从c＃执行相同查询时的结果不同（不正确）

Question

我正在尝试在SQL Server中进行一些复杂的操作。为了防止任何性能损失，我试图在数据库端进行所有操作。

以下是我在C＃中的查询

"DECLARE @t AS TABLE (StartTime DATETIME, EndTime DateTime, Value REAL, V1 REAL, V2 REAL) " +

"DECLARE @i AS DateTime = @startDate " +
"WHILE @i <= @endDate " +
"BEGIN " +

    "DECLARE @min AS Real = " +
        "(SELECT TOP 1 RawStatus FROM CustomPollerStatistics_Detail " +
        "WHERE RawStatus IS NOT NULL AND CustomPollerAssignmentID = @pollerID " +
        "AND [DateTime] <= @i ORDER BY [DateTime] DESC) " +

    "DECLARE @max AS Real = " +
        "(SELECT TOP 1 RawStatus FROM CustomPollerStatistics_Detail " +
        "WHERE RawStatus IS NOT NULL AND CustomPollerAssignmentID = @pollerID " +
        "AND [DateTime] <= DATEADD(MI, @interval, @i) ORDER BY [DateTime] DESC) " +

    "INSERT INTO @t VALUES (@i, DATEADD(MI, @interval, @i), @max - @min, @min, @max) " +

    "SET @i = DATEADD(MI, @interval, @i) " +
"END " +

"SELECT * FROM @t";

我在调试模式下仔细检查了参数的值。我尝试使用相同的参数运行查询，方法是在SQL Server Management Studio中使用C＃传递的相同值在查询顶部声明它，在这种情况下，它会按预期生成正确的结果。我已按以下方式添加参数。

com.Parameters.Add("@pollerID", System.Data.SqlDbType.UniqueIdentifier).Value = assignmentID;
com.Parameters.Add("@interval", System.Data.SqlDbType.Int).Value = accuracyMinutes;
com.Parameters.Add("@startDate", System.Data.SqlDbType.DateTime).Value = startDate;
com.Parameters.Add("@endDate", System.Data.SqlDbType.DateTime).Value = endDate;

当我从C＃运行此查询时，列@max - @min, @min, @max（第3列，第4列和第5列）总是评估为0。我无法自己找出原因。请帮忙。见下面的截图。

示例数据https://github.com/scipy/scipy/pull/3574。 C＃代码http://pastebin.com/raw.php?i=QuBCSRw3。

Answer 1

首先，理想情况下，数据库中应该有两个带有索引的标准表。我看到有60-70％的报告使用这些表格。

1. 表示日期，即1900-01-01至2079-06-06（smalldatetime）最大值 范围
2. 号码表（1到1百万）

这些将帮助您解决报告中的性能问题。

如果您不能，请先使用以下脚本生成日期时间范围。这将运行n次迭代，即2 ^ n

DROP TABLE #tbl;
CREATE TABLE #tbl( num INT,
                   startDt DATETIME2( 0 ),
                   endDt DATETIME2( 0 ),
                   dateWithMin AS DATEADD( mi,num,startDt ) PERSISTED
                                                            PRIMARY KEY
--will create clustered index for performance
);
DECLARE @startDt DATETIME2( 0 ) = GETDATE(),
        @endDt DATETIME2( 0 ) = DATEADD( Mi,15,GETDATE());
DECLARE @rc INT = 1,
        @max INT = DATEDIFF( MI,@startDt,@endDt );
SELECT @rc,
       @max;
INSERT INTO #tbl
VALUES( 0,@startDt,@endDt );
WHILE @rc * 2 <= @max
    BEGIN
        INSERT INTO #tbl
        SELECT num + @rc,
               @startDt,
               @endDt
          FROM #tbl;
        SET @rc = @rc * 2;
    END;
INSERT INTO #tbl
SELECT num + @rc,
       @startDt,
       @endDt
  FROM #tbl
  WHERE num + @rc <= @max;
SELECT *
  FROM #tbl;

这会产生类似这样的结果

+-----+---------------------+---------------------+---------------------+
| num |       startDt       |        endDt        |     dateWithMin     |
+-----+---------------------+---------------------+---------------------+
|   0 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:37:24 |
|   1 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:38:24 |
|   2 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:39:24 |
|   3 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:40:24 |
|   4 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:41:24 |
|   5 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:42:24 |
|   6 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:43:24 |
|   7 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:44:24 |
|   8 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:45:24 |
|   9 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:46:24 |
|  10 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:47:24 |
|  11 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:48:24 |
|  12 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:49:24 |
|  13 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:50:24 |
|  14 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:51:24 |
|  15 | 2015-05-22 19:37:24 | 2015-05-22 19:52:24 | 2015-05-22 19:52:24 |
+-----+---------------------+---------------------+---------------------+

假设您每分钟都有数据，以下查询应该会给您结果。否则你需要一个如上所述的时间序列表。

修改

SELECT dt,
       mini,
       maxi,
       ( 
         SELECT RawStatus
           FROM CustomPollerStatistics_Detail
           WHERE [DateTime] = maxi ) - ( 
                                         SELECT RawStatus
                                           FROM CustomPollerStatistics_Detail
                                           WHERE [DateTime] = mini ) AS PowerConsumed 
INTO #tmp
  FROM( 
        SELECT DATEADD( Mi,DATEDIFF( Mi,0,[DateTime] ),0 ) AS dt,
               MIN( [DateTime] ) mini,
               MAX( [DateTime] ) maxi
          FROM CustomPollerStatistics_Detail t
          WHERE [DateTime] BETWEEN @StartDate AND @EndDate
          GROUP BY DATEADD( Mi,DATEDIFF( Mi,0,[DateTime] ),0 )) x;
SELECT l.dt,
       DATEADD( Mi,@TimeInterval,l.dt ) AS NextDt,
       SUM( PowerConsumed ) AS Total
  FROM #tmp l
  WHERE l.dt BETWEEN l.dt AND( 
                               SELECT TOP 1 r.maxi AS NextDt
                                 FROM #tmp r
                                 WHERE r.dt = l.dt
                                 ORDER BY r.maxi DESC )
  GROUP BY l.dt,
           DATEADD( Mi,@TimeInterval,l.dt );