根据列值将SQL select数据集多行合并为一行答案

【问题标题】：Combine SQL select data-set multiple rows into a single row based on column values根据列值将SQL select数据集多行合并为一行
【发布时间】：2018-06-13 16:26:35
【问题描述】：

我有一个 SQL 查询，而不是将多个数据库表中的数据组合成一个数据集，例如：

SELECT 
    CS.Id As ID,
    CP.Country As Country,
    S.Title As Site,
    CONCAT(V.Title,' ',M.Model,' ',M.Version) As Machine,
    CP.Title As Protocol,
    CS.Title As Sequence,
    IQS.SequenceType AS Type,
    IQS.ImageQuality AS IQ,
    _IQBeforeOpt =
        CASE IQS.SequenceType
            WHEN 1 THEN IQS.ImageQuality
        END,
    _IQDuringOpt = 
        CASE IQS.SequenceType
            WHEN 2 THEN IQS.ImageQuality
        END,
    _IQAfterOpt = 
        CASE IQS.SequenceType
            WHEN 3 THEN IQS.ImageQuality
        END,
   ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY))) AS SeqPI,
   _SeqPIBeforeOpt = 
        CASE IQS.SequenceType
            WHEN 1 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
       END,
  _SeqPIDuringOpt =
      CASE IQS.SequenceType
          WHEN 2 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END,
  _SeqPIAfterOpt =
      CASE IQS.SequenceType
          WHEN 3 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END       
FROM ClientSequence CS
JOIN ImageQualitySequencePool IQS ON CS.Id = IQS.SequenceId
JOIN ClientProtocol CP ON CS.ClientProtocolId = CP.Id
JOIN Team S ON CP.TeamID = S.Id
JOIN Machine M On CP.MachineId = M.Id
JOIN Vendor V ON M.VendorId = V.Id
ORDER BY CP.Country,S.Title,CP.MachineId,CP.Title,CS.Title,IQS.SequenceType

上面的查询用于组合来自多个表（主表是 ClientSequence 和 ImageQualitySequencePool）的数据，以构建我想要用于报告的数据集。

以下是一些指示性值：

ID  Country    Site    Machine   Protocol   Sequence          Type  IQ   _IQBeforeOpt   _IQDuringOpt   _IQAfterOpt  SeqPI   _SeqPIBeforeOpt   _SeqPIDuringOpt    _SeqPIAfterOpt
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------
41  Country_1  Site_1  Mach_1    ProtocolA  Ax_PD_FSE_FS      1     2    2              NULL            NULL         0.805   0.805             NULL               NULL
86  Country_1  Site_1  Mach_1    ProtocolA  Ax_PD_FSE_FS      2     4    NULL           4               NULL         1.38    NULL              1.38               NULL
91  Country_1  Site_1  Mach_1    ProtocolA  Ax_PD_FSE_FS      3     3    NULL           NULL            3            1.03    NULL              NULL               1.03

76  Country_1  Site_1  Mach_1    ProtocolB  Cor_PD_FSE_FS     1     3    3              NULL            NULL         0.83    0.83              NULL               NULL
88  Country_1  Site_1  Mach_1    ProtocolB  Cor_PD_FSE_FS     2     4    NULL           4               NULL         1.62    NULL              1.62               NULL
92  Country_1  Site_1  Mach_1    ProtocolB  Cor_PD_FSE_FS     3     3    NULL           NULL            3            1.21    NULL              NULL               1.21

91  Country_2  Site_3  Mach_3    ProtocolA  Seg_SF_FSE_FS     1     3    1              NULL            NULL         1.41    1.41              NULL               NULL
94  Country_2  Site_3  Mach_3    ProtocolA  Seg_SF_FSE_FS     2     2    NULL           2               NULL         1.28    NULL              1.28               NULL

72  Country_2  Site_3  Mach_3    ProtocolC  Ger_SE_FSE_FS     1     3    3              NULL            NULL         0.83    0.83              NULL               NULL

因此“国家、站点、机器、协议、序列”的每个组合都可以是类型：1,2 或 3，含义：

类型 1：优化前的序列
类型 2：序列期间优化
类型 3：优化后的序列

正如您在查询中看到的，最初我在数据库中存储了一些 IQ 和 SeqPI 值，并且我根据序列。

一些“序列”具有对应于所有 3 种类型的记录，而另一些则没有。

我想要的是将“国家、站点、机器、协议、序列”的特定组合的所有记录合并到一行中。所以最后我想有一个查询，结果是这样的数据集：

Country    Site    Machine   Protocol    Sequence       Type     _IQBeforeOpt   _IQDuringOpt   _IQAfterOpt  _SeqPIBeforeOpt   _SeqPIDuringOpt    _SeqPIAfterOpt
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Country_1  Site_1  Mach_1    ProtocolA   Ax_PD_FSE_FS   1,2,3    2              4             3            0.805              1.38               1.03
Country_1  Site_1  Mach_1    ProtocolB   Cor_PD_FSE_FS  1,2,3    3              4             3            0.83               1.62               1.21
Country_2  Site_3  Mach_3    ProtocolA   Seg_SF_FSE_FS  1,2      3              2             NULL         1.41               1.28               NULL
Country_2  Site_3  Mach_3    ProtocolC   Ger_SE_FSE_FS  1        3              NULL          NULL         0.83               NULL               NULL

