从一天中的某个小时起按日期分组

Question

SQLServer 2008r2。我有一张桌子，每天每小时过去10分钟就有一张唱片。工作每运行一小时，它也会输入48条记录，代表对未来48小时内可能发生的事情的预测。注 - 在它进入48小时预报之前，它将删除上次输入的预测。因此，尽管它每小时进行48小时的预测，但系统中只有一次预测。在表中的相关字段是这样的：

currentScore  obsDate 
    9   2017-06-22 08:10:00 
    9   2017-06-22 07:10:00 
    9   2017-06-22 06:10:00 
    10   2017-06-22 05:10:00 
    ...  ... 

我如何可以查询该表，并按照日期分组从一天中的某个时间？我希望这一天在上午6点开始，并在当天上午6点结束。我只需要表中的五条记录，一天前两位，未来两位。所以如果它的6月20日我想要6月18日，19日，20日，21日和22日。这是按日历日获得正确结果的查询。

SELECT cast(obsDate AS DATE) AS theDate 
    ,sum(CASE 
      WHEN currentScore < 8 
       THEN 1 
      ELSE 0 
      END) AS currentscore_low 
    ,sum(CASE 
      WHEN currentScore >= 8 
       AND currentScore < 17 
       THEN 1 
      ELSE 0 
      END) AS currentscore_medium 
    ,sum(CASE 
      WHEN currentScore >= 17 
       THEN 1 
      ELSE 0 
      END) AS currentscore_high 
FROM diseaseScores 
WHERE siteID = 8315 
    AND obsDate >= cast(getdate() - 2 AS DATE) 
GROUP BY cast(obsDate AS DATE) 
ORDER BY cast(obsDate AS DATE); 

返回这个结果：

theDAte  low med high 
2017-06-18  23 0  0 
2017-06-19  22 0  0 
2017-06-20  5  19 0 
2017-06-21  0  24 0 
2017-06-22  0  9  0 

有一个新的要求得到同样的结果，而是由集团和随后数必须是从早上6点到早上6点。例如：

第一个记录应该是从2017-06-17 06:00到2017-06-18 06:00 am 第二个记录应该是从2017-06-18 06:00到2017-06-19 06 ：00am .... etc

我该怎么做？在此先感谢

UPDATE，我已经做了两两件事：

1..introduce蒂姆斯想法

2..I还增加额外的字段“numOfScores”显示有多少个小时的数据 的每一行代表

select 
     cast(dateadd(hour, -6, obsDate) as date) as theDate, count(currentScore) as numOfScores, 
     sum(case when currentScore < 8 then 1 else 0 end) as currentscore_low, 
     sum(case when currentScore >= 8 and currentScore < 17 
       then 1 else 0 end) as currentscore_medium, 
     sum(case when currentScore >= 17 then 1 else 0 end) as currentscore_high 
    from diseaseScores 
    where siteID = 8315 and 
      obsDate >= cast(getdate() - 2 as date) 
    group by cast(dateadd(hour, -6, obsDate) as date) 
    order by cast(dateadd(hour, -6, obsDate) as date); 

现在我得到这样的结果：

2017-06-18 5 5 0 0 
    2017-06-19 24 23 1 0 
    2017-06-20 24 1 23 0 
    2017-06-21 24 8 16 0 
    2017-06-22 24 1 23 0 
    2017-06-23 9 0 9 0 

这告诉我，2017-06-18只有5小时值得分。我希望这第一行24小时值得。从17日上午6点到18日上午6点。这让我觉得我没有得到的结果，我希望

的23只具有9-11个小时OK，因为这是最近期的预测

更新：

我不认为它很容易在一个查询完成（如果可能的话），所以我会只使用五个查询，并特别指出日期&次来获得我的结果。这里e.g是前两个条件：

这可能会在这里工作

select 
     sum(case when currentScore < 9 then 1 else 0 end) as numOfLOWRecs, 
     sum(case when currentScore > 8 and currentScore < 17 then 1 else 0 end) as currentscore_medium, 
     sum(case when currentScore >= 17 then 1 else 0 end) as currentscore_high 
    from diseaseScores where siteID = 9999 
    and obsDate >= '2017-06-18 06:00' and obsDate < '2017-06-19 06:00' 

    select 
     sum(case when currentScore < 9 then 1 else 0 end) as numOfLOWRecs, 
     sum(case when currentScore > 8 and currentScore < 17 then 1 else 0 end) as currentscore_medium, 
     sum(case when currentScore >= 17 then 1 else 0 end) as currentscore_high 
    from diseaseScores where siteID = 9999 
    and obsDate >= '2017-06-19 06:00' and obsDate < '2017-06-20 06:00' 

Answer 1

一招是简单地向后 6小时每个观测转移。这会将2017-06-17 06:00:00转变为2017-06-17 00:00:00，即现在上午6点成为该实际日子的开始。

select 
    cast(dateadd(hour, -6, obsDate) as date) as theDate, 
    sum(case when currentScore < 8 then 1 else 0 end) as currentscore_low, 
    sum(case when currentScore >= 8 and currentScore < 17 
      then 1 else 0 end) as currentscore_medium, 
    sum(case when currentScore >= 17 then 1 else 0 end) as currentscore_high 
from diseaseScores 
where siteID = 8315 and 
     obsDate >= cast(getdate() - 2 as date) 
group by cast(dateadd(hour, -6, obsDate) as date) 
order by cast(dateadd(hour, -6, obsDate) as date);