Question

我的数据如下：

WorkTable
---------
WorkId  
1  
2  

ProcessTable
WorkId    Process    ProcessDate  
1         Receive    1/20/2017  
1         Triage     1/21/2017  
1         Triage     1/22/2017  
2         Receive    1/20/2017

我想创建一个如下所示的视图：

WorkId    Received    Triaged  
1         1/20/2017   1/22/2017  
2         1/20/2017   NULL

我只需要最新的流程执行日期，因此Id 1我为Triage选择了1/22，因为它是最新的

我当前的查询看起来像这样，但效率非常低

Select w.WorkId,    
(SELECT     TOP (1) ProcessDate
     FROM          ProcessTable
     WHERE      (w.WorkId = WorkId) AND (Process = 'Receive')
     ORDER BY ProcessDate DESC) AS Received,
(SELECT     TOP (1) ProcessDate
     FROM          ProcessTable
     WHERE      (w.WorkId = WorkId) AND (Process = 'Triage')
     ORDER BY ProcessDate DESC) AS Triaged
From WorkTable w

我可以做些什么来加快速度？

Answer 1

更新：使用外部apply()：

select 
    w.WorkId
  , r.Received
  , r.ReceivedBy
  , t.Triaged
  , t.TriagedBy
from WorkTable
  outer apply (
    select top 1 
        Received = i.ProcessDate 
      , ReceivedBy = i.UserName
    from ProcessTable i
    where i.WorkId = w.WorkId
      and i.Process = 'Receive'
    order by i.ProcessDate desc
    ) r
  outer apply (
    select top 1 
        Triaged = i.ProcessDate 
      , TriagedBy = i.UserName
    from ProcessTable i
    where i.WorkId = w.WorkId
      and i.Process = 'Triage'
    order by i.ProcessDate desc
    ) t

使用条件聚合：

select 
    WorkId 
  , Received = max(case when Process='Receive' then ProcessDate end)
  , Triaged  = max(case when Process='Triage'  then ProcessDate end)
from ProcessTable w
group by WorkId

rextester演示：http://rextester.com/HNB72821

返回（rextester重新格式化日期时间值，将其格式化回来）：

+--------+------------+------------+
| WorkId |  Received  |  Triaged   |
+--------+------------+------------+
|      1 | 2017-01-20 | 2017-01-22 |
|      2 | 2017-01-20 | NULL       |
+--------+------------+------------+

或与pivot()一样：

select 
    WorkId 
  , Receive
  , Triage 
from ProcessTable w
pivot (max(ProcessDate) for Process in ([Receive],[Triage]))p

如果ProcessTable中还有其他未使用的列，您可能需要选择pivot()所需的列，如下所示：

select 
    WorkId 
  , Receive
  , Triage 
from (select WorkId, Process, ProcessDate from ProcessTable) w
pivot (max(ProcessDate) for Process in ([Receive],[Triage]))p

Answer 2

您可以使用连接和条件聚合。但是，这很棘手，因为排序是CreatedDate，但您想要的字段是ProcessDate。

我建议您按原样保留查询并创建最佳索引索引：ProcessDate(WorkId, Process, CreatedDate, ProcessDate)。

SQL在视图中将一对多子查询到单个数据列的最佳方式

2 个答案: