按 Scan.ScanUserId 分组所有扫描，然后获取 ScanItems.Qty 的总和，例如每个用户。我的查询看起来像这样，EF 给出了以下错误：

'NavigationExpandingExpressionVisitor' 处理 LINQ 表达式 'AsQueryable((Unhandled parameter: x).Items)' 失败

from scan in Scans
        .Include(x=>x.ScanUser)
        .Include(x=>x.Items)
    group scan by new { scan.ScanUser.Name, scan.ScanUser.Id } into g
    select new
    {
        UserId = g.Key.Id,
        Name = g.Key.Name,
        LastSyncTime = g.Max(x => x.ScanDate),
        ScanItems = g.Sum(x=>x.Items.Sum(i=>i.Qty))
    }

如何在嵌套表的属性上运行聚合函数而不在客户端对其进行评估？

Answer 1

EF Core 仍然无法转换GroupBy结果（分组）上的嵌套聚合。

您必须通过使用GroupBy（或element查询语法group element by key）的元素选择器预先计算嵌套聚合：

from scan in Scans
group new { scan.ScanDate, Qty = scan.Items.Sum(i => i.Qty) } // <--
by new { scan.ScanUser.Name, scan.ScanUser.Id } into g
select new
{
    UserId = g.Key.Id,
    Name = g.Key.Name,
    LastSyncTime = g.Max(x => x.ScanDate),
    ScanItems = g.Sum(x => x.Qty) // <--
}

**更新：** 对于 SqlServer，上面的 LINQ 查询转换为这样的 SQL 查询：

from scan in Scans
group new { scan.ScanDate, Qty = scan.Items.Sum(i => i.Qty) } // <--
by new { scan.ScanUser.Name, scan.ScanUser.Id } into g
select new
{
    UserId = g.Key.Id,
    Name = g.Key.Name,
    LastSyncTime = g.Max(x => x.ScanDate),
    ScanItems = g.Sum(x => x.Qty) // <--
}

如注释中所述，它会生成 SQL 执行异常“无法对包含聚合或子查询的表达式执行聚合函数”。.

所以你真的需要另一种方法 - 用于left join在分组之前展平结果集，然后在该集合上执行分组/聚合：

from scan in Scans
from item in scan.Items.DefaultIfEmpty() // <-- left outer join
group new { scan, item } by new { scan.ScanUser.Name, scan.ScanUser.Id } into g
select new
{
    UserId = g.Key.Id,
    Name = g.Key.Name,
    LastSyncTime = g.Max(x => x.scan.ScanDate),
    ScanItems = g.Sum(x => (double?)x.item.Qty) ?? 0
};

现在转换为希望有效的 SqlServer SQL 查询：

 SELECT [s1].[Id] AS [UserId], [s1].[Name], MAX([s0].[ScanDate]) AS [LastSyncTime], SUM((
      SELECT SUM([s].[Qty])
      FROM [ScanItem] AS [s]
      WHERE [s0].[Id] = [s].[InventoryScanId])) AS [ScanItems]
  FROM [Scan] AS [s0]
  INNER JOIN [ScanUser] AS [s1] ON [s0].[ScanUserId] = [s1].[Id]
  GROUP BY [s1].[Name], [s1].[Id]