>In my system, delete record is not an avoidable event.
>I delete the record by delete command only(mark the record deleted), I physically remove them sometime in the program by the Pack.
>When the record is marked for deleted, it still violate the pirmar key rule. If the deleted record have the same key value.
>How can I solve this problem? I solve it by include !deleted() in the primary key definition. Is it the best method?

The problem should be resolved by sophisticated inserting of new records, i.e. you check for new record key uniqueness for all records in table (deleted and non-deleted). If you want to use key from deleted record anyway, you can just recall the record and replace values instead of inserting.