How can I create a composite key on multiple columns, one of which can have some value but not null (or some constant value)?
For example:
PK Loc_ID Date Time Cancelled
1 1 01/01/2010 10:00AM YES
2 1 01/01/2010 10:00AM YES
3 1 01/01/2010 10:00AM null
4 1 01/01/2010 10:00AM null - Not Acceptable
Insertion of the fourth record should raise a composite key violation error.
So what you what is to enforce a rule where only record cannot be cancelled for any given permutation of LOC_ID, DATE, TIME? We can do this with a function-based unique index.
This is what we want to avoid:
SQL> select * from t34
2 /
PK LOC_ID SOMEDATE SOMETIM CAN
---------- ---------- ---------- ------- ---
1 1 01/01/2010 10:00AM YES
2 1 01/01/2010 10:00AM YES
3 1 01/01/2010 10:00AM
SQL> insert into t34
2 values (4 , 1 , to_date('01/01/2010','DD/MM/YYYY') , '10:00AM', null )
3 /
1 row created.
SQL>
Let's build an index to enforce the rule
SQL> rollback
2 /
Rollback complete.
SQL> create unique index t34_uidx
2 on t34 (loc_id, somedate, some_time, nvl2(cancelled, pk, null) )
3 /
Index created.
SQL>
The NVL2()
function is a special form of CASE which returns the second argument if the first argument is NOT NULL otherwise the third. The index uses the PK col as the second argument because it is the primary key and hence unique. So the index allows duplicate values of CANCELLED unless they are null:
SQL> insert into t34
2 values (4 , 1 , to_date('01/01/2010','DD/MM/YYYY') , '10:00AM', null )
3 /
insert into t34 values (4 , 1 , to_date('01/01/2010','DD/MM/YYYY') , '10:00AM', null )
*
ERROR at line 1:
ORA-00001: unique constraint (APC.T34_UIDX) violated
SQL>
Could this be done with a unique function based index? Something like:
create unique index ix on tb (
loc_id, date, time, decode(cancelled, null, 1, null));
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With