How to prevent OLAP from doing unneccessary aggredations during data load?

Question

Hi,
I'm trying to create a relatively simple two-dimensional OLAP cube (you might want to call it "OLAP square"). My current environment is 11.2EE with AWM for workspace management.
One dimension is date, all years->year->month->day, the other one is production unit, implemented as a hierarchy with a certain machine at the bottom level. The fact is defined by a pair of  bottom-level values of these dimensions; for instance, a measure is taken once a day from each machine. I would like to store these detailed facts in a cube together with aggregates, so they could be easily drilled down to without querying the original fact table.
The aggregation rules are set to "Aggregate from level = default" (which is day and machine respectively) for both of my dimensions, the cube is mapped to fact table with dimension tables, the data is loaded, and the whole thing is working as expected.
The problem is with the load itself, I noticed it being too slow for my amount of sample data. After some investigation of the issue I found out a query in cube_build_log table, a query the data is actually being loaded with.
<SQL>
  <![CDATA[
SELECT /*+  bypass_recursive_check  cursor_sharing_exact  no_expand  no_rewrite */
  T4_ID_DAY ALIAS_37,
  T1_ID_POT ALIAS_38,
  MAX(T7_TEMPERATURE)  ALIAS_39,
  MAX(T7_TEMPERATURE)  ALIAS_40,
  MAX(T7_METAL_HEIGHT)  ALIAS_41
FROM
  SELECT /*+  no_rewrite */
    T1."DATE_TRUNC" T7_DATE_TRUNC,
    T1."METAL_HEIGHT" T7_METAL_HEIGHT,
    T1."TEMPERATURE" T7_TEMPERATURE,
    T1."POT_GLOBAL_ID" T7_POT_GLOBAL_ID
  FROM
    POTS."POT_BATH" T1   )
  T7,
  SELECT /*+  no_rewrite */
    T1."ID_DIM" T4_ID_DIM,
    T1."ID_DAY" T4_ID_DAY
  FROM
    RI."DIM_DATES" T1   )
  T4,
  SELECT /*+  no_rewrite */
    T1."ID_DIM" T1_ID_DIM,
    T1."ID_POT" T1_ID_POT
  FROM
    RI."DIM_POTS" T1   )
  T1
WHERE
  ((T4_ID_DIM = T7_DATE_TRUNC)
    AND (T1_ID_DIM = T7_POT_GLOBAL_ID)
    AND ((T7_DATE_TRUNC)  IN  < a long long list of dates for currently processed cube partition is clipped >  ) ) ) 
GROUP BY
  (T1_ID_POT, T4_ID_DAY) 
ORDER BY
  T1_ID_POT ASC NULLS LAST ,
  T4_ID_DAY ASC NULLS LAST ]]>>
</SQL>
Notice T4_ID_DAY,  T1_ID_POT in the top level column list - these are bottom-level identifiers of my dimensions, which means the query isn't actually doing any aggregation here, as there is only one fact per each pair of (ID_DAY, ID_POT).
What I want to do is to somehow load the data without doing this (totally useless in my case) intermediate aggregation. Basically, I want it to be something like
SELECT /*+  bypass_recursive_check  cursor_sharing_exact  no_expand  no_rewrite */
  T4_ID_DAY ALIAS_37,
  T1_ID_POT ALIAS_38,
  T7_TEMPERATURE  ALIAS_39,
  T7_TEMPERATURE  ALIAS_40,
  T7_METAL_HEIGHT  ALIAS_41
FROM etc...
without any aggregations. In fact, I can live even with this loading query, as the amounts of data are not that large, but I want things to work the right way (more or less ).
Any chance to do it?
Thanks.

Answer 1

I defined a primary key over all the dimension keys in the fact table but for some reason the build still uses an aggregation. Probably because the aggregation operator for the cube I'm currently playing with is actually set, and I don't see a way to undefine it from the UI toolbar you are referring to. This is a piece of mapping section from the workspace file I exported using AWM.:
<CubeMap
      Name="MAP1"
      IsSolved="False"
      Query="POT_BATH_T"
      AggregationMethod="SUM">
Looks like the build aggregates because it is clearly told to so by the AggregationMethod attribute? Any way to override it?