In my work I need to convert data from one system to another all the time and most of these comes from God knows where and you simply don´t understand how come someone could work with a table with no primary key on it and with lot of duplicated rows.

Let´s  transform words into picture.

Create a table and fill in some data with the following script.

create table test (
          cdempgrp integer,
          cdfilial integer,
          nrbem integer,
          nrinc integer 
);

insert into test values( 1, 1, 1, null );
insert into test values( 1, 1, 1, null );
insert into test values( 1, 1, 1, null );
insert into test values( 1, 1, 2, null );
insert into test values( 1, 1, 2, null );
insert into test values( 1, 1, 3, null ); 
insert into test values( 1, 1, 3, null );
insert into test values( 1, 1, 3, null );
insert into test values( 1, 1, 4, null ); 
insert into test values( 1, 1, 4, null );

As you can see there´s lot of duplicated rows. Suppose you need a diferent value for NRINC column for each NRBEM.  In  fact in this case we don´t need a different NRINC for each line.  NRINC is supposed to go from zero to n for each NRBEM.

That´s a problem that you will run over all the time.   And it doesn´t seem to have an easy solution.  Because when you think  you have to make an update you simply don´t know how to write your WHERE clause.

The following script solves this problem.

declare   
       vint   integer;
       vnrbem integer;
begin
       vnrbem := 0;    
      for rec in ( select nrbem from test2 order by nrbem ) 
      loop
             if rec.nrbem <> vnrbem then
                  vint := -1;
                  vnrbem := rec.nrbem; 
           end if;
           vint := vint + 1;
           update test2 set nrinc = vint 
           where nrbem = rec.nrbem and
                       nrinc is null and 
                       rownum = 1;
   end loop;

end;

After all. A simple solution for a big problem. Tranks to Oracle.

That´s all for now.

Reginaldo