Hello, the application I'm working on reads data from kafka topics that have consistent keys. If I understand correctly, there's always only a single Flink task that would read from a single partition, this guarantees that records from the same partition always come in order. Now in tests, I always used

env.fromCollection(...)

to build test data, but this does not have any source task topic/partition awareness, is it possible to build such a source function where I could provide keys to simulate kafka's partition awareness? I looked at the RichParallelSourceFunction but couldn't find a way to specify partitions a record should go to.