Question

我编写了这个T-SQL脚本，用于汇总由重建过程创建的数据库中的重复行。为此，它执行以下操作：

获取并存储表变量（@minColTargetPKTable）中每组重复条目的最小目标表主键（ColTargetPK）;通过匹配三列来确定重复项 - ColIntA，ColIntB和ColDateTimeA。
将要汇总的目标列（TargetColVarchar）设置为等于相应重复条目的目标列的串联。
将重复的条目标记为非活动状态（ColTargetStatus = 0）
报告成功（或失败）

由于涉及的数据集的大小，此脚本需要一段不合适的时间才能运行。

任何人都可以看到如何将其转换为基于设置的，如果是这样，你能提供一个例子吗？

如果我的描述有点令人困惑，我提前道歉......

declare @MinColTargetPKTable table
    (ColIntA int,
     ColIntB int,
     ColDateTimeA nvarchar(25),
     minColTargetPK int
    )

insert @minColTargetPKtable
    select ColIntA, ColIntB, convert(nvarchar(25),ColDateTimeA,120) as ColDateTimeA, 
    min(ColTargetPK) as MinColTargetPK from TargetColTable
    group by ColIntA, ColIntB, convert(nvarchar(25),ColDateTimeA,120) 


declare @TargetColVarchar varchar(max)

declare @updatedColTargetPKs table
(updatedColTargetPKs int)

declare @minColTargetPK int

declare cur cursor
for
select minColTargetPK
from @minColTargetPKtable

open cur

fetch next from cur into @minColTargetPK

while @@FETCH_STATUS = 0
begin
    begin try

    set @TargetColVarchar =
        convert(nvarchar(max),(
        select replace(convert(nvarchar(max), isnull(TargetColVarchar,'')) +   convert (nvarchar(max),' \par \par \par'), '\par } ', '\par') as 
        TargetColVarchar
        from TargetColTable v1
        where ColIntA = (select ColIntA from TargetColTable where ColTargetPK = @minColTargetPK)
        and ColIntB = (select ColIntB from TargetColTable where ColTargetPK = @minColTargetPK)
        and convert(nvarchar(25),ColDateTimeA,120) = (select convert(nvarchar(25),ColDateTimeA,120) from TargetColTable where ColTargetPK = @minColTargetPK)
        order by ColTargetPK
        for xml path(''), type  
        ))

        set @TargetColVarchar = REPLACE(REPLACE (REPLACE (@TargetColVarchar,'<TargetColVarchar>',''),'</TargetColVarchar>',''), '&#x0D;','')


        update TargetColTable
        set TargetColVarchar = @TargetColVarchar
        where ColTargetPK = @minColTargetPK

        update TargetColTable
        set ColTargetStatus = 0
        from TargetColTable v1
        where ColIntA = (select ColIntA from TargetColTable where ColTargetPK = @minColTargetPK)
        and ColIntB = (select ColIntB from TargetColTable where ColTargetPK = @minColTargetPK)
        and convert(nvarchar(25),ColDateTimeA,120) = (select convert(nvarchar(25),ColDateTimeA,120) from TargetColTable where ColTargetPK = @minColTargetPK)
        and ColTargetPK != @minColTargetPK

        Print 'Merge complete for ColTargetPK '+ convert(varchar(50),  @minColTargetPK)

    end try
    begin catch

        Print 'Merge failed for ColTargetPK '+ convert (varchar(20),@minColTargetPK)

    end catch

    fetch next from cur into @minColTargetPK
end

close cur
deallocate cur

编辑：好的，下面是使用Preet的建议将脚本移动到基于集合的操作。为了给出一些额外的背景，TargetTable约为110万行。奇怪的是，在2次试验中，下面基于集合的脚本并不比下面基于光标的脚本在相同的数据子集（大约20000行）上快得多。对于为什么这不会更快的任何想法？

declare @minColTargetPKTable table
    (
    ColIntA int,
    ColIntB int,
    ColDateTimeA nvarchar(25),
    ColTargetPK int,
    concTargetCol varchar(max)
    )

insert @minColTargetPKtable (minColIntA,ColIntB,minColDateTimeA,minColTargetPK)
select ColIntA, ColIntB, convert(nvarchar(25),ColDateTimeA,120) as ColDateTimeA, min(ColTargetPK) as minColTargetPK from TargetTable
group by ColIntA, ColIntB, convert(nvarchar(25),ColDateTimeA,120) 


update @minColTargetPKTable 
set concTargetCol  = 
(REPLACE(REPLACE(REPLACE(replace(convert(nvarchar(max),
    (
        select convert(nvarchar(max), isnull(TargetColVarchar,'')) + convert (nvarchar(max),' \par \par \par ') as 
        TargetColVarchar
        from TargetTable v1
        where ColIntA = (select ColIntA from TargetTable where ColTargetPK = minColTargetPK)
        and ColIntB = (select ColIntB from TargetTable where ColTargetPK = minColTargetPK)
        and convert(nvarchar(25),ColDateTimeA,120) = (select convert(nvarchar(25),ColDateTimeA,120) from TargetTable where ColTargetPK = minColTargetPK)
        order by ColTargetPK
        for xml path(''), type  
    ))
, '\par } ', '\par '),'<TargetColVarchar>',''),'</TargetColVarchar>',''), '&#x0D;',''))

update TargetTable 
set TargetColVarchar = mv.concTargetCol
from @minColTargetPKTable mv
where mv.minColTargetPK = TargetTable.ColTargetPK 

update TargetTable 
set TargetColStatus = 0 
from TargetTable v
inner join @minColTargetPKTable mv on
mv.minColIntA = v.ColIntA
and mv.minColDateTimeA = convert(nvarchar(25),v.ColDateTimeA,120)
and mv.ColIntB = v.ColIntB 
where ColTargetPK not in (select minColTargetPK from @minColTargetPKTable)

Answer 1

好的，我建议以下内容：

在临时表中添加一个额外的列到@TargetColVarchar值，执行此操作
将临时表和TargetColTable加入到执行更新

然后，您可以根据执行计划进行优化

更新

查看修改后的结果，我会说顺序如下：

使用#temp表，这些表在大型数据集上更具竞争力。

一个。在临时表中添加更多列，以便在大前锋中记录诸如(select ColIntA from TargetColTable where ColTargetPK = @minColTargetPK和(select ColIntB from TargetColTable where ColTargetPK = @minColTargetPK之类的内容

湾我认为字符串替换很慢。这仍然很慢。我知道XML不是世界上最快的东西。你能用SQL XML specfic代码替换字符串comp

℃。在底部的第二次更新中，where ColTargetPK not in (select minColTargetPK from @minColTargetPKTable)可能比精确连接慢，并且您应该同时执行两次更新

但是，请使用“实际查询计划”来解决此问题。

Answer 2

尝试对您的第一个查询执行此操作

    from tct v1
    join tct v2 on v2.pk = @pk
    where v1.a = v2.a and v1.b = v2.b and v1.dt = v2.dt

这是你的第二个问题

    from tct v1
    join tct v2 on v2.pk = @pk and v1.pk <> @pk
    where v1.a = v2.a and v1.b = v2.b and v1.dt = v2.dt

这个逐行的T-SQL脚本可以转换为基于集合的脚本吗？

2 个答案: