--- In firebird-support@yahoogroups.com, Ales Smodis <aless@z...> > So
I'm left with two rows with the same key and don't know how to delete

> one of them since the delete statement doesn't take a plan and whatever
> condition I use to identify the row I want to delete, it always uses an
> index.

Ales, I usually make it following next idea, which you can adopt to
multy-segment PK modifying Where, and for deleting many duplicate by
removing Exit and skipping first row, selected column have no meaning:

CREATE PROCEDURE DELETE_ONE_DUPLICATE(ID INTEGER)
DECLARE VARIABLE K INTEGER;
BEGIN
FOR SELECT ID FROM DAMAGED_TABLE
WHERE ID=:ID
INTO :K
AS CURSOR TMPCURSOR
DO
BEGIN
DELETE FROM DAMAGED_TABLE
WHERE CURRENT OF TMPCURSOR;
EXIT;
END
END

Theorethically the same can be done selecting RDB$DB_KEY on client
side and using it in delete, but personally I have not much experience
with this, if you are interested in general sense, you can try.

Best regards,
Alexander