Using the analytic function ROW_NUMBER(), unique values can be selected as shown in 
the query below.

WITH q_dup AS
(SELECT 
   MOD(level,2)+1 dbt_idx,
   TO_CHAR(sysdate+MOD(level,2),'yyyymmdd') dbt_date_id
 FROM DUAL
 CONNECT BY LEVEL <= 6),
w_unique AS
(SELECT  
   dbt_idx,  dbt_date_id,
   ROW_NUMBER() OVER (PARTITION BY dbt_idx ORDER BY dbt_date_id) row_val
 FROM q_dup)
SELECT
 dbt_idx,  dbt_date_id
FROM  w_unique
WHERE  row_val = 1;



Unique rows of data can be inserted into the table using the query below restricting the
data to row_val of 1 to create unique data set.  The row_number window setup can be 
customzied based on data being inserted. 

INSERT INTO dbt_nf11g_demo
 (dbt_idx, dbt_date_id)
WITH q_dup AS
(SELECT 
   MOD(level,2)+1 dbt_idx,
   TO_CHAR(sysdate+MOD(level,2),'yyyymmdd') dbt_date_id
 FROM DUAL
 CONNECT BY LEVEL <= 6),
w_unique AS
(SELECT  
   dbt_idx,  dbt_date_id,
   ROW_NUMBER() OVER (PARTITION BY dbt_idx ORDER BY dbt_date_id) row_val
 FROM q_dup)
SELECT
 dbt_idx,  dbt_date_id
FROM  w_unique
WHERE  row_val = 1;



Use of Oracle 11g hint - IGNORE_ROW_ON_DUPKEY_INDEX