Sequence in combination with function

Question

Hi,
I have an issue with a sequence in combination with a function.
I want to generate some kind of an inventory material report that contains some kind of a material ID (a non-repeating, auto-incrementing number, gaps allowed). However I have to keep track of which material IDs I have used for which inventory material, for traceability reasons. Another important requirement is that whenever the report is run, that the returned material IDs are never repeating. So repeatedly runs for the same inventory material should always return new material ID.
So I was about to create a simple PL/SQL function where I pass my data from traceability as parameters. Then the function uses internally a sequence (nextval) to generate a new material ID, then stores the new material ID with the related values from the input parameters into a table (autonomous transaction) and returns the new ID.
This works pretty fine so far.
Now the tricky part: I want have a SQL statement where I want to return one and the same ID in different columns.
Let me give an example:
CREATE SEQUENCE XXX_TEST_SEQ
MINVALUE 1 MAXVALUE 999999999999999999999999999
INCREMENT BY 1 START WITH 1
NOCACHE NOORDER NOCYCLE;
CREATE PACKAGE XXX_TEST_PKG AUTHID CURRENT_USER AS
FUNCTION get_id(p_someparameter NUMBER) RETURN NUMBER;
END XXX_TEST_PKG;
CREATE PACKAGE BODY XXX_TEST_PKG AS
FUNCTION get_id(p_someparameter NUMBER) RETURN NUMBER IS
  l_new_id NUMBER;
BEGIN
  SELECT XXX_TEST_SEQ.NEXTVAL
    INTO l_new_id
    FROM DUAL;
    /* do some more logic, e.g. store the freshly generated ID
        together with the input parameter values in some tracing table */
  RETURN l_new_id;
END get_id;
END XXX_TEST_PKG;This is a simplified version of the function for demo.
Let's use it:
-- 1) this works as expected:
select xxx_test_pkg.get_id(1) from dual;
XXX_TEST_PKG.GET_ID(1)
1                       
-- 2) this returns two different numbers:
select xxx_test_pkg.get_id(1) as id_one
    ,  xxx_test_pkg.get_id(1) as id_two
from dual;
ID_ONE                 ID_TWO                
2                      3                     
-- 3) This is the problematic case: I want to return the same ID two times:
select my_id as id
     , my_id as same_id
from (select xxx_test_pkg.get_id(1) as my_id from dual);
ID                     SAME_ID               
4                      5                     
-- 4) CURVAL on these sequence does not work:
select my_id as id
     , xxx_test_seq.curval as same_id
from (select xxx_test_pkg.get_id(1) as my_id from dual);
SQL Error: ORA-00904: "XXX_TEST_SEQ"."CURVAL": invalid identifier
-- 5) This gives the output that I need, but why?
select my_id as id
     , xxx_test_seq.nextval as same_id
from (select xxx_test_pkg.get_id(1) as my_id from dual);
ID                     SAME_ID               
7                      7                      So query 3 is the one that gives me headaches. I need to return one and the same ID in different columns. But this returns just two different IDs. I know that the optimizer will merge the two select statements into one and therefore execute the function twice. Does anybody know how I could avoid this?
What is a bit strange to me is that query 5 really returns what I need, but at a first glance I would say it should not do so. I would say that using nextval would return me again another sequence number.
This is confusing... Can anybody explain this and advise how I could build a query were I can get the same ID two times per row?
I know I can totally get rid of the sequence in my function and use something different, but I was wondering if I can avoid changing my function.
Thanks,
David.

Answer 1

David Weber wrote:
thanks lot for all the great feedback. That's really helpful. I think I'll go for the NO_MERGE hint. A first test with this looks good.Keep in mind, only undocumented hint MATERIALIZE along with subquery factoring clause WITH (for which this hint was created) will work in real queries:
SQL> select my_id as id
  2       , my_id as same_id
  3  from (select /*+ NO_MERGE */ xxx_test_pkg.get_id(1) as my_id from dual),emp where rownum < 6;
        ID    SAME_ID
        20         20
        21         21
        21         21
        21         21
        21         21
SQL> select my_id as id
  2       , my_id as same_id
  3  from (select xxx_test_pkg.get_id(1) as my_id from dual where rownum=1),emp where rownum < 6;
        ID    SAME_ID
        22         22
        23         23
        23         23
        23         23
        23         23
SQL> select my_id as id
  2       , my_id as same_id
  3  from (select /*+ MATERIALIZE */ xxx_test_pkg.get_id(1) as my_id from dual),emp where rownum < 6;
        ID    SAME_ID
        24         25
        26         27
        28         29
        30         31
        32         33
SQL> with t as (select /*+ MATERIALIZE */ xxx_test_pkg.get_id(1) as my_id from dual)
  2  select my_id as id
  3       , my_id as same_id
  4  from t,emp where rownum < 6;
        ID    SAME_ID
        34         34
        34         34
        34         34
        34         34
        34         34
SQL> SY.