【问题讨论】：

您的报告工具可能已经具有用于交叉表或数据透视报告的内置功能。您可以直接使用它，而无需对查询进行任何修改。试试看是否有效

标签： sql sql-server database relational-database

【解决方案1】：

窗口函数是一种有效的聚合方式： Over Clause

使用 OVER 子句，您可以定义自己的聚合函数。

如果失败，您可以使用 CTE 破解它 Concatenation with CTEs

【讨论】：

【解决方案2】：

你可以试试这个吗？

    SELECT 
    CS.Id As ID,
    CP.Country As Country,
    S.Title As Site,
    CONCAT(V.Title,' ',M.Model,' ',M.Version) As Machine,
    CP.Title As Protocol,
    CS.Title As Sequence,
 --   IQS.SequenceType AS Type,
    STUFF(OQS.types,1,1,'') AS [type],
  --  IQS.ImageQuality AS IQ,
    _IQBeforeOpt =MAX(
        CASE IQS.SequenceType
            WHEN 1 THEN IQS.ImageQuality
        END),
    _IQDuringOpt =MAX( 
        CASE IQS.SequenceType
            WHEN 2 THEN IQS.ImageQuality
        END),
    _IQAfterOpt = 
        MAX(CASE IQS.SequenceType
            WHEN 3 THEN IQS.ImageQuality
        END),
 --  ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY))) AS SeqPI,
   _SeqPIBeforeOpt = MAX(
        CASE IQS.SequenceType
            WHEN 1 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
       END),
  _SeqPIDuringOpt =MAX(
      CASE IQS.SequenceType
          WHEN 2 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END),
  _SeqPIAfterOpt =MAX(
      CASE IQS.SequenceType
          WHEN 3 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END)       
FROM ClientSequence CS
JOIN ImageQualitySequencePool IQS ON CS.Id = IQS.SequenceId
JOIN ClientProtocol CP ON CS.ClientProtocolId = CP.Id
JOIN Team S ON CP.TeamID = S.Id
JOIN Machine M On CP.MachineId = M.Id
JOIN Vendor V ON M.VendorId = V.Id
OUTER APPLY(SELECT ','+LTRIM(QS.SequenceType) FROM ImageQualitySequencePool AS QS WHERE CS.Id = QS.SequenceId FOR XML PATH('')) OQS(types)
GROUP BY CS.Id, CP.Country, S.Title ,CONCAT(V.Title,' ',M.Model,' ',M.Version),CP.MachineId,CP.Title,CS.Title,oQS.types
ORDER BY CP.Country,S.Title,CP.MachineId,CP.Title,CS.Title--,IQS.SequenceType

【讨论】：

试过但没用。不过会更接近您的查询

【解决方案3】：

试试这个查询：

SELECT
    CP.Country As Country,
    S.Title As Site,
    CONCAT(V.Title,' ',M.Model,' ',M.Version) As Machine,
    CP.Title As Protocol,
    CS.Title As Sequence,
    STUFF(CONCAT(MAX(IIF(IQS.SequenceType = 1, ',1', '')), MAX(IIF(IQS.SequenceType = 2, ',2', '')), MAX(IIF(IQS.SequenceType = 3, ',3', ''))),1,1,'') AS Type,
    _IQBeforeOpt =
        MAX(CASE IQS.SequenceType
            WHEN 1 THEN IQS.ImageQuality
        END),
    _IQDuringOpt = 
        MAX(CASE IQS.SequenceType
            WHEN 2 THEN IQS.ImageQuality
        END),
    _IQAfterOpt = 
        MAX(CASE IQS.SequenceType
            WHEN 3 THEN IQS.ImageQuality
        END),
   _SeqPIBeforeOpt = 
        MAX(CASE IQS.SequenceType
            WHEN 1 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
       END),
  _SeqPIDuringOpt =
      MAX(CASE IQS.SequenceType
          WHEN 2 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END),
  _SeqPIAfterOpt =
      MAX(CASE IQS.SequenceType
          WHEN 3 THEN ((IQS.ImageQuality*IQS.SeqSlices)/(IQS.SeqTime*(CS.slicetkndef*CS.pixelspacingX*CS.pixelspacingY)))
      END)      
FROM 
    ClientSequence CS
    JOIN ImageQualitySequencePool IQS ON CS.Id = IQS.SequenceId
    JOIN ClientProtocol CP ON CS.ClientProtocolId = CP.Id
    JOIN Team S ON CP.TeamID = S.Id
    JOIN Machine M On CP.MachineId = M.Id
    JOIN Vendor V ON M.VendorId = V.Id
GROUP BY CP.Country, S.Title, CONCAT(V.Title,' ',M.Model,' ',M.Version), CP.Title, CS.Title

【讨论】：