Question

我有一个SQL服务器数据库，其中包含一个包含100个测量点的测量数据的表。以大约5秒的间隔测量每一个。它具有以下格式：

MeasurementPointID           timestamp              value
001234                       03-01-2015 00:02:03    100
001234                       03-01-2015 00:02:08    120
001234                       03-01-2015 00:02:13    130
001234                       03-01-2015 00:02:19    160
001234                       03-01-2015 00:02:22    200
001236                       03-01-2015 00:02:04    400
001236                       03-01-2015 00:02:09    405
001236                       03-01-2015 00:02:14    420
001236                       03-01-2015 00:02:19    445
001236                       03-01-2015 00:02:25    470
Etc.

我想知道如何根据以前的时间戳除以时间段（自前时间戳）以来测量值的变化来制作视图

MeasurementPointID           timestamp              changeDivPeriod
001234                       03-01-2015 00:02:03    0 <-- Zero would be nice here
001234                       03-01-2015 00:02:08    4
001234                       03-01-2015 00:02:13    2
001234                       03-01-2015 00:02:19    5
001234                       03-01-2015 00:02:22    13,333333
001236                       03-01-2015 00:02:04    0 <-- Zero would be nice here
001236                       03-01-2015 00:02:09    1
001236                       03-01-2015 00:02:14    3
001236                       03-01-2015 00:02:19    5
001236                       03-01-2015 00:02:25    4,166666
Etc.

我将如何做到这一点？首先用一种观点解决这个问题是否明智？

我认为将触发器添加到此表并使用计算值填充新表（而不是视图）会更快/更容易。如果是这样，我该怎么做？

Answer 1

你可以使用窗口函数和自连接（如果需要用视图包装它）：

WITH cte AS
(
  SELECT *,
    rn = ROW_NUMBER() OVER (PARTITION BY MeasurementPointID ORDER BY [timestamp])
  FROM #tab
)
SELECT c1.MeasurementPointID,
       c1.[timestamp],
       [changeDivPeriod] = 
           CASE WHEN c1.rn = 1 THEN 0 
                ELSE 1.0 * (c1.[value] - c2.[value]) 
                     / NULLIF(DATEDIFF(second, c2.[timestamp], c1.[timestamp]),0)
           END
FROM cte c1
LEFT JOIN cte c2
  ON c1.MeasurementPointID = c2.MeasurementPointID 
  AND c1.rn = c2.rn+1;

的 LiveDemo

输出：

╔════════════════════╦═════════════════════╦═════════════════╗ ║ MeasurementPointID ║ timestamp ║ changeDivPeriod ║ ╠════════════════════╬═════════════════════╬═════════════════╣ ║ 001234 ║ 2015-03-01 00:02:03 ║ 0 ║ ║ 001234 ║ 2015-03-01 00:02:08 ║ 4 ║ ║ 001234 ║ 2015-03-01 00:02:13 ║ 2 ║ ║ 001234 ║ 2015-03-01 00:02:19 ║ 5 ║ ║ 001234 ║ 2015-03-01 00:02:22 ║ 13.333333333333 ║ ║ 001236 ║ 2015-03-01 00:02:04 ║ 0 ║ ║ 001236 ║ 2015-03-01 00:02:09 ║ 1 ║ ║ 001236 ║ 2015-03-01 00:02:14 ║ 3 ║ ║ 001236 ║ 2015-03-01 00:02:19 ║ 5 ║ ║ 001236 ║ 2015-03-01 00:02:25 ║ 4.166666666666 ║ ╚════════════════════╩═════════════════════╩═════════════════╝

使用SQL Server 2012+，您可以使用 LEAD/LAG ：

SELECT MeasurementPointID, [timestamp], [changeDivPeriod] = COALESCE(1.0 * ([value] - LAG ([value]) OVER ( PARTITION BY MeasurementPointID ORDER BY [timestamp] )) / NULLIF(DATEDIFF(second, LAG ([timestamp]) OVER ( PARTITION BY MeasurementPointID ORDER BY [timestamp]), [timestamp]),0),0) FROM #tab

的 LiveDemo2

SQL Server之间的差异很多不规则的测量值

1 个答案: