在不连续的时间段 YYYYMM00 上查找 MIN 和 MAX 日期答案

【问题标题】：Find MIN and MAX date on a discontinuous period of time YYYYMM00在不连续的时间段 YYYYMM00 上查找 MIN 和 MAX 日期
【发布时间】：2019-05-12 14:11:12
【问题描述】：

我有以下两张表：

DimensionTime 是一个包含每个月的表格，按 ID 排序，采用 YYYMM00 形式
LogPlayer 是一个表格，其中包含与玩家和特定月份相关的一些统计信息。

我想要得到的是以下内容：

+--------+--------+----------+----------+
| Player |  Team  |  Start   |   End    |
+--------+--------+----------+----------+
| John   | Red    | 20180100 | 20180300 |
| John   | Red    | 20180600 | 20180700 |
| Luke   | Yellow | 20180100 | 20180100 |
| Luke   | Yellow | 20190100 | 20190100 |
+--------+--------+----------+----------+

我不能使用 MIN 和 MAX 函数，因为周期不连续...我该如何解决？我已经尝试将 MIN/MAX 与 GROUP BY 结合使用，但我没有得到任何有用的信息。我在 Stackoverflow 上找不到任何问题或答案。

SELECT *
    INTO #DimensionTime
    FROM (
        SELECT 1 AS [ID], 20180100 AS [TIMEID]
        UNION ALL
        SELECT 2 AS [ID], 20180200 AS [TIMEID]
        UNION ALL
        SELECT 3 AS [ID], 20180300 AS [TIMEID]
        UNION ALL
        SELECT 4 AS [ID], 20180400 AS [TIMEID]
        UNION ALL
        SELECT 5 AS [ID], 20180500 AS [TIMEID]
        UNION ALL
        SELECT 6 AS [ID], 20180600 AS [TIMEID]
        UNION ALL
        SELECT 7 AS [ID], 20180700 AS [TIMEID]
        UNION ALL
        SELECT 8 AS [ID], 20180800 AS [TIMEID]
        UNION ALL
        SELECT 9 AS [ID], 20180900 AS [TIMEID]
        UNION ALL
        SELECT 10 AS [ID], 20181000 AS [TIMEID]
        UNION ALL
        SELECT 11 AS [ID], 20181100 AS [TIMEID]
        UNION ALL
        SELECT 12 AS [ID], 20181200 AS [TIMEID]
        UNION ALL
        SELECT 13 AS [ID], 20190100 AS [TIMEID]
        UNION ALL
        SELECT 14 AS [ID], 20190200 AS [TIMEID]
        UNION ALL
        SELECT 15 AS [ID], 20190300 AS [TIMEID]
    ) A

SELECT *
INTO #LogPlayer
FROM (
    SELECT 'John' AS [Player], 'Red' AS [Team], 20180100 AS [TIMEID]
    UNION ALL
    SELECT 'John' AS [Player], 'Red' AS [Team], 20180200 AS [TIMEID]
    UNION ALL
    SELECT 'John' AS [Player], 'Red' AS [Team], 20180300 AS [TIMEID]
    UNION ALL
    SELECT 'John' AS [Player], 'Red' AS [Team], 20180600 AS [TIMEID]
    UNION ALL
    SELECT 'John' AS [Player], 'Red' AS [Team], 20180700 AS [TIMEID]
    UNION ALL
    SELECT 'Luke' AS [Player], 'Yellow' AS [Team], 20180100 AS [TIMEID]
    UNION ALL
    SELECT 'Luke' AS [Player], 'Yellow' AS [Team], 20190100 AS [TIMEID]
) B

【问题讨论】：

您是否真的在使用 SQL Server，尤其是 SQL Server 2005（已不再支持）？ SQL Server 不支持yyyyMM00 日期；这是 MySQL 的事情。
timeID varchar 是什么数据类型？
@Larnu 我认为他的字段是 INT 或 BIG INT 类型，无论如何如果 SQL Server 2005 确实不是最佳方案
除了奇怪的“日期”之外，这似乎是您需要对连续日期进行分组的情况。 Jeff Moden 有一个很好的方法来处理这个here。您只需要稍微修改一下以使用月份而不是日期。
@Lamu 是的，不幸的是我使用的是老式 SQL Server 2005 和 YYYYMM00，由于这些原因，存储为 VARCHAR。如果有帮助，我可以通过链接服务器在新的 SQL Server 2018 上运行存储过程。

标签： sql sql-server datetime sql-server-2005

【解决方案1】：

这是一种孤岛问题。即使在 SQL Server 2005 等不受支持的古老软件中也可以解决，因为该版本具有 row_number()。

一个技巧是将时间 ID 转换为真正日期/时间。另一个技巧是通过从日期/时间值中减去连续的月份数来定义组：

select player, team, min(timeid), max(timeid)
from (select lp.*,
             row_number() over (partition by player, team order by timeid) as seqnum,
             cast(cast(timeid + 1 as varchar(255)) as datetime) as yyyymm
      from logplayer lp
     ) lp
group by player, team, dateadd(month, - seqnum, yyyymm)
order by player, team, min(timeid);

