sql - SQL 循环遍历 800 万条记录并更新它们

Question

我有一个包含大约 800 万条记录的审计表。我最近添加了两个新列，我需要使用一些规则/条件从现有列进行更新。基本上，最初，每当在表中更新 FK 时，它都会将新旧 FK id 存储到审计表中。例如

Table A
ID      Name
1       First A
2       Second A
3       Third A

Table B
ID   AID  Name
1     1   First B
2     1   Second B
3     2   Third B

Audit 
ID  TableName FieldName OldValue NewValue 

现在，如果我将表 B 的第一条记录从 1 1 First B 更新为 1 3 First B 那么审计表会将更改存储为

Audit 
ID  TableName FieldName OldValue NewValue 
1    Table B  AID        1          3

现在我更新了审计表来存储 FK 的实际文本值，即上述更改将存储为

Audit 
ID  TableName FieldName OldValue  NewValue  OldText   NewText
1    Table B   AID       1         3        First A   Third A

问题是我已经有大约 800 万条记录需要创建新列。我写了下面的查询来做到这一点

declare @sql nvarchar(max);
declare @start int = 1

while @start <= 8000000 
begin
  select top 10000  @sql = COALESCE(@sql+'Update Audit set ','Update Audit set') + 
    isnull(' OldText = ('+ dbo.GetFKText(i.TableName, i.FieldName)+case when len(isnull(i.OldValue,'')) < 1 then null else i.OldValue end +'),',' OldText = OldValue, ')  + 
    isnull(' NewText = ('+ dbo.GetFKText(i.TableName, i.FieldName)+case when len(isnull(i.NewValue,'')) < 1 then null else i.NewValue end +')',' NewText = NewValue ')    +  
    ' where AuditID = '+cast(i.AuditID as nvarchar(200))+' and lower(ltrim(rtrim(TableName))) <> ''audit'';' 
    from Audit i where i.AuditID >= @start 

  exec sp_executesql @sql

  set @start = @start+10000;
end

获取文本函数（基本上我得到的列具有 name = (TableName)+'Name' 或 (TablName)+(SomeText)+'Name' 这只是我在所有表格中遵循的约定）

     declare @res nvarchar(max)=''; 
     declare @fn nvarchar(200);
     declare @ttn nvarchar(200);
     declare @tcn nvarchar(200);

     SELECT top 1
         @ttn = kcu.table_name  
        ,@tcn = kcu.column_name  
     FROM INFORMATION_SCHEMA.CONSTRAINT_COLUMN_USAGE ccu
     INNER JOIN INFORMATION_SCHEMA.REFERENTIAL_CONSTRAINTS rc
        ON ccu.CONSTRAINT_NAME = rc.CONSTRAINT_NAME 
     INNER JOIN INFORMATION_SCHEMA.KEY_COLUMN_USAGE kcu 
        ON kcu.CONSTRAINT_NAME = rc.UNIQUE_CONSTRAINT_NAME  
     Where ccu.TABLE_NAME = @TableName and  ccu.COLUMN_NAME = @FieldName

     if isnull(@ttn,'') != '' and ISNULL(@tcn,'') != ''
     begin
       select @fn= COLUMN_NAME 
       from (SELECT top 1 COLUMN_NAME , 
                   case when COLUMN_NAME like (@ttn+'Name') then 0 
                   when COLUMN_NAME like (@ttn+'%Name') then 1 
                   when COLUMN_NAME like (@ttn+'Code') then 2 
                   when COLUMN_NAME like (@ttn+'%Code') then 3 else 4 end as CPriority
             FROM JVO.INFORMATION_SCHEMA.COLUMNS
             WHERE TABLE_NAME = @ttn and (COLUMN_NAME like '%Name' or COLUMN_NAME like '%Code'
            )
       order by CPriority) as aa; 

       RETURN 'select '+@fn+' from '+@ttn+' where '+@tcn+' = ';
     end

 return null;

它工作但非常慢，它在 13 小时内更新了大约 100 万条记录。任何人都可以帮助改进此查询或建议更新它的替代方法。

谢谢

标签： sqlsql-serversql-updatesqlperformancebulkupdate