用组的上一个填充日期的数据值填充缺失的日期

Mar*_*man 13 sql-server optimization sql-server-2012 date

图片在部门之间转移的服务台票。我们想知道在票打开的每一天,每个票在一天结束时的部门是什么。该表包含每张票的最后一个部门,在该部门发生变化的每一天开放(包括该票最初打开日期和关闭日期的一行)。数据表如下所示:

CREATE TABLE TicketAssigment (
    TicketId     INT NOT NULL,
    AssignedDate DATE NOT NULL,
    DepartmentId INT NOT NULL);
Run Code Online (Sandbox Code Playgroud)

我需要的是使用按日期排序的前一个 TicketAssigment 行中的 DepartmentId 为每个 TicketId 填写任何缺失的日期。

如果我有像这样的 TicketAssigment 行:

1, '1/1/2016', 123 -- Opened
1, '1,4,2016', 456 -- Transferred and closed
2, '1/1/2016', 25  -- Opened
2, '1/2/2016', 52  -- Transferred
2, '1/4/2016', 25  -- Transferred and closed
Run Code Online (Sandbox Code Playgroud)

我想要这个输出:

1, '1/1/2016', 123
1, '1/2/2016', 123
1, '1/3/2016', 123
1, '1/4/2016', 456
2, '1/1/2016', 25
2, '1/2/2016', 52
2, '1/3/2016', 52
2, '1/4/2016', 25
Run Code Online (Sandbox Code Playgroud)

这看起来可能接近我所需要的,但我没有耐心让它完成,并且估计的计划成本有 6 位数字:

SELECT  l.TicketId, c.Date, MIN(l.DepartmentId)
FROM    dbo.Calendar c 
        OUTER APPLY (SELECT TOP 1 TicketId, DepartmentId FROM TicketAssigment WHERE AssignedDate <= c.Date ORDER BY AssignedDate DESC) l
WHERE   c.Date <= (SELECT MAX(AssignedDate) FROM TicketAssigment)
GROUP   BY l.TicketId, c.Date
ORDER   BY l.TicketId, c.Date;
Run Code Online (Sandbox Code Playgroud)

我怀疑有一种方法可以使用 LAG 和窗口框架来做到这一点,但我还没有完全弄清楚。满足要求的更有效方法是什么?

Rob*_*ley 15

使用LEAD()得到TicketId分区中的下一行。然后加入日历表以获取之间的所有日期。

WITH TAwithnext AS
(SELECT *, LEAD(AssignmentDate) OVER (PARTITION BY TicketID ORDER BY AssignmentDate) AS NextAssignmentDate
 FROM TicketAssignment
)
SELECT t.TicketID, c.Date, t.DepartmentID
FROM dbo.Calendar c
JOIN TAwithnext t
    ON c.Date BETWEEN t.AssignmentDate AND ISNULL(DATEADD(day,-1,t.NextAssignmentDate),t.AssignmentDate)
;
Run Code Online (Sandbox Code Playgroud)

各种获取Calendar表的方法...


Kin*_*hah 4

这是一种快速的方法(我尚未测试性能或可扩展性)

-- 创建日历表

-- borrowed from @Aaron's post http://sqlperformance.com/2013/01/t-sql-queries/generate-a-set-3 
CREATE TABLE dbo.Calendar(d DATE PRIMARY KEY);

INSERT dbo.Calendar(d) SELECT TOP (365)
 DATEADD(DAY, ROW_NUMBER() OVER (ORDER BY number)-1, '20160101')
 FROM [master].dbo.spt_values
 WHERE [type] = N'P' ORDER BY number;
Run Code Online (Sandbox Code Playgroud)

--- 创建您的测试表

CREATE TABLE dbo.TicketAssigment (
    TicketId     INT NOT NULL,
    AssignedDate DATE NOT NULL,
    DepartmentId INT NOT NULL);

--  truncate table dbo.TicketAssigment;

insert into dbo.TicketAssigment values (1   ,   '1-1-2016'  ,   123 )
insert into dbo.TicketAssigment values (1   ,   '1-4-2016'  ,   456 )
insert into dbo.TicketAssigment values (2   ,   '1-1-2016'  ,   25  )
insert into dbo.TicketAssigment values (2   ,   '1-2-2016'  ,   52  )
insert into dbo.TicketAssigment values (2   ,   '1-4-2016'  ,   25  )
Run Code Online (Sandbox Code Playgroud)

--- 查询以获得所需的输出

;with Cte as
(
  select TicketID, 
         min(AssignedDate) minAD, -- This is the min date
         max(AssignedDate) maxAD  -- This is the max date
  from TicketAssigment
  group by TicketID
)
select Cte.TicketID,
       c.d as AssignedDate,

       ( -- Get DeptID
       select top(1) T.departmentID
       from dbo.TicketAssigment as T
       where T.TicketID = cte.TicketID and
             T.AssignedDate <= c.d
       order by T.AssignedDate desc
       ) as DepartmentID
from Cte
  left outer join dbo.Calendar as c
      on c.d between Cte.minAD and Cte.maxAD
    order by Cte.TicketID
Run Code Online (Sandbox Code Playgroud)

在此输入图像描述