Fastest way to update column in all tables of schema

Question

In our schema we have two columns ColA and ColB common in all tables in our schema.
Suppose these columns have values as below in all tables
ColA     ColB     
A1     B11     
A12     B22     
ABC     DEF     
Now we have to update ColA and ColB where we have alphanumeric values in all tables, some tables have few hundred records and some tables have millions of records.
Could you gurus suggest me with a fastest way to acheive this.
What we are thinking is to write a procedure where we can input multiple tables which could be updated simultaneously and make a collection within procedure with following values
ColA     ColA_R     ColB     ColB_R     
A1     aa     B11     bb     
A12     aaa     B22     bbb     
ABC     No Update DEF     No Update     
So whenever we have value matching A1 update it with value aa if we have value matching B11 update it with value bb and so on.
Your inputs are welcome so that to acheive this in fastest manner.
Thanks,
Tony
Edited by: tony29743 on Nov 9, 2010 9:15 AM

Answer 1

I would be tempted to do it something like this:
Create an index organized table for the cola updates (old_val, new_val) with a PK on old_val and another one for the colb Values. This could possibliy be a single table, depending on how many distinct values there were for cola and colb and if you are sure that "But if colA and colB have value A1 then it will be updated with aa".
Then do the updates as an updateable join view something like:
UPDATE (SELECT t1.cola, iot.new_val
        from tab1 t1, new_values_iot iot
        where t1.cola = iot.old_val)
SET cola = new_valThis would require two rounds of updates, one for cola and one for colb, but they could be parallelized somewhat by distributing the tables to be updated through several pl/sql blocks each updating a different set of tables.
You may be able to do it in a single query like:
UPDATE (SELECT t1.cola, t1.colb, iota.new_val new_vala, iotb.new_val new_valb
        from tab1 t1, new_values_iot iota, new_values_iot iotb,
        where t1.cola = iota.old_val and
              t1.cola = iotb.old_val)
SET cola = new_vala,
    colb = new_valbHowever, given that you said there were some values in both cola and colb that did not require updating, that may not work since the join will fail on one of cola or colb if that value is not in the IOT, so you will not get all of the rows updated. If, and it is a big if, either both of cola and colb or neither of cola annd colb need to be updated in a single row, it might work. So, looking at your original examples (ABC and DEF do not require updates but A1 does), if there could be as case where cola = 'A1' and colb = 'DEF' then you will have to do it in two updates per table.
John