Sql server 如何计算两个datetime2列之间的间隔（SQL Server）？

嗨，我正在计算datetime2类型的两列之间的差异

但是，SQL server（2012）似乎不喜欢以下内容：

select cast ('2001-01-05 12:35:15.56786' as datetime2)
    - cast ('2001-01-01 23:45:21.12347' as datetime2);

Msg 8117, Level 16, State 1, Line 2
Operand data type datetime2 is invalid for subtract operator.

现在，如果我将其转换为datetime类型，它就可以工作了：

select cast (cast ('2001-01-05 12:35:15.56786' as datetime2) as datetime) 
    - cast (cast ('2001-01-01 23:45:21.12348' as datetime2) as datetime);

1900-01-04 12:49:54.443

---

但是，当我将其转换为datetime时，我正在失去精度（请注意上面的3位小数精度）。在这种情况下，我实际上需要所有5个小数点。有没有办法获得两个datetime2列之间的间隔，并且仍然保持5个小数点的精度？谢谢。

要查找两个日期之间的差异，您需要使用该函数

你可以简单地使用

返回指定日期部分的计数（有符号整数） 在指定的开始日期和结束日期之间跨越边界

不幸的是，试图通过以下方式获得所需的精度：

select DATEDIFF(MICROSECOND, cast('20010101 23:45:21.12347' as datetime2), 
                             cast('20010105 12:35:15.56786' as datetime2)) 

导致溢出错误：

The datediff function resulted in an overflow. 
The number of dateparts separating two date/time instances is too large. 
Try to use datediff with a less precise datepart.

实现所需精度的一种方法是迭代地分解成粒度时间组件（天、小时、分钟、秒等），并使用DateAdd（）将其从值中减去，例如

---

5年后，这不太可能对slavoo有所帮助，但这个冗长的例子可能满足了人们的要求：

    @FromDateTime DATETIME2 = CAST('20010101 23:45:21.12347' AS DATETIME2),
    @ToDateTime DATETIME2 = CAST('20010105 12:35:15.56786' AS DATETIME2),
    @FromMicroSecs NUMERIC,
    @FromDateTimeNoMicroSecs DATETIME,
    @ToMicroSecs NUMERIC,
    @ToDateTimeNoMicroSecs DATETIME;

SELECT
    @FromMicroSecs = DATEPART(MICROSECOND, @FromDateTime),
    @FromDateTimeNoMicroSecs = CAST(DATEADD(MICROSECOND, -1 * @FromMicroSecs, @FromDateTime) AS DATETIME),
    @ToMicroSecs = DATEPART(MICROSECOND, @ToDateTime),
    @ToDateTimeNoMicroSecs = CAST(DATEADD(MICROSECOND, -1 * @ToMicroSecs, @ToDateTime) AS DATETIME);

SELECT
    CAST(DATEDIFF(SECOND, @FromDateTimeNoMicroSecs, @ToDateTimeNoMicroSecs) AS NUMERIC) * 1000000 + @ToMicroSecs - @FromMicroSecs AS AnswerInMicroseconds;```

---

这是为了计算两个科学指标之间的间隔。存储间隔值，然后将其添加到另一个datetime2以进行预测。虽然两个日期之间的间隔通常没有那么大（上例中为4天）。也许这更现实一点。选择DATEDIFF（微秒，cast（'20010101 23:45:21.12347'作为datetime2），cast（'20010102 00:05:15.56786'作为datetime2）），它似乎在足够小的时间间隔内工作。也许要注意，这种精度有一个微秒的上限？如果只有SQL server datetime2在这方面的表现与datetime类似，那么我就不必经历这样的麻烦。

The datediff function resulted in an overflow. 
The number of dateparts separating two date/time instances is too large. 
Try to use datediff with a less precise datepart.

remainingAtLowerGranularity = DateAdd(granularity, -1 * numFoundInStep, value)

    @FromDateTime DATETIME2 = CAST('20010101 23:45:21.12347' AS DATETIME2),
    @ToDateTime DATETIME2 = CAST('20010105 12:35:15.56786' AS DATETIME2),
    @FromMicroSecs NUMERIC,
    @FromDateTimeNoMicroSecs DATETIME,
    @ToMicroSecs NUMERIC,
    @ToDateTimeNoMicroSecs DATETIME;

SELECT
    @FromMicroSecs = DATEPART(MICROSECOND, @FromDateTime),
    @FromDateTimeNoMicroSecs = CAST(DATEADD(MICROSECOND, -1 * @FromMicroSecs, @FromDateTime) AS DATETIME),
    @ToMicroSecs = DATEPART(MICROSECOND, @ToDateTime),
    @ToDateTimeNoMicroSecs = CAST(DATEADD(MICROSECOND, -1 * @ToMicroSecs, @ToDateTime) AS DATETIME);

SELECT
    CAST(DATEDIFF(SECOND, @FromDateTimeNoMicroSecs, @ToDateTimeNoMicroSecs) AS NUMERIC) * 1000000 + @ToMicroSecs - @FromMicroSecs AS AnswerInMicroseconds;```