>>>I have a table where 3 character fields are combined to form the primary key. Because of the manner in which this table is now used (not as originally designed), a record can be deleted from the table by one user and later another user can add a record with the same key values. Auditing constraints require the deleted record (with datetime stamps) remain in the database; packing and recycling is not allowed.
>>>
>>>Of couse I understand using a surrogate key would be best, but I'm not a liberty to add fields to the table. I'm changing the index via a new "re-indexing" routine added to the executable (a one-time deliverable).
>>>
>>>In effect, I need as the index "cField1 + cField2 + cField3 + NOT DELETED()".
>>>
>>>or, if you prefer, "Field1_c + Field2_c + Field3_c + NOT DELETED()".
>>>
>>>Any help will be greatly appreciated.
>>
>>Could you use this:
>>INDEX ON Field1_c + Field2_c + Field3_c .... FOR NOT DELETED()
>
>This sounds good, I'll give it a try. However, since this creates a "regular" index, (I was trying alter table add primary...), how do I ensure only one active record is allowed for the combination of field1 + field2 + field3. Note, any number of deleted records can exist, but only one can be "active". Also, this is a networked, multi-user application, if that helps.

You can maintain 'uniqueness' either by using 'UNIQUE' or 'CANDIDATE' (depending on your purpose) clause, or by validating user entry when he/she saves a record.