查找状态字段与当前字段不同的下一条记录

Find next record where status field is different from current

我有一个 table 用于记录事件。具体有两种类型:ON 和 OFF。

有时会有重叠的日志条目,因为可以同时记录 2 个设备。这并不重要,因为最终报告应该 [大部分] 正确概述开启 -> 关闭期间。

下面是一个示例,第三列只是为了说明:它不存在。

ActionTaken    ID   ID_of_next_OFF
Switched ON    1    3
Switched ON    2    6
Switched OFF   3    
Switched ON    4    7
Switched ON    5    8
Switched OFF   6    
Switched OFF   7    
Switched OFF   8    
Switched On    9    10
Switched OFF   10   
Switched On    11   12
Switched OFF   12   

给定前两列,如何计算第三列?

这不起作用:

SELECT actionTaken, Id, LEAD(Id) 
OVER (PARTITION BY ActionTaken ORDER BY ID) nextConn 
FROM dbo.Events

因为它基于 ID_of_Next 下一个匹配的 actionTaken 值,而不是下一个替代值。

像这样的东西应该会让你到达那里。

下面我使用了 2 个 CTE 来拆分关闭和打开数据,然后为第一次打开第一次关闭提供了一个排名项,然后我使用联合查询来匹配它们

declare @Events table (
    ActionTaken nvarchar(25),
    ID int
);

insert @Events
values

--ActionTaken    ID   ID_of_next_OFF
('Switched ON' ,   1), -- 3
('Switched ON' ,   2),-- 6
('Switched OFF',   3),    
('Switched ON' ,   4),--    7
('Switched ON' ,   5),--    8
    ('Switched OFF',   6),    
    ('Switched OFF',   7),    
    ('Switched OFF',   8),    
    ('Switched On' ,   9),--    10
    ('Switched OFF',   10),   
    ('Switched On' ,   11),--   12
    ('Switched OFF',   12);

    with onrank as (
    select row_number()over(order by id) ranking, * from @Events where ActionTaken like '%ON')
    , offrank as (
    select row_number()over(order by id) ranking, * from @Events where ActionTaken like '%OFF')

    select o.ActionTaken, o.ID, case when o.ranking=f.ranking then cast(f.id as nvarchar(3)) end as Id_next_off
    from onrank o inner join offrank f on o.ranking=f.ranking
    union
    select ActionTaken, ID, '' from offrank
    order by o.ID;

你走对了。您只需要 'Switched ON' 部分的 LEFT JOIN'Switched OFF' 部分在相同的行号上。

with Events as (
  select 'Switched ON' as ActionTaken, 1 as ID union all -- 3
  select 'Switched ON', 2 union all -- 6
  select 'Switched OFF', 3 union all
  select 'Switched ON', 4 union all -- 7
  select 'Switched ON', 5 union all -- 8
  select 'Switched OFF', 6 union all
  select 'Switched OFF', 7 union all
  select 'Switched OFF', 8 union all
  select 'Switched On', 9 union all -- 10
  select 'Switched OFF', 10 union all
  select 'Switched On', 11 union all -- 12
  select 'Switched OFF', 12
), E as (
  select
    *, row_number() over(partition by ActionTaken order by ID) as rn
  from Events
)
select
  a.ActionTaken, a.ID, b.ID
from E as a
left join E as b
  on a.ActionTaken = 'Switched ON' and
     b.ActionTaken = 'Switched OFF' and
     a.rn = b.rn
order by a.ID, a.ActionTaken;

输出:

+--------------+----+------+
| ActionTaken  | ID |  ID  |
+--------------+----+------+
| Switched ON  |  1 | 3    |
| Switched ON  |  2 | 6    |
| Switched OFF |  3 | NULL |
| Switched ON  |  4 | 7    |
| Switched ON  |  5 | 8    |
| Switched OFF |  6 | NULL |
| Switched OFF |  7 | NULL |
| Switched OFF |  8 | NULL |
| Switched On  |  9 | 10   |
| Switched OFF | 10 | NULL |
| Switched On  | 11 | 12   |
| Switched OFF | 12 | NULL |
+--------------+----+------+

使用 SQL Fiddle 在线测试。