11g: automatic group by unused column

Question

Hi,
I've got this problem:
I have report with just one measure. This measure should be aggregated over a selectable dimension.
Therefor I build a dashboard prompt, which saves the selected dimension in a presentation variable.
(btw: a column selector would be the easier way, but there are more than one report which should react on this dashboard prompt)
Then I used a "case when" in the report to differ between the selection:
case when [presentation-variable] = 'Dim1' then Dim1.col1
when [presentation-variable] = 'Dim2' then Dim2.col1
end
Now I would expect such a physical SQL:
select
case when [presentation-variable] = 'Dim1' then Dim1.col1 when [presentation-variable] = 'Dim2' then Dim2.col1 end,
sum(...)
from
group by
case when [presentation-variable] = 'Dim1' then Dim1.col1 when [presentation-variable] = 'Dim2' then Dim2.col1 end;
But OBIEE creates such one:
select
case when [presentation-variable] = 'Dim1' then Dim1.col1 when [presentation-variable] = 'Dim2' then Dim2.col1 end,
sum(...)
from
group by
Dim1.col1,
Dim2.col1,
case when [presentation-variable] = 'Dim1' then Dim1.col1 when [presentation-variable] = 'Dim2' then Dim2.col1 end;
So of course the wrong the measure is aggregated wrong, because it is grouped by a column, which i do not want to use.
Could it be the Problem, that this 2 columns are logical level key of a dimension hierarchy?
Anybody got a idea, how I could suppress this group by?

Answer 1

Hi,
Just 4 hours ago, someone else had a very similar question:
Group by fails on ORA-00979
Personally, I consider that a bug in Oracle 10. When you do a GROUP BY on a view, you can't expect the SQL engine to look at the view definition, and determine if one of the columns can be derived from another.
In practice, it's easy enough to add c1u to the GROUP BY clause, or to use MAX (c1u) instead of c1u, to make your query work.