Hi Xiaowei,
I like the idea to reuse a partitioning and thus saving a shuffle
operation. It would be great if we could fail at runtime in case the
partitioning changed somehow. That way a logical user failure won't go
unnoticed.
Would it make sense to name the method partitionedByKey(...) because
After we do any interesting operations (e.g. reduce) on KeyedStream, the
result becomes DataStream. In a lot of cases, the output still has the same
or compatible keys with the KeyedStream (logically). But to do further
operations on these keys, we are forced to use keyby again. This works
semantic
Xiaowei Jiang created FLINK-4855:
Summary: Add partitionedKeyBy to DataStream
Key: FLINK-4855
URL: https://issues.apache.org/jira/browse/FLINK-4855
Project: Flink
Issue Type: Improvement
