---------------------------------------------------------------------------------------------------
| Id  | Operation                    | Name     | Rows  | Bytes | Cost  |    TQ  |IN-OUT| PQ Distri
---------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT             |          |     3 |     9 |     6 |        |      |
|   1 |  PX COORDINATOR              |          |       |       |       |        |      |
|   2 |   PX SEND QC (RANDOM)        | :TQ10002 |     3 |     9 |     6 |  Q1,02 | P->S | QC (RAND)
|   3 |    HASH GROUP BY             |          |     3 |     9 |     6 |  Q1,02 | PCWP |
|   4 |     PX RECEIVE               |          |     3 |     9 |     6 |  Q1,02 | PCWP |
|   5 |      PX SEND HASH            | :TQ10001 |     3 |     9 |     6 |  Q1,01 | P->P | HASH
|   6 |       HASH GROUP BY          |          |     3 |     9 |     6 |  Q1,01 | PCWP |
|   7 |        VIEW                  |          |    14 |    42 |     6 |  Q1,01 | PCWP |
|   8 |         SORT ORDER BY        |          |    14 |    42 |     6 |  Q1,01 | PCWP |
|   9 |          PX RECEIVE          |          |    14 |    42 |     2 |  Q1,01 | PCWP |
|  10 |           PX SEND RANGE      | :TQ10000 |    14 |    42 |     2 |  Q1,00 | P->P | RANGE
|  11 |            PX BLOCK ITERATOR |          |    14 |    42 |     2 |  Q1,00 | PCWC |
|  12 |             TABLE ACCESS FULL| EMP      |    14 |    42 |     2 |  Q1,00 | PCWP |
---------------------------------------------------------------------------------------------------

--------------------------------------------------------------------------------------------------
| Id  | Operation                | Name     | Rows  | Bytes | Cost  |    TQ  |IN-OUT| PQ Distrib |
--------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT         |          |     3 |     9 |    27 |        |      |            |
|   1 |  PX COORDINATOR          |          |       |       |       |        |      |            |
|   2 |   PX SEND QC (RANDOM)    | :TQ10001 |     3 |     9 |    27 |  Q1,01 | P->S | QC (RAND)  |
|   3 |    HASH GROUP BY         |          |     3 |     9 |    27 |  Q1,01 | PCWP |            |
|   4 |     PX RECEIVE           |          |     3 |     9 |    27 |  Q1,01 | PCWP |            |
|   5 |      PX SEND HASH        | :TQ10000 |     3 |     9 |    27 |  Q1,00 | P->P | HASH       |
|   6 |       HASH GROUP BY      |          |     3 |     9 |    27 |  Q1,00 | PCWP |            |
|   7 |        PX BLOCK ITERATOR |          |    14 |    42 |     2 |  Q1,00 | PCWC |            |
|   8 |         TABLE ACCESS FULL| EMP      |    14 |    42 |     2 |  Q1,00 | PCWP |            |
--------------------------------------------------------------------------------------------------

alter table emp parallel 4;

You’re absolutely right. The application code is wrong if it’s expecting an ordered result without supplying an “order by” clause – and that’s true under any circumstances. It just happens that sometimes we put in an “order by” clause and the execution plan doesn’t end with a “sort order by” operation because the optimizer has determined that previous operations will guarantee that the data will be appearing in the right order as a guaranteed side effect of the earlier operatins.

The only point I was making with the note was that there are still cases even with the latest patching where the code with the inline view doesn’t produce the same execution plan as the code with a CTE that has been inlined.

in any case… Is the same true if you use a full view with an order by clause.. ?

ie

create or replace view e
as
  select
        /*+ gather_plan_statistics */
        deptno
  from    emp
  order by deptno
;

select   deptno, count(*)
from  e
group by deptno
;

“what is the advantage of putting the order by in the “inner” query?”

Don’t ask me, I didn’t write the application: “… this third party application is … “. The sample above is just demonstrating the issue, perhaps if we saw an example of the real application code we might some rationale for doing things this way.

i would always have written this query as

select   deptno, count(*)
from (
    select
        /*+ gather_plan_statistics */
        deptno
    from    emp
) e
group by deptno
order by deptno
;