SQL Server 获取构成聚合的第一行和最后一行

Posted 2023-03-25

【中文标题】SQL Server 获取构成聚合的第一行和最后一行

【英文标题】：SQL Server Getting First & Last Row that makes up aggregation

【发布时间】：2011-11-22 02:39:20

【问题描述】：

我有一个存储一分钟内价格变动数据的表格。每条记录包含最后一分钟的开盘价、最高价、最低价、收盘价和成交量

CREATE TABLE TimeBar (
  Instrument varchar(20),
  BarTimeStamp datetimeoffset(7),
  Open decimal(18, 5),
  High decimal(18, 5),
  Low decimal(18, 5),
  Close decimal(18, 5),
  Volume int
)

我要做的是创建一个查询，我可以在其中将数据聚合到更高的时间范围内，例如，我希望能够显示每个小时的开盘价、最高价、最低价和收盘价。

以下是我目前的查询，我已经设法获得了最高价和最低价，但是您如何获得开盘价和收盘价？

  SELECT MIN(BarTimeStamp) AS TimeStamp,
         MAX(High) AS High,
         MIN(Low) AS Low,
         SUM(Volume) AS Volume
    FROM TimeBar
   WHERE Instrument = 'XYZ'
GROUP BY DATEPART(YEAR, BarTimeStamp), DATEPART(MONTH, BarTimeStamp), DATEPART(DAY, BarTimeStamp), DATEPART(HOUR, BarTimeStamp)

【参考方案1】：

将主查询用作子查询，为每条记录获取最小时间戳和最大时间戳，因此为它们获取各自的开盘价和收盘价。

SELECT x.*, TI.Open, TE.Close
FROM
(
SELECT   Instrument,
         MIN(BarTimeStamp) AS TimeStampIni,
         MAX(BarTimeStamp) AS TimeStampEnd,
         MAX(High) AS High,
         MIN(Low) AS Low,
         SUM(Volume) AS Volume
    FROM TimeBar
GROUP BY Instrument, DATEPART(YEAR, BarTimeStamp), DATEPART(MONTH, BarTimeStamp), DATEPART(DAY, BarTimeStamp), DATEPART(HOUR, BarTimeStamp)
) x
inner join TimeBar TI on ti.Instrument = x.Instrument AND TI.BarTimeStamp = x.TimeStampIni
inner join TimeBar Te on te.Instrument = x.Instrument AND Te.BarTimeStamp = x.TimeStampEnd

【讨论】：

谢谢，正是我想要的。

【参考方案2】：

首先要做的是按时间范围分组：

SELECT MIN(BarTimeStamp) AS StartTimeStamp,
         MAX(BarTimeStamp) AS EndTimeStamp
    FROM #TimeBar
   WHERE Instrument = 'TEST'
GROUP BY DATEPART(YEAR, BarTimeStamp), DATEPART(MONTH, BarTimeStamp), DATEPART(DAY,       BarTimeStamp), DATEPART(HOUR, BarTimeStamp)

之后，问题是：

进行连接以获取范围， 进行连接以开始 进行连接以获得最小值 按原始范围分组以选择所有内容

我喜欢使用公用表表达式来简化：

;WITH times as (
    SELECT MIN(BarTimeStamp) AS StartTimeStamp,
             MAX(BarTimeStamp) AS EndTimeStamp
        FROM #TimeBar
       WHERE Instrument = 'TEST'
    GROUP BY DATEPART(YEAR, BarTimeStamp), DATEPART(MONTH, BarTimeStamp), DATEPART(DAY, BarTimeStamp), DATEPART(HOUR, BarTimeStamp)
)
SELECT 
    StartTimeStamp as TimeStamp
    ,MIN([first].[Open]) as [Open]
    ,MAX(ranged.High) as High
    ,MAX(ranged.Low) as Low
    ,MIN([last].[Close]) as [Close]
    ,SUM(ranged.Volume) as Volume
FROM times 
INNER JOIN #TimeBar ranged ON times.StartTimeStamp <= ranged.BarTimeStamp AND times.EndTimeStamp >= ranged.BarTimeStamp
INNER JOIN #TimeBar [first] ON times.StartTimeStamp = [first].BarTimeStamp 
INNER JOIN #TimeBar [last] ON times.EndTimeStamp = [last].BarTimeStamp 
GROUP BY [times].StartTimeStamp

这是我的测试数据：

CREATE TABLE #TimeBar (
  Instrument varchar(20),
  BarTimeStamp datetimeoffset(7),
  [Open] decimal(18, 5),
  High decimal(18, 5),
  Low decimal(18, 5),
  [Close] decimal(18, 5),
  Volume int
)

insert into #TimeBar values ('TEST', '2011-11-21 1:00', 5, 6, 4, 8, 100)
insert into #TimeBar values ('TEST', '2011-11-21 1:10', 1, 7, 3, 4, 100)
insert into #TimeBar values ('TEST', '2011-11-21 2:10', 15, 16, 17, 18, 100)
insert into #TimeBar values ('TEST', '2011-11-21 2:30', 12, 16, 17, 19, 100)
insert into #TimeBar values ('TEST', '2011-11-21 2:50', 13, 14, 15, 20, 100)

结果是：

TimeStamp                          Open                                    High                                    Low                                     Close                                   Volume
---------------------------------- --------------------------------------- --------------------------------------- --------------------------------------- --------------------------------------- -----------
2011-11-21 01:00:00.0000000 +00:00 5.00000                                 7.00000                                 4.00000                                 4.00000                                 200
2011-11-21 02:10:00.0000000 +00:00 15.00000                                16.00000                                17.00000                                20.00000                                300

注意，我为此使用了一个临时表，只需将 #TimeBar 更改为 TimeBar 即可为您的真实表更改它。

---

除此之外，我不喜欢将Instrurment varchar 视为字段定义。您应该为此使用surrogate key。