Question

我正在使用RODBC在R中进行SQL查询。

查询以格式

提供数据

Date(POSIXct); var1:var29
%Y-%m-%d %H:%M:%S; numeric_values

问题是，一旦查询超过一定长度，6960左右有29-30个变量，包括日期，传递给R的信息开始如下：

Date(POSIXct); var1:var30
%Y-%m-%d; numeric_values

因此，我丢失了"%H:%M:%S"信息。我不知道为什么。如果我减少变量的数量，我可以在此之前增加时间长度。

在Windows服务器2007上是ts-sql。（我相信）

R中的SQL调用示例：

sqlQuery(database, "SELECT [datetime], [0] as SYS, [1] as NO1, [2] as NO2, [7] as NO3, [9] as NO4, [19] as NO5, [5] as DK1,[6] as DK2, [25] as SE1,
         [26] as SE2,[27] as SE3, [28] as SE4, [4] as FIN, [13] as DE, [14] as NL, [16] as FR, [15] as CH, [17] as AT, [20] as EE,
         [36] as LT, [45] as LV, [42] as SI, [50] as IT, [44] as ES, [43] as BE, [74] as HU, [75] as CZ, [41] as UK

         From
         (
         SELECT [area_id],[pris],[datetime]
         FROM [BigData].[dbo].[Prices]
         WHERE area_id in (0,1,2,7,9,19,5,6,25,26,27,28,4,13,14,16,15,17, 20, 36, 45, 42, 50, 44, 43, 74, 75, 41)
         AND [datetime]>= cast(GETDATE()-290 as date) 
         AND [datetime]< cast(GETDATE()+0 as date)
         ) p
         PIVOT(SUM([pris])
         FOR [area_id] IN
         ([0], [1], [2], [7], [9], [19], [5],[6], [25],[26],[27], [28], [4], [13], [14], [16], [15], [17], [20],
         [36], [45], [42], [50], [44], [43], [74], [75], [41]))
         AS pvt
         ORDER BY [datetime] asc   ") -> prices

Answer 1

解决方案＃1

您可以使用answer from kristang :(使用as.is选项调用sqlQuery，在字符串中获取时间戳，并在R中使用as.POSIXct转换列。

解决方案＃2

但我认为更有效的解决方案是通过SQL表达式（SQL Server的示例）获取数值类型的日期时间值：

sqlQuery( "select convert(float, my_date)*3600*24 as my_date from ...")

并将其从数字转换为POSIXct：

df1$my_date <- as.POSIXct(df1$my_date, origin = "1900-01-01", tz = "UTC")

由于POSIXct本质上是数字，因此获取和类型转换比使用RODBC的普通sqlQuery运行得更快。 RODBC从文本字符串转换每个时间戳（在sqlGetResults函数中看起来为.POSIXct用法）。因此，即使RODBC返回完整的时间戳，解决方案也是合理的。

P.S。如果您真的想从文本转换，请参阅fasttime包中的fastPOSIXct。

SQL查询丢失信息

1 个答案:

解决方案＃1

解决方案＃2