EF Core 2.1，linq 查询没有按 sql 查询产生分组，我最后只能看到 Order By

Question

我有一个包含 Group By 子句的 linq 查询，但 Group By 没有在 sql server 上发生。 我尝试了一个简单的查询，并且 Group By 发生在 sql server 上。 请指导我为什么会有这种不同的行为？ 我希望在服务器上进行分组以提高性能。

如果我记录 sql 查询，我会在其中得到分组的简单查询：

var testt = (from doc in _patientRepository.Documents
                     group doc by doc.DocumentType into G
                     select new
                     {
                         Key = G.Key

                     }).ToList();

生成的sql：

Executed DbCommand (247ms) [Parameters=[], CommandType='Text', 
CommandTimeout='30']
SELECT [doc].[DocumentType] AS [Key]
FROM [Document] AS [doc]
GROUP BY [doc].[DocumentType]

问题查询：

var patX = (from doc in _patientRepository.Documents
                                               join pat in _patientRepository.Patients
                                               on doc.PatientId.ToString().ToLower() equals pat.PatientId.ToString().ToLower()
                                               where doc.Source.ToLower() != "testclient.server.postman" &&
                                               pat.Deleted == false && sfHCPs.Contains(pat.HcpId.ToLower())
                                               select new Document()
                                               {
                                                   DocumentId = doc.DocumentId,
                                                   CreationDateTime = doc.CreationDateTime,
                                                   DocumentType = doc.DocumentType,
                                                   PatientId = doc.PatientId,
                                                   DocumentTypeVersion = doc.DocumentTypeVersion,
                                                   Source = doc.Source,
                                                   PayloadLeft = DocumentMapper.DeserializePayload(doc.PayloadLeft),
                                                   PayloadRight = DocumentMapper.DeserializePayload(doc.PayloadRight),
                                                   PayloadBoth = DocumentMapper.DeserializePayload(doc.PayloadBoth),
                                                   IsSalesforceSynced = doc.IsSalesforceSynced,
                                                   HcpId = pat.HcpId
                                               }).GroupBy(p => new { p.PatientId, p.DocumentType })
        .Select(g => g.OrderByDescending(p => p.CreationDateTime).FirstOrDefault())
        .Where(x => x.IsSalesforceSynced == false)
        .ToList();

为什么不生成group-by sql：

Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (200ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
SELECT [doc].[DocumentId], [doc].[CreationDateTime], [doc].[DocumentType], [doc].[PatientId], [doc].[DocumentTypeVersion], [doc].[Source], [doc].[PayloadLeft], [doc].[PayloadRight], [doc].[PayloadBoth], [doc].[IsSalesforceSynced], [pat].[HcpId]
FROM [Document] AS [doc]
INNER JOIN [Patient] AS [pat] ON LOWER(CONVERT(VARCHAR(36), [doc].[PatientId])) = LOWER(CONVERT(VARCHAR(36), [pat].[PatientId]))
WHERE ((LOWER([doc].[Source]) <> N'testclient.server.postman') AND ([pat].[Deleted] = 0)) AND LOWER([pat].[HcpId]) IN (N'4e7103a9-7dff-4fa5-b540-a32a31be2997', N'abc1', N'def2', N'ghi3')
ORDER BY [doc].[PatientId], [doc].[DocumentType]

我尝试了以下方法，但生成了相同的 sql：

    var patX = ((from doc in _patientRepository.Documents
                                                       join pat in _patientRepository.Patients
                                                       on doc.PatientId.ToString().ToLower() 
                                                       equals pat.PatientId.ToString().ToLower()
                                                       where doc.Source.ToLower() != "testclient.server.postman" &&
                                                       pat.Deleted == false && sfHCPs.Contains(pat.HcpId.ToLower())
                                                       select new Document()
                                                       {
                                                           DocumentId = doc.DocumentId,
                                                           CreationDateTime = doc.CreationDateTime,
                                                           DocumentType = doc.DocumentType,
                                                           PatientId = doc.PatientId,
                                                           DocumentTypeVersion = doc.DocumentTypeVersion,
                                                           Source = doc.Source,
                                                           PayloadLeft = DocumentMapper.DeserializePayload(doc.PayloadLeft),
                                                           PayloadRight = DocumentMapper.DeserializePayload(doc.PayloadRight),
                                                           PayloadBoth = DocumentMapper.DeserializePayload(doc.PayloadBoth),
                                                           IsSalesforceSynced = doc.IsSalesforceSynced,
                                                           HcpId = pat.HcpId
                                                       }).GroupBy(p => new { p.PatientId, p.DocumentType })
                .Select(g => g.OrderByDescending(p => p.CreationDateTime).FirstOrDefault())
                .Where(x => x.IsSalesforceSynced == false))
                .ToList();

Answer 1

在 2.1 版之前，在 EF Core 中，GroupBy LINQ 运算符将始终在内存中进行计算。现在支持在大多数常见情况下将其转换为 SQL GROUP BY 子句。

更改代码：尝试将 .GroupBy 放在 .select 方法之前（第一次选择）

Answer 2

考虑重新排序查询，以便新类的 select 位于最后：

var p1 = from doc in _patientRepository.Documents
         join pat in _patientRepository.Patients on doc.PatientId.ToString().ToLower() equals pat.PatientId.ToString().ToLower()
         where doc.Source.ToLower() != "testclient.server.postman" && pat.Deleted == false && sfHCPs.Contains(pat.HcpId.ToLower())
         group new { doc, pat.HcpId } by new { doc.PatientId, doc.DocumentType } into dpg
         select dpg.OrderByDescending(dp => dp.doc.CreationDateTime).FirstOrDefault();

var patX = (from dp in p1
            where !dp.doc.IsSalesforceSynced
            select new Document() {
                DocumentId = dp.doc.DocumentId,
                CreationDateTime = dp.doc.CreationDateTime,
                DocumentType = dp.doc.DocumentType,
                PatientId = dp.doc.PatientId,
                DocumentTypeVersion = dp.doc.DocumentTypeVersion,
                Source = dp.doc.Source,
                PayloadLeft = DocumentMapper.DeserializePayload(dp.doc.PayloadLeft),
                PayloadRight = DocumentMapper.DeserializePayload(dp.doc.PayloadRight),
                PayloadBoth = DocumentMapper.DeserializePayload(dp.doc.PayloadBoth),
                IsSalesforceSynced = dp.doc.IsSalesforceSynced,
                HcpId = dp.HcpId
            })
            .ToList();