Question

SQL Server 2008.

I have a parent row with pk id 1. While blocking all other DB users (this is a cleanup operation so resource contention is not an issue), I would like to insert a new row, then take all of the child rows and change their fk column to the new row. With the below DDL for example, I would like to insert a new row and give all of the #chi.parid values a value of ‘3’ so they would essentially now belong to the new row so the old one can be deleted.

Help!

create table #par ( parid int identity(1,1) , note varchar(8) )
create table #chi ( chiid int identity(1,1) , parid int , thing varchar(8) )
insert into #par values ( 'note1' )
insert into #par values ( 'note2' )
insert into #chi values ( 1 , 'a' )
insert into #chi values ( 1 , 'b' )
insert into #chi values ( 1 , 'c' )

Answer 1

Thanks all for the input. It appears I “can’t” do a set-based operation on this with SQL Svr 2008, so I did RBAR solution with a loop (I think it performs better than a cursor). Anyone who can comment on making this safer with try..catch or enlighten me more on doing this in a set, please comment. 🙂

Thanks.

    Select 
                [parid]
            ,   [name]
    Into    #redo
    From    partable 
    Where   DateDiff( Hour , donewhen ,SysDateTimeOffset() ) > 23
Begin Transaction
Declare @rows int  = ( Select COUNT(*) From #redo )
Declare @parid int 
Create Clustered Index redoix on #redo([parid]) With FillFactor = 100
While @rows > 0
Begin
    Select Top 1 @parid = [parid] from #redo Order By parid Asc

        Insert partable 
            (
                [name]
            )
        Select 
                [name]      
        From #redo 
        Where parid = @parid
        Update chitable
            Set parid = Scope_Identity()
            Where   parid = @parid
        Delete From partable
            Where   parid = @parid

    Delete from #redo where [parid] = @parid 
    Set @rows  = ( Select COUNT(*) From #redo )
End
Commit Transaction