SQL - 将许多类别分组为几个类别答案

【问题标题】：SQL - Group many categories into few categoriesSQL - 将许多类别分组为几个类别
【发布时间】：2017-10-04 03:29:08
【问题描述】：

我有一个问题。它返回行，其中一列是状态列，一列是经过时间列。有大约 8 个状态类别。我想将某些状态及其经过的时间总结为〜3个类别。有人知道我该怎么做吗？

这是当前查询，它按天将条目拆分为 bin。

declare @STARDTZ as datetime2
declare @LASTDTZ as datetime2

declare @STARTDT as datetime2
declare @LASTDT as datetime2

set @LASTDT = getdate()
set @STARTDT = DATEADD(dd,-7,getdate())

SET @STARDTZ = DATEADD(dd, DATEDIFF(dd, 0, @STARTDT), 0) --set the start time to 12am (zero) given input datetime
SET @LASTDTZ = DATEADD(dd, DATEDIFF(dd, 0, @LASTDT), 0) --set the stop time to 12am (zero) given input stoptime
;
with dates(dt) as (select @STARDTZ union all select dateadd(day,1,dt) from dates where dt < @LASTDTZ)
select t.Name
,case when t.StartDT>=d.dt then t.StartDT else d.dt end as StartDT
,case when datediff(day,LastDT,dt)=0 then t.LastDT else dateadd(day,1,d.dt) end as LastDT 
,datediff(millisecond,case when t.StartDT>=d.dt then t.StartDT else d.dt end
,case when datediff(day,LastDT,dt)=0 then t.LastDT else dateadd(day,1,d.dt) end)/60000.0 as IntervalMin
,t.FullName
,DATEADD(dd, DATEDIFF(dd, 0, dt), 0) as DateBin

from 
(
    SELECT
    CoatingChamberStates.Name as Name
    ,c.StartDT as StartDT
    ,c.LastDT as LastDT
    ,c.IntervalMin
    ,CoatingChambers.FullName
    ,DATEADD(dd, DATEDIFF(dd, 0, StartDT), 0) as DateBin

    FROM [REO].[dbo].CoatingChamberStateLogs as c
    INNER JOIN CoatingChamberStates on c.CoatingChamberStatesID = CoatingChamberStates.CoatingChamberStatesID
    INNER JOIN CoatingChambers on c.CoatingChambersID = CoatingChambers.CoatingChambersID

    where c.StartDT >= @STARDTZ
    and c.LastDT <= @LASTDTZ


) as t

join dates d on d.dt >= cast(t.StartDT as date) and d.dt<=cast(t.LastDT as date)

group by DateBin, StartDT, Name, dt, LastDT, FullName

这是它返回的内容：

Name    StartDT LastDT  IntervalMin FullName    DateBin
UNKNOWN 2017-04-28 00:05:31.1335380 2017-04-28 11:36:40.1200328 691.1497833 SPECTOR7                2017-04-28 00:00:00.000
UNKNOWN 2017-04-28 00:06:31.6555661 2017-04-28 00:07:31.5121862 0.9976166   LOTUS3                  2017-04-28 00:00:00.000
IDLE    2017-04-28 00:07:31.5121862 2017-04-28 11:53:39.7066464 706.1365666 LOTUS3                  2017-04-28 00:00:00.000
UNKNOWN 2017-04-28 00:50:32.1654631 2017-04-28 00:51:32.0670071 0.9983666   LOTUS1                  2017-04-28 00:00:00.000
RUN_PARTS   2017-04-28 00:51:32.0670071 2017-04-28 00:53:32.0966491 2.0004833   LOTUS1                  2017-04-28 00:00:00.000
IDLE    2017-04-28 00:53:32.0966491 2017-04-28 05:50:36.3819295 297.0714166 LOTUS1                  2017-04-28 00:00:00.000
RUN_OTHER   2017-04-28 00:53:32.1747858 2017-04-28 12:03:40.0639753 670.1314833 OXFORD2                 2017-04-28 00:00:00.000
IDLE    2017-04-28 01:00:32.3043454 2017-04-28 02:26:33.3361640 86.0172000  SPECTOR1                2017-04-28 00:00:00.000
IDLE    2017-04-28 01:50:32.8690160 2017-04-28 07:07:37.3384170 317.0744833 IONTECH                 2017-04-28 00:00:00.000
IDLE    2017-04-28 01:57:32.9759673 2017-04-28 02:36:33.6234811 39.0108000  LOTUS8                  2017-04-28 00:00:00.000
IDLE    2017-04-28 02:04:33.0037095 2017-04-28 03:10:33.8922959 66.0148166  LOTUS6                  2017-04-28 00:00:00.000
UNKNOWN 2017-04-28 02:12:33.2932821 2017-04-28 02:13:33.1640310 0.9978500   SPECTOR6                2017-04-28 00:00:00.000
IDLE    2017-04-28 02:12:33.3089156 2017-04-28 05:27:36.3023807 195.0499000 SPECTOR8                2017-04-28 00:00:00.000
RUN_PARTS   2017-04-28 02:13:33.1640310 2017-04-28 02:31:33.4447516 18.0046666  SPECTOR6                2017-04-28 00:00:00.000
RUN_TEST    2017-04-28 02:26:33.3361640 2017-04-28 03:43:34.4053900 77.0178166  SPECTOR1                2017-04-28 00:00:00.000
IDLE    2017-04-28 02:31:33.4447516 2017-04-28 04:27:35.0346386 116.0265000 SPECTOR6                2017-04-28 00:00:00.000

