更新查询与最近的DateTime匹配与内部联接

Question

我有以下两个表

CREATE TABLE Ep
    ([E] varchar(9), [M] varchar(9), [DTE] DATETIME)
;

INSERT INTO Ep
    ([E], [M], [DTE])
VALUES
    ('1595861-1', '1595861-1', CONVERT(datetime, '2002-11-26 14:18:00', 20)),
    ('1595904-1', '1595904-1', CONVERT(datetime, '2002-11-24 15:15:00', 20)),
    ('1596298-1', '1596298-1', CONVERT(datetime, '2002-12-17 11:12:00', 20)),
    ('1596357-1', '1596357-1', CONVERT(datetime, '2002-12-09 19:57:00', 20)),
    ('1596369-1', '1596369-1', CONVERT(datetime, '2002-12-11 06:00:00', 20)),
    ('1596370-1', '1596370-1', CONVERT(datetime, '2002-12-19 12:31:00', 20)),
    ('1596473-2', '1596473-1', CONVERT(datetime, '2002-12-15 08:39:00', 20)),
    ('1596473-3', '1596473-1', CONVERT(datetime, '2002-12-20 08:39:00', 20)),
    ('1596473-4', '1596473-1', CONVERT(datetime, '2002-12-13 08:39:00', 20)),
    ('1596473-5', '1596473-1', CONVERT(datetime, '2002-12-16 08:39:00', 20)),
    ('1596473-1', '1596473-1', CONVERT(datetime, '2002-12-14 08:39:00', 20))
;

CREATE TABLE Mp
    ([E] varchar(9), [M] varchar(9), [DTE] DATETIME)
;

INSERT INTO Mp
    ([E], [M], [DTE])
VALUES
    ('', '1595861-1', CONVERT(datetime, '2002-11-26 14:18:00', 20)),
    ('', '1595904-1', CONVERT(datetime, '2002-11-24 15:15:00', 20)),
    ('', '1596298-1', CONVERT(datetime, '2002-12-17 11:12:00', 20)),
    ('', '1596357-1', CONVERT(datetime, '2002-12-09 19:57:00', 20)),
    ('', '1596369-1', CONVERT(datetime, '2002-12-11 06:00:00', 20)),
    ('', '1596370-1', CONVERT(datetime, '2002-12-19 12:31:00', 20)),
    ('', '1596473-1', CONVERT(datetime, '2002-12-17 08:39:00', 20))
;

目前，我正在通过[E]上的匹配更新Mp表格中的[M]字段，其中DTE字段（位于Mp）位于certian范围（比如+ -3天）。目前执行此操作的查询是

UPDATE [Mp] 
SET [E] = [Ep].[E] 
FROM [Mp] INNER JOIN [Ep] 
    ON [Mp].[M] = [Ep].[M] 
WHERE [Mp].[DTE] BETWEEN [Ep].[DTE] - 3 AND [Ep].[DTE] + 3;

这会将[Mp].[E]的{​​{1}}更新为[Mp].[M] = N'1596473-1'。 Essentailly SQL Server发现的第一个有效的条目。但是，我想更新此查询，以便SQL Server在所需日期范围内的1596473-2字段上匹配（就像现在一样），但是对于最接近[M]值的[Ep].[DTE]值。 {1}} [Mp].[DTE]的值。

我已经考虑过以下方式添加2002-12-17 08:39:00子句

DATEDIFF

显然，我无法做到这一点，但我不确定如何对此进行修改以使其有效。更新后[Mp]的最终数据应为

UPDATE [Mp] 
SET [E] = [Ep].[E] 
FROM [Mp] INNER JOIN [Ep] 
    ON [Mp].[M] = [Ep].[M] 
WHERE [Mp].[DTE] BETWEEN [Ep].[DTE] - 3 AND [Ep].[DTE] + 3 
ORDER BY DATEDIFF(minutes, [Mp].[DTE], [Ep].[DTE]);

感谢您的时间。

Answer 1

请先检查CTE产生的数据作为输出

对于最近的数据，我根据DATEDIFF（）函数提取的时差计算使用SQL ROW_NUMBER function和Partition By子句。为了消除以前和以后的记录，我使用了ABS（）数学函数。

select
*,
ROW_NUMBER() over (partition by [Mp].[M] order by abs(datediff(mi, [Mp].[DTE], [Ep].[DTE]))) as rn,
abs(datediff(mi, [Mp].[DTE], [Ep].[DTE])) diff
from Mp
left join Ep 
    on [Mp].[M] = [Ep].[M] 
WHERE [Mp].[DTE] BETWEEN dateadd(dd,-3,[Ep].[DTE]) AND dateadd(dd,3,[Ep].[DTE])

然后在UPDATE命令中使用相同的CTE表达式，如下所示，可以将所需数据填充到目标数据库表中

;with cte as (
    select
        [Mp].[M] as M,
        [Ep].E as E,
    ROW_NUMBER() over (partition by [Mp].[M] order by abs(datediff(mi, [Mp].[DTE], [Ep].[DTE]))) as rn,
    abs(datediff(mi, [Mp].[DTE], [Ep].[DTE])) diff
    from Mp
    left join Ep 
        on [Mp].[M] = [Ep].[M] 
    WHERE [Mp].[DTE] BETWEEN dateadd(dd,-3,[Ep].[DTE]) AND dateadd(dd,3,[Ep].[DTE])
)
update [Mp]
set [E] = cte.E
from [Mp]
inner join cte on [Mp].M = cte.M and cte.rn = 1
where 
    cte.E is not null

执行UPDATE语句后，目标表数据如下

我希望这是你需要的