HOWTO: Add a UNIQUE constraint to a populated column

Question

Here is the way to add a unique constraint to a populated column, ensuring that the existing values are unique as well.
ALTER TABLE t1
ADD CONSTRAINT t1_uk UNIQUE (col1,col2)
EXCEPTIONS INTO my_exceptions
/If the alter table statement fails this will populate the MY_EXCEPTIONS table with the rows that contain duplicate values for (col1,col2). These are identified by ROWID. We can then chose to:
[list]
[*]delete rows with a bad key;
[*]amend the values of the key columns;
[*]re-apply the constraint with the NOVALIDATE option.
[list]
We should be wary of choosing the NOVALIDATE option. There's usually a good reason why the unique constraint is required and we should not circumvent it. Apart from anything else, Oracle allows us to build foreign keys referencing NOVALIDATE unique keys. This could result in child rows that have two parents, which is normal in biology, but very wrong in a database.
NOVALIDATE is useful in datawarehouses and suchlike, because little updating occurs and the data integrity issues are of less imporatance. I don't think it ought to be used in OLTP situations. The code example at the end of this posting illustrates why we should exercise caution with NOVALIDATE.
Once we have handled the duplicate values we can re-run the alter table statement and apply the constraint.
If you don't already have an EXCEPTIONS table (it can be called anything, it's the structure that counts) you may need to run (or get a DBA to run) a script called UTLEXCPT.SQL, which will be in the $ORACLE_HOME/rdbms/admin directory.
Cheers, APC
SQL> SELECT col1, col2, col3 FROM t1;
      COL1       COL2       COL3                                               
         1          1          0                                               
         2          1          0                                               
         2          1          0                                               
SQL> SELECT col1, col2, cola FROM t2;
      COL1       COL2 COLA                                                     
         1          1  I'm child #1                                                        
         1          1  I'm child #2                                                        
         2          1  I'm child #3                                                        
SQL> ALTER TABLE t1
  2  ADD CONSTRAINT t1_uk UNIQUE (col1, col2);
ALTER TABLE t1 ADD CONSTRAINT t1_uk UNIQUE (col1, col2)
ERROR at line 1:
ORA-02299: cannot validate (TST2.T1_UK) - duplicate keys found We can't add a unique key, so we decide to use the NOVALIDATE clause.
SQL> CREATE INDEX t1_i ON t1(col1, col2);
Index created.
SQL> ALTER TABLE t1
  2  ADD CONSTRAINT t1_uk UNIQUE (col1, col2) ENABLE NOVALIDATE;
Table altered.
SQL> INSERT INTO t1 VALUES (2, 1, 1);
insert into t1 values (2, 1, 1)
ERROR at line 1:
ORA-00001: unique constraint (TST2.T1_UK) violated Well, we can't add any more duplicate keys, so that's alright isn't it? Nope. We can add a foreign key to table T2 referencing T1 but we don't know which row in T1 is the parent of the rows in T2.
SQL> ALTER TABLE t2
  2  ADD CONSTRAINT t2_t1_fk foreign key  (col1, col2)
  3  REFERENCES t1(col1, col2)
Table altered.
SQL> SELECT rowid FROM t1 WHERE col1 = 2;
ROWID                                                                          
AAAVnMAANAAAB96AAB                                                             
AAAVnMAANAAAB96AAC                                                             
SQL> DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAB';
1 row deleted.
SQL> DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAC';
DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAC'
ERROR at line 1:
ORA-02292: integrity CONSTRAINT (TST2.T2_T1_FK) violated - child record found
SQL> ROLL
Rollback complete.
SQL> DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAC';
1 row deleted.
SQL> DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAB';
DELETE FROM t1 WHERE rowid = 'AAAVnMAANAAAB96AAB'
ERROR at line 1:
ORA-02292: integrity CONSTRAINT (TST2.T2_T1_FK) violated - child record found
SQL>  ROLL
Rollback complete.
Why this matters: queries joining child and parent tables return more rows than we'd expect normally.
SQL> SELECT t2.cola
  2  FROM   t2
  3  WHERE  t2.col1 = 2
  4  and    t2.col2 = 1;
COLA                                                                           
I'm child #3                                                                   
SQL> SELECT t2.cola, t1.col4
  2  FROM   t1, t2
  3  WHERE  t2.col1 = 2
  4  AND    t2.col2 = 1
  5  AND    t1.col1 = 2
  6  AND    t2.col2 = 1;
COLA                    COL4                                                   
I'm child #3            I'm the daddy                                          
I'm child #3            No, I'm the daddy!!                                    
SQL> Caveat emptor
This posting is issued on behalf of the Rogue Moderators. It is posted with the best of intentions but is not guaranteed in any way, shape or form. In particular, the code is presented "as is" and you assume full responsibility for running it on your system. Furthermore, you must not download and install software from the internet unless you know what you are doing and have the permission of whoever owns your system. It was never this hard for the Lone Ranger.

Answer 1

Hi,
You can not create unique key with duplicate values as you know. Now if you want to create unique key then you have to delete duplicate records from the table but this again if you can afford to delete because this data may be more importante to you and you can not delete any record.
or you can use ENABLE NOVALIDATE option of oracle explain by Tom on below link.
http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:8806498660292
or if you dont want to use Tom's way then you go with primary key like
Drop old primary key if there is already on the column.
create new primary key with enable novalidate option.