merge data from multi data source

Shuyu Kameda

Hello 

I want to merge data from multi data source.
the replace option in datareader clear the Infocube cells so not suitable for import multi data source.

I have two cubes, 'tmp' and 'data'.
I made the data reader loading data from data source to 'tmp'.
and I made the data flow procedure. Datail is the following,
    block A: tmp
    block B: data
    flow: B = if(A<>0, A,

The problem with this method is that it takes longer to process more.
Do you have any other solution?

Thanks for your help,

------------------------------
Shuyu Kameda
??????
ZEAL CORPORATION
Japan
------------------------------

Etienne CAUSSE

Hi @Shuyu Kameda,
I also have some cases where this kind of calculations is necessary, and the subsequent algorithm (Cellbased) is much slower than other calculations. I did not find a good solution so far with Board 10.6.
There might be solutions using the calculation domain options of Board 11, but I don't use that version so I'm not sure if that would fit the requirement. @Pietro Ferrari, @Björn Reuber, @Antonio Speca : you're the experts, any idea ?

------------------------------
Etienne CAUSSE
Finance Project Manager
France
------------------------------
-------------------------------------------

Helmut Heimann

Hi @Shuyu Kameda and @Etienne CAUSSE,

the Dataflow works in replace-mode, as well as the DataReader would --but, it works select-sensitive. 
My idea would be to select prior to the flow, restricting data to those dimensions having values in "tmp". Thus, you wouldn't need the algorithm and the flow should be faster.

Kind Regards,
Helmut

------------------------------
Helmut Heimann
Senior Consultant
Board Community
Germany
------------------------------
-------------------------------------------

Etienne CAUSSE

Hallo @Helmut Heimann,
I also had that idea, but the selection cannot be properly performed through a "select based on cube", because it works one dimension at a time.
For example I have in my cube Product and Organization dimensions.

If the "tmp" cube has a value for Product A and Organization B, and a value for Product B and Organization A, then all 4 combinations (AxA, AxB, BxA, BxB) will be selected and within the scope of the calculation.

So we would need a selection by combination at the cube / sparsity level, not by dimension. It is my understanding that the Board 11 calculation domain was one way to achieve that.
If you know a way to do the selection precisely with Board 10 I'm all ears

Etienne

------------------------------
Etienne CAUSSE
Finance Project Manager
France
------------------------------
-------------------------------------------

Helmut Heimann

Hi @Etienne CAUSSE,

that's correct, but after all - we don't know the dimensions of the cube ;-)

Kind Regards,
Helmut

------------------------------
Helmut Heimann
Senior Consultant
Board Community
Germany
------------------------------
-------------------------------------------

Antonio Speca

Hello @Shuyu Kameda ,

If the cubes have the same structure you may try to use the "try force hbmp" mode in dataflow.
If the cubes have not the same structure you may create some temp cube with the same structure of the target to obtain that result.
Please also check the density/sparsity of the target. If you can't find a solution probably the best way to move on is to ask the support.

@Etienne CAUSSE B11 is natively faster on some calculations, domain is not meant to optimize but only to decide where to write.
Basically the tecniques to make df faster are the same as previous versions

------------------------------
Antonio Speca
Product Specialist
Board Community
Switzerland
------------------------------
-------------------------------------------

Etienne CAUSSE

Hi @Antonio Speca,
I get your point.

In the case presented, could we replace the current formula "B = if(A<>0, A, " with a simpler formula B = A + the calculation domain defined by the A cube ?
In that case the algorithm is simpler (so I guess faster) and the result would be the same ; right ?

Thanks for your clarification

------------------------------
Etienne CAUSSE
Finance Project Manager
France
------------------------------
-------------------------------------------

Antonio Speca

Hello @Etienne CAUSSE

If the cubes have exactly the same srtucture it might have some benefit.
I didn't notice before that the target cube is also one of the sources in the formula.
For sure @Shuyu Kameda can get a benefit by using a third cube instead and using the try force hbmp switch.

------------------------------
Antonio Speca
Product Specialist
Board Community
Switzerland
------------------------------
-------------------------------------------

Shuyu Kameda

Thanks, @Etienne CAUSSE @Helmut Heimann @Antonio Speca

In my situation, using the selection and the try force hbmp may solve my problem.

Thanks to the professionals!

king regards,


------------------------------
Shuyu Kameda
??????
ZEAL CORPORATION
Japan
------------------------------
-------------------------------------------