Primary Key supported by a non-unique index?

Question

Encountered a weird situation today. A utility we setup which allows Analysts to restore data into their tables, started failing when it attempted to drop an index. The index was supporting a Primary Key. Makes sense. But our script was supposed to only be attempting to drop/recreate non-unique indexes. Turns out the supporting index on the Primary Key was non-unique, and to the best of my knowledge came about as follows:
SQL> create table junk (f number(1));
Table created.
SQL> create index junk_ix on junk(f);
Index created.
SQL> select UNIQUENESS from DBA_INDEXES where index_name = 'JUNK_IX';
UNIQUENES
NONUNIQUE
SQL> alter table forbesc.junk add constraint junk_pk primary key (f) using index junk_ix;
Table altered.
SQL> select UNIQUENESS from DBA_INDEXES where index_name = 'JUNK_IX';
UNIQUENES
NONUNIQUE
SQL> insert into junk values (1);
1 row created.
SQL> insert into junk values (1);
insert into junk values (1)
ERROR at line 1:
ORA-00001: unique constraint (FORBESC.JUNK_PK) violated
SQL> select index_name from dba_constraints where constraint_name = 'JUNK_PK';
INDEX_NAME
JUNK_IXWhat I can't figure out is how a non-unique index is enforcing uniqueness. I thought that it was the key in that very same process. I thought that perhaps an index with the 'SYS_123456' was getting created, perhaps, but I couldn't find one:
SQL> select object_name, object_type from dba_objects order by created desc;
OBJECT_NAME   OBJECT_TYPE
JUNK_IX     INDEX
JUNK     TABLE
...How is the uniqueness getting enforced in this case? This is in Oracle 11.1.0.7
Thanks,
--=Chuck

Answer 1

It has always been that way. Oracle can, and will, use a non-unique index to enforce a PK constraint, The existing index just needs to have the PK column(s) as the leading column(s) of the index:
SQL> create table t (id number, id1 number, descr varchar2(10));
Table created.
SQL> create index t_ids on t(id, id1);
Index created.
SQL> select index_name from user_indexes
  2  where table_name = 'T';
INDEX_NAME
T_IDS
SQL> alter table t add constraint t_pk
  2  primary key (id);
Table altered.
SQL> select index_name from user_indexes
  2  where table_name = 'T';
INDEX_NAME
T_IDS
SQL> insert into t values (1, 1, 'One');
1 row created.
SQL> insert into t values (1, 2, 'Two');
insert into t values (1, 2, 'Two')
ERROR at line 1:
ORA-00001: unique constraint (OPS$ORACLE.T_PK) violatedJohn