Which sql processes fast?

Question

I am writing a program that will change the emplids in a database from one value to another.
The way that I am doing it is to get the tables where emplid is a column (and the ones that have a column name ending in ID and checking if they prompt against a table that has emplid)
I am storing the one to new values in a table ep_t_epxh900
So as I process each table from the list above I can do one of these processes
begin-SQL ON-ERROR=UPDATE-ERROR
Update [$Table] t
set t.emplid = (select emplid from PS_EP_epxh900
where old_emplid = t.emplid)
where t.emplid in (select old_emplid from PS_EP_epxh900
where old_emplid = t.emplid)
end-SQL
or
Read through the ep_t_epxh900 one emplid at a time and issue the above update just for one employee.
Caveat, some tables have over 32 million rows.
I will be using the large rollback segment
If I use the latter solution I can issue a commit at the change of employee id. If I use the former I will need to do it at the change of a table.
Which in your opinion is the better solution. I seem to think that the latter is safer, what do you think?

Answer 1

I generally use merge statement for this kind of logic because is easy.
Note: Applicable to Oracle 9i and above
merge  /*+ parallel(t1,6) append */ into t1
using  PS_EP_epxh900 p on t1.emplid=p.old_emplid
when matched then update set t.emplid=p.new_empidOr try this --not testd
UPDATE ( SELECT tgt.emplid       AS  tgt_name
                       , src.new_empid AS src_name
                       ,src.old_empid    AS src_old_namel
              FROM   t1 tgt
                       , PS_EP_epxh900  src
             WHERE  tgt.emplid = src.src_old_namel)
    SET tgt_name = src_name;
Edited by: user627047 on Jul 18, 2012 10:58 AM