Question

For example, if I do

SELECT ROWID, name from emp where age > 30;

As ROWID does not take any storage space, is it calculated every time a query like this runs?

---

From here,

Each table in an Oracle database
internally has a pseudocolumn named
ROWID. This pseudocolumn is not
evident when listing the structure of
a table by executing a SELECT * FROM
… statement, or a DESCRIBE …
statement using SQL*Plus, nor does the
pseudocolumn take up space in the
table.

---

Taking clue from the comments,

ROWID is stored when you create an index. Suppose I have no other index than the implicit one for the primary key (emp_id). In this case, will the above query go to this implicit index? How will the ROWID calculation happen?

Please note that name and age columns are not a part of the index.

Answer 1

“Suppose I have no other index than
the implicit one for the primary key
(emp_id). In this case, will the above
query go to this implicit index? How
will the ROWID calculation happen?”

Firstly, the “implicit index” is a real index. If we create a primary or unique key on a table and no index exists on the key column(s) Oracle we create an index, with the same name as the constraint.

SQL> create table t72
  2      ( emp_id number not null primary key
  3        , name varchar2(10) not null
  4        , age number(3,0) )
  5  /

Table created.

SQL> select constraint_name from user_constraints
  2  where table_name = 'T72'
  3  and constraint_type='P'
  4  /

CONSTRAINT_NAME
------------------------------
SYS_C001145039

1 row selected.

SQL> select index_type, uniqueness
  2  from user_indexes
  3  where index_name = 'SYS_C001145039'
  4  /

INDEX_TYPE                  UNIQUENES
--------------------------- ---------
NORMAL                      UNIQUE

1 row selected.

SQL>

Secondly, the query filters on the AGE column. So the optimizer would ignore any index on EMP_ID. On that case the database will do a full table scan of EMP, evaluating the value of each AGE column it retrieves. For each record where AGE < 30 it will concatenate the table’s object number, the block number, the slot number and the file number into a ROWID.

If you want to understand more about ROWID have a play around with the DBMS_ROWID package. René Nyffenegger has a useful tutorial on his website. Find out more.

---

“Suppose it was SELECT ROWID, name
from emp where emp_id > 100;. Would
the query fetch the ROWID from the
emp_id index? “

There’s one easy way to tell: experimentation. First we create an index on a table with a lot of records, and freshen the statistics:

SQL> create unique index big_i on big_emp (empno)
  2  /

Index created.

SQL> exec dbms_stats.gather_table_stats(user, 'BIG_EMP', cascade=>true)

PL/SQL procedure successfully completed.

SQL>

Then we see how Oracle tackles the query:

SQL> explain plan for
  2      select empno, rowid from big_emp
  3      where empno > 10000;

Explained.

SQL> select * from table(dbms_xplan.display)
  2  /

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------
Plan hash value: 3238483832

------------------------------------------------------------------------------
| Id  | Operation            | Name  | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |       | 24319 |   403K|    16   (0)| 00:00:01 |
|*  1 |  INDEX FAST FULL SCAN| BIG_I | 24319 |   403K|    16   (0)| 00:00:01 |
------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter("EMPNO">10000)

13 rows selected.

SQL>

If Oracle can satisfy the query with just indexed columns it doesn’t touch the table. Clearly here it is retreiving the ROWID from the index.