MDX Calculate the sum of distinct values - distinct

Question

I'm wondering if it is possible to create a calculated member to obtain the sum of distinct values for a fact. I will try to explain it with the following example:
I have a fact where the primary key is related with two dimensions (one to many cardinality). The fact contains a measure and its value is the same for all members of each distinct combination of FACT_ID and DIM_1_ID. For the total, I don't want to consider multiple times the same values. So, with the following values the total should be 450 and not 850 (default Mondrian behavior).
| FACT_ID | DIM_1_ID | DIM_2_ID | MEASURE |
|---------|----------|----------|---------|
| 1 | A | D | 100 |
| 1 | A | E | 100 |
| 1 | B | F | 50 |
| 2 | A | D | 300 |
| 2 | A | E | 300 |
|---------|----------|----------|---------|
TOTAL | 450 |
Is it possible? How can it be done with Mondrian?
Thanks in advance
UPDATE - Current status
As described in one of the comments bellow, base on #whytheq's answer, I managed to calculate the right value for the total, using the following MDX formula for the measure:
Sum(
Order(
[dActivity.hActivity].[lActivity].MEMBERS*[dFacility.hFacility].[lFacilit‌​y].MEMBERS,
[dActivity.hActivity].[lActivity].currentmember.name
) as [m_set] ,
iif(
[m_set].currentordinal = 0
OR
not(
[m_set]
.item([m_set].currentordinal)
.item(0).NAME
=
[m_set]
.item([m_set].currentordinal-1)
.item(0).NAME
) ,
[Measures].[mBudget]
,
0
)
)
However, this expression is using the complete set for every single row, so the result overrides the measure real value for the different fact rows.
| FACT_ID | DIM_1_ID | DIM_2_ID | MEASURE |
|---------|----------|----------|---------|
| 1 | A | D | 450 |
| 1 | A | E | 450 |
| 1 | B | F | 450 |
| 2 | A | D | 450 |
| 2 | A | E | 450 |
|---------|----------|----------|---------|
TOTAL | 450 |

Answer 1

Great question - really tricky to do in MDX.
If we do the following then there are 158 rows returned - a handful have duplicate values for [Measures].[Internet Sales Amount]:
SELECT
[Measures].[Internet Sales Amount] ON 0
,NON EMPTY
Order
(
[Product].[Product].[Product]
,[Measures].[Internet Sales Amount]
,bdesc
) ON 1
FROM [Adventure Works];
This only counts them if the member above is different for the respective measure:
WITH
SET [x] AS
Order
(
NonEmpty
(
[Product].[Product].[Product]
,[Measures].[Internet Sales Amount]
)
,[Measures].[Internet Sales Amount]
,bdesc
)
SET [FILTERED] AS
Filter
(
[x]
,
(
[x].Item(
[x].CurrentOrdinal - 1)
,[Measures].[Internet Sales Amount]
)
<>
(
[x].Item(
[x].CurrentOrdinal)
,[Measures].[Internet Sales Amount]
)
)
MEMBER [Measures].[distCount] AS
Count([FILTERED])
SELECT
[Measures].[distCount] ON 0
FROM [Adventure Works];
Maybe try adding the EXISTING keyword into your calculatio:
Sum
(
Order
(
EXISTING //<<<
[dActivity.hActivity].[lActivity].MEMBERS
*
[dFacility.hFacility].[lFacilit‌​y].MEMBERS
,[dActivity.hActivity].[lActivity].CurrentMember.Name
) AS [m_set]
,IIF
(
[m_set].CurrentOrdinal = 0
OR
(NOT
[m_set].Item(
[m_set].CurrentOrdinal).Item(0).Name
=
[m_set].Item(
[m_set].CurrentOrdinal - 1).Item(0).Name)
,[Measures].[mBudget]
,0
)
)

Answer 2

You could try to obtain the average over the set. The code is a bit complex.
WITH SET SomeSet AS
{
Fact.FactID.FactID.MEMBERS
*
Fact.DimID1.DimID1.MEMBERS
*
Fact.DimID2.DimID2.MEMBERS
}
MEMBER Measures.AvgVal AS
AVG
(
{Fact.FactID.CURRENTMEMBER}
*
{Fact.DimID1.CURRENTMEMBER}
*
NonEmpty
(
Fact.DimID2.DimID2.MEMBERS,
{{Fact.FactID.CURRENTMEMBER} *
{Fact.DimID1.CURRENTMEMBER}} *
[Measures].[TheMeasure]
)
,
[Measures].[TheMeasure]
)
SELECT NON EMPTY SomeSet ON 1,
NON EMPTY {
[Measures].[TheMeasure],
Measures.AvgVal
} on 0
from [YourCube]
What I am doing is, for the current FactID- DimID1 combination on the axis, I am getting the list of all possible DimID2s and then, over the internally generated non-empty tuples of FactID-DimID1-DimID2, deriving the average value of the measure TheMeasure
So, for example (100+100)/2 = 100 value would be displayed for the combination of FactID = 1 and DimID1 = A