DATEDIFF 函数导致 SQL Server 2016 溢出

Question

我在 SQL Server 2016 中有一个 table，其中包含以下数据：

id     | t                       | memory
-------+-------------------------+------------
620255 | 2017-07-17 16:11:25.100 | 11632640
620127 | 2017-07-17 16:11:24.100 | 11632640
619999 | 2017-07-17 16:11:23.097 | 11632640
619872 | 2017-07-17 16:11:22.097 | 11632640

我正在尝试获取每秒 memory 的平均值。因此，为了做到这一点，我尝试了以下查询：

SELECT 
    AVG(memory) AS avgMemory, 
    DATEADD(second, DATEDIFF(second, 0, t), 0) AS t 
FROM 
    Table1 AS Table1 
WHERE 
    t BETWEEN '2017-07-17 16:11:00.000' AND '2017-07-17 16:12:00.000' 
GROUP BY 
    DATEADD(second, DATEDIFF(second, 0, t), 0) 
ORDER BY 
    t ASC;

但是在执行时，出现以下错误：

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.

根据 official docs and this SO post，秒数的最大差异可以是 68 年，但在我的例子中只有 60 秒。

那么为什么我会收到此错误？我能做些什么来解决它？

这是 SQL解决上述问题的方法：Fiddle

Answer 1

SQL 服务器为此目的提供 datediff_big()：

SELECT AVG(memory) AS avgMemory,
       DATEADD(second, DATEDIFF_BIG(second, 0, t), 0) AS t 
FROM Table1 AS Table1 
WHERE t BETWEEN '2017-07-17 16:11:00.000' AND '2017-07-17 16:12:00.000' 
GROUP BY DATEADD(second, DATEDIFF_BIG(second, 0, t), 0) 
ORDER BY t ASC;

您也可以通过将值转换为字符串来执行此操作：

SELECT AVG(memory) AS avgMemory,
       CONVERT(VARCHAR(19), t, 120) as t
FROM Table1 AS Table1 
WHERE t BETWEEN '2017-07-17 16:11:00.000' AND '2017-07-17 16:12:00.000' 
GROUP BY CONVERT(VARCHAR(19), t, 120)
ORDER BY t ASC;

此版本适用于 SQL 服务器的所有受支持版本。

Answer 2

您也可以使用不同的日期（例如 2017-07-17 00:00:00.000）作为计算的固定点：

SELECT AVG(memory) AS memory, 
  DATEADD(second, 
          DATEDIFF(second, '2017-07-17 00:00:00.000', t),
          '2017-07-17 00:00:00.000') AS t 
FROM Table1 AS Table1 
WHERE t BETWEEN '2017-07-17 16:11:00.000' AND '2017-07-17 16:12:00.000' 
GROUP BY DATEADD(second, 
                 DATEDIFF(second, '2017-07-17 00:00:00.000', t), 
                 '2017-07-17 00:00:00.000') 
ORDER BY t ASC;

SQL Fiddle

Answer 3

这里的问题是"starting date"。

在select DATEDIFF(second, 0, t)中，0表示1900-01-01离2017年太远了，所以从1900-01-01到2017-07-17的时间以秒为单位溢出了简单的整数，这不仅仅是“60 秒”

Answer 4

pre 2016 db，在一组依据上，我不得不将 datediff 转换为 bigint，然后获取平均值，对于 datediff by ms。 avg(cast(datediff(ms,sdate,edate)) as BIGINT)) as theavg

DATEDIFF 函数导致 SQL Server 2016 溢出

DATEDIFF function resulted in an overflow in SQL Server 2016

sql

sql-server

sql-server-2016