我想将某些类别的经过时间合并为 3 个类别：“idle”、“running”和“waste”，例如，“run_parts”和“run_test”将汇总为“running”。

这有意义吗？我该如何处理？谢谢！

【问题讨论】：

标签： sql sql-server if-statement sum conditional

【解决方案1】：

您可以使用case 表达式将名称排序到派生表中所需的存储桶中：

with dates(dt) as (select @stardtz union all select dateadd(day,1,dt) from dates where dt < @lastdtz)
  select
     t.Name
    ,case when t.Startdt>=d.dt then t.Startdt else d.dt end as Startdt
    ,case when datediff(day,Lastdt,dt)=0 then t.Lastdt else dateadd(day,1,d.dt) end as Lastdt 
    ,datediff(millisecond,case when t.Startdt>=d.dt then t.Startdt else d.dt end
    ,case when datediff(day,Lastdt,dt)=0 then t.Lastdt else dateadd(day,1,d.dt) end)/60000.0 as IntervalMin
    ,t.FullName
    ,dateadd(day, datediff(day, 0, dt), 0) as DateBin
  from (
    select
    case when CoatingChamberStates.Name in ('run_parts','run_test') then 'running'
      when CoatingChamberStates.Name = 'unknown' then 'waste'
      else CoatingChamberStates.Name
      end as Name 
    ,c.Startdt as Startdt
    ,c.Lastdt as Lastdt
    ,c.IntervalMin
    ,CoatingChambers.FullName
    ,dateadd(day, datediff(day, 0, Startdt), 0) as DateBin
    from [reo].[dbo].CoatingChamberStateLogs as c
      inner join CoatingChamberStates 
        on c.CoatingChamberStatesid = CoatingChamberStates.CoatingChamberStatesid
      inner join CoatingChambers 
        on c.CoatingChambersid = CoatingChambers.CoatingChambersid
    where c.Startdt >= @stardtz
      and c.Lastdt <= @lastdtz
    ) as t
    join dates d 
      on d.dt >= cast(t.Startdt as date) 
     and d.dt<=cast(t.Lastdt as date)
  group by DateBin, Startdt, dt, Lastdt, FullName, t.Name

或者你可以用select 和group by case 表达式代替t.Name：

with dates(dt) as (select @stardtz union all select dateadd(day,1,dt) from dates where dt < @lastdtz)
  select
     case when t.Name in ('run_parts','run_test') then 'running'
      when t.Name = 'unknown' then 'waste'
      else t.Name
      end as Name
    ,case when t.Startdt>=d.dt then t.Startdt else d.dt end as Startdt
    ,case when datediff(day,Lastdt,dt)=0 then t.Lastdt else dateadd(day,1,d.dt) end as Lastdt 
    ,datediff(millisecond,case when t.Startdt>=d.dt then t.Startdt else d.dt end
    ,case when datediff(day,Lastdt,dt)=0 then t.Lastdt else dateadd(day,1,d.dt) end)/60000.0 as IntervalMin
    ,t.FullName
    ,dateadd(day, datediff(day, 0, dt), 0) as DateBin
  from (
    select
    CoatingChamberStates.Name as Name
    ,c.Startdt as Startdt
    ,c.Lastdt as Lastdt
    ,c.IntervalMin
    ,CoatingChambers.FullName
    ,dateadd(day, datediff(day, 0, Startdt), 0) as DateBin
    from [reo].[dbo].CoatingChamberStateLogs as c
      inner join CoatingChamberStates 
        on c.CoatingChamberStatesid = CoatingChamberStates.CoatingChamberStatesid
      inner join CoatingChambers 
        on c.CoatingChambersid = CoatingChambers.CoatingChambersid
    where c.Startdt >= @stardtz
      and c.Lastdt <= @lastdtz
    ) as t
    join dates d 
      on d.dt >= cast(t.Startdt as date) 
     and d.dt<=cast(t.Lastdt as date)
  group by DateBin, Startdt, dt, Lastdt, FullName
    , case when t.Name in ('run_parts','run_test') then 'running'
           when t.Name = 'unknown' then 'waste'
      else t.Name
      end

【讨论】：