Difference in writing feeders

[deepakjain2020]

Hi All,

Feeder1: [A] => [B,'Dollar'];
Feeder2: [A,'Dollar']=>;

What is the difference between both the feeders? Does it impact memory?
Which is the most preferred way to be followed?

Regards,
Deepak Jain

[dan.kelleher]

Hi Deepak,

Let's call your dimension with values 'A' and 'B' your measures dimension, and the one with 'Dollar' your currency dimension.

Feeder1 is explicit as to which area it is feeding (right-hand side of feeder). This means that if you have values in any currency for measure 'A', this will feed ['B', 'Dollar'].

Feeder2 is explicit on the currency dimension from where you are feeding from (left-hand side of feeder). Because you are only feeding from ['A', 'Dollar'], and not the other currencies, you don't need to be explicit on the right-hand side of the feeder, but it is effectively the same as:

['A', 'Dollar']=>['B', 'Dollar'];

I am assuming your rule is performing a currency calculation based on a dollar value being present in the cube. If this is the case you will want to feed from 'Dollar' to all other currencies. There are 2 ways in which you can do this. Either feed individually to each currency, or write 1 feeder which feeds to a consolidation ('All Currencies') which has the calculated currencies as children. Feeding to a consolidation is the same as feeding each n-level node in that consolidation individually.

There's a good intro to feeders in the IBM documentation and here. Plus there are countless threads in this forum which will help you to understand feeders.

[deepakjain2020]

Hi Dan,

Good description from your end, thanks for that.

Can you please let me know about how the impact will be on memory of feeders?

Feeder1: [A] => [B,'Dollar'];
Feeder2: [A,'Dollar']=>;

Will the memory consumed by feeders will be same in both the cases?

Regards,
Deepak Jain

[Michel Zijlema]

Hi Dan,

Good description from your end, thanks for that.

Can you please let me know about how the impact will be on memory of feeders?

Feeder1: [A] => [B,'Dollar'];
Feeder2: [A,'Dollar']=>;

Will the memory consumed by feeders will be same in both the cases?

Regards,
Deepak Jain

Hi,

I don't expect there will be a difference in memory used for the feeders between the two, as they're both feeding the exact same area.
Of course if you have a non-zero value on f.i. ['A', 'Euro'] and no value on ['A', 'Dollar'], then Feeder1 will trigger and place a feeder, while Feeder2 won't. In that case there is a difference in memory consumption. Feeder1 potential leads to more feeder triggering (overhead) than Feeder2, so Feeder2 is the more efficient one.
But in the end it depends on what you need - if the rule you're feeding is calculating Dollars based on other currency values, it could be that Feeder1 is the only valid one...

Michel

[beek]

Hi there,

I'm facing a similar situation here. Assuming below is my rule & feeders, any idea which way of putting the feeders would be more efficient ?
A=B+C;
D=A+E;

FEEDERS;
B=>A;
C=>A;
A=>D;
E=>D;

OR
B=>A,D;
C=>A,D;
E=>D;

[Wim Gielis]

beek

I would try to solve this with consolidation structures, since you have sums...

If not, set 2 seems more appropriate. Try to feed from values instead of feeding from cells fed by other cells.

[beek]

Hi Wim Gielis,
Thanks for your reply.

I would try to solve this with consolidation structures, since you have sums...

Refering to the above, can I assume you are refering to the 1st set of FEEDERS?

[java_to_tm1]

beek,
Wim Gielis was suggesting that you re-do the structure of your measures dimension so that it looks somewhat like this

D
|- E
|- A
|- B
|- C

Assuming you give each child element a weight of 1, this structure takes care of the calculations without the requirement for feeding.

If you have simple addition / subtraction to do, using weights and a hierarchy in is cheaper (memory consumption-wise) and faster (performance-wise) than rules and feeders.

~Java_to_tm1

[beek]

I see. Got it. Thank you Wim & Java

[Wim Gielis]

That's what I meant, java_to_tm1