Oracle sql merge to insert and delete but not update

Question

Is there a way to use oracle merge to insert and delete but not update?

I have a table representing a set of values related to a single row in another table. I could change the set of values by deleting them all and adding back the new set, or by selectively deleting some and adding others, but I am interested in making it a single statement if possible.

Here is a working example with update. In order to make this work, I had to add dummy so that a column was available to update that was not in the on condition. Is there some way to only delete and insert without a dummy column to update?

No column from the on condition may be in the update set list even if it is not actually updated.

create table every_value ( the_value varchar2(32) );
create table paired_value ( the_id number, a_value varchar2(32) , dummy number default 0 );
-- the_id is a foreign_key to a row in another table

insert into every_value ( the_value ) values ( 'aaa' );
insert into every_value ( the_value ) values ( 'abc' );
insert into every_value ( the_value ) values ( 'ace' );
insert into every_value ( the_value ) values ( 'adg' );
insert into every_value ( the_value ) values ( 'aei' );
insert into every_value ( the_value ) values ( 'afk' );

-- pair ace and afk with id 3
merge into paired_value p using every_value e
on ( p.the_id = 3 and p.a_value = e.the_value )
when matched then update set dummy=dummy+1
delete where a_value not in ('ace','afk')
when not matched then insert (the_id,a_value)
values (3,e.the_value)
where e.the_value in ('ace','afk');

-- pair ace and aei with id 3
-- should remove afk, add aei, do nothing with ace
merge into paired_value p using every_value e
on ( p.the_id = 3 and p.a_value = e.the_value )
when matched then update set dummy = dummy+1
delete where a_value not in ('ace','aei')
when not matched then insert (the_id,a_value)
values (3,e.the_value)
where e.the_value in ('ace','aei');

-- pair aaa and adg with id 4
merge into paired_value p using every_value e
on ( p.the_id = 4 and p.a_value = e.the_value )
when matched then update set dummy = dummy+1
delete where a_value not in ('aaa','adg')
when not matched then insert (the_id,a_value)
values (4,e.the_value)
where e.the_value in ('aaa','adg');

select * from paired_value;

I have tried this in oracle 10g and, with this sqlfiddle, oracle 11g.

Answer 1

No, you cannot delete rows that have not been updated by the merge command.
Here is documentation: http://docs.oracle.com/cd/B28359_01/server.111/b28286/statements_9016.htm

Specify the DELETE where_clause to clean up data in a table while populating or updating it. The only rows affected by this clause are those rows in the destination table that are updated by the merge operation. The DELETE WHERE condition evaluates the updated value, not the original value that was evaluated by the UPDATE SET ... WHERE condition. If a row of the destination table meets the DELETE condition but is not included in the join defined by the ON clause, then it is not deleted. Any delete triggers defined on the target table will be activated for each row deletion.

That means, that rows must be updated. Hovewer, you don't need to update all rows, after UPDATE use the same WHERE clause as you are using after DELETE

when matched then update set dummy=dummy
    where a_value not in ('ace','afk')
delete 
    where a_value not in ('ace','afk')

Answer 2

I have found you can set the column to itself:

MERGE ...
WHEN MATCHED THEN 
   UPDATE SET a_value = a_value WHERE a_value not in ('ace','afk')
   DELETE WHERE a_value not in ('ace','afk')

This negates the need for the dummy column.