SAP Integration Suite - Message Mapping - addContextChange in UDF is ignored

[PAST]

Hi expterts,

I am trying to merge several single values into one structure using an UDF with the addContextChange function.
Unfortunately it does not work as expected. 

My source structure look similar to this all with occurence 0..1:
[...]
costCenter1 
costCenter2
costCenter3
percentage1
percantage2
percentage3
[...]

Inside the target message I would like to group them in a structure like this:
[...]
costDistribution (0..*)
- costCenter (1..1)
- percentage (1..1)
[...]

To map the values I am using an UDF function:
def void costDistribution(String[] costCenter1, String[] costCenter2, String[] costCenter3, String[] percentage1, String[] percentage2, String[] percentage3, Output parentNode, Output costCenters, Output percentages, MappingContext context){ ... }

For test purposes I tried this with static values to get a feeling for the context change function and tried this:
parentNode.addValue("004025235")
costCenters.addValue("004025235")
percentages.addValue("80")

parentNode.addContextChange()
costCenters.addContextChange()
percentages.addContextChange()

parentNode.addValue("004025240")
costCenters.addValue("004025240")
percentages.addValue("20")

My expection in the target would be to have several costDistribution nodes with the corresponding values like this:
[...]
costDistribution
- costCenter 004025235
- percentage 80
costDistribution
- costCenter004025240
- percentage 20
[...]

In Reality the first block is generated. May someone explain me how I need to use the contextChange function to achieve the wanted structure?
Thank you all in advance.

Best regards

[Ryan-Crosby]

There are ways to accomplish this without a UDF at all with adding additional occurrences and mapping 1 -> 1st index, 2 -> 2nd index, and so on.

[PAST]

You mean adding additional occurences inside the target structure? Since it is possible that I receive between 0 and six costCenter data sets I would appreciate to do thie dynamically using the UDF and at least in my understanding it should work or is this not possible at all?

[Ryan-Crosby]

Why write code when not necessary at all? The index approach will still handle the case when the source node does not exist by not producing target output, and it doesn't require maintaining any code directly. The goal should not be to write code but rather solve a problem, which in this case can be done with zero lines of code.

[PAST]

Hi Ryan, I do agree that this would work in my case. But in general I would still like to understand what I did wrong in my coding approach, to understand how it could be used in other cases in the future.