Here 是一个 dbfiddle。

【讨论】：

【解决方案2】：

您可以执行此类操作来查找日期运行的开始和结束。

通过 CTE 转换为“日期”（我认为它存在于 2005 年）然后使用 Cross Apply an EXIST 来查找日期运行的开始和结束

你没有提供球员和球队的数据，但你可以添加到 EXISTS 中的 WHERE 条件，然后 GROUP BY - 如果需要

;WITH dats as (SELECT CAST(LEFT(timeid, 6) + '01' as datetime) as DT from #DimensionTime)
    select CONVERT(varchar(7),d1.DT,112) +'0'  as strt, 
           CONVERT(varchar(7),dq.dt,112) +'0' as [end] from dats d1 
        CROSS  APPLY
          (SELECT TOP 1 d3.dt from dats d3 where 
                                            d3.dt >  d1.dt 
                                            and 
                                            not exists(
                                                    select 0 from dats d4 where d4.DT = dateadd(month,1,d3.DT)
                                                      )
        ORDER BY d3.dt asc) DQ           
        where not exists(select 0 from dats d2 where d2.DT = dateadd(month,-1,d1.DT)) ;

对一些样本数据进行猜测，我尝试了

SELECT *
INTO #DimensionTime
FROM (
    SELECT 1 AS [ID], 20180100 AS [TIMEID], 'john' as player, 'red' as team 
    UNION ALL
    SELECT 2 AS [ID], 20180200 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 3 AS [ID], 20180300 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 4 AS [ID], 20180400 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 5 AS [ID], 20180500 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 7 AS [ID], 20180700 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 8 AS [ID], 20180800 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 9 AS [ID], 20180900 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 11 AS [ID], 20181100 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 12 AS [ID], 20181200 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 13 AS [ID], 20190100 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 14 AS [ID], 20190200 AS [TIMEID], 'john','red'
    UNION ALL
    SELECT 15 AS [ID], 20190300 AS [TIMEID], 'john','red'
    UNION ALL 
    SELECT 1 AS [ID], 20180100 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 2 AS [ID], 20180200 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 4 AS [ID], 20180400 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 5 AS [ID], 20180500 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 8 AS [ID], 20180800 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 9 AS [ID], 20180900 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 12 AS [ID], 20181200 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 13 AS [ID], 20190100 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 14 AS [ID], 20190200 AS [TIMEID], 'luke','yellow'
    UNION ALL
    SELECT 15 AS [ID], 20190300 AS [TIMEID], 'luke','yellow'


) A



;WITH dats as (SELECT CAST(LEFT(timeid, 6) + '01' as datetime) as DT,player,team from #DimensionTime)
    select d1.team,d1.player,
            CONVERT(varchar(7),d1.DT,112) +'0'  as strt, 
           CONVERT(varchar(7),dq.dt,112) +'0' as [end] from dats d1 
        CROSS  APPLY
          (SELECT TOP 1 d3.dt from dats d3 where 
                                            d3.dt >  d1.dt 
                                            and
                                            d3.player = d1.player
                                            and
                                            d3.team = d1.team
                                            and 
                                            not exists(
                                                    select 0 from dats d4 where d4.DT = dateadd(month,1,d3.Dt)
                                                         and  d4.team = d3.team
                                                         and d4.player = d3.player
                                                      )
        ORDER BY d3.dt asc) DQ           
        where not exists(select 0 from dats d2 where 
                    d2.player=d1.player
                    and
                    d2.team = d1.team
                    and 
                    d2.DT = dateadd(month,-1,d1.DT) and d1.team=d2.team and d1.player = d2.player ) ;

drop table #DimensionTime;

很抱歉我错过了最新的表格，我设计了

;WITH dats as (SELECT CAST(LEFT(timeid, 6) + '01' as datetime) as DT,player,team from #LogPlayer)
    select d1.team,d1.player,
            CONVERT(varchar(7),d1.DT,112) +'0'  as strt, 
           CONVERT(varchar(7),dq.dt,112) +'0' as [end] from dats d1 
        CROSS  APPLY
          (SELECT TOP 1 d3.dt from dats d3 where 
                                            d3.dt >  d1.dt 
                                            and
                                            d3.player = d1.player
                                            and
                                            d3.team = d1.team
                                            and 
                                            not exists(
                                                    select 0 from dats d4 where d4.DT = dateadd(month,1,d3.Dt)
                                                         and  d4.team = d3.team
                                                         and d4.player = d3.player
                                                      )
        ORDER BY d3.dt asc) DQ           
        where not exists(select 0 from dats d2 where 
                    d2.player=d1.player
                    and
                    d2.team = d1.team
                    and 
                    d2.DT = dateadd(month,-1,d1.DT) and d1.team=d2.team and d1.player = d2.player ) ;

【讨论】：

抱歉，我不明白如何运行它。我已经运行了代码，我只得到了 strt 20180100 AS [strt] 和 20190300 AS [end]，这与计算 #DimensionTime 表的 MIN 和 MAX 相同
那么您希望从样本数据中列出哪些日期？ 20180100 是连续月份运行的开始，20190300 是运行的结束 - 如果删除中间的记录，您将获得两个序列
我想要每个球员和每支球队的每个时期的开始/结束数据，就像第一个问题中的示例一样。
您的样本数据中没有提到任何球队或球员 - 我会尝试为您添加一些
对不起，我在这里看到了一些数据，我添加了这样做的尝试