Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/5236
If we want to implement something like that, we should tap into what the
data structures give us anyways, so we don't add additional overhead. Its a bit
tricky to do that, though, because
Github user bowenli86 commented on the issue:
https://github.com/apache/flink/pull/5236
@aljoscha @StephanEwen This metadata of timers helps users to understand
their timer usage and write logic in code to handle it. I think it's really
helpful and convenient to provide the data
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/5236
I am with Aljoscha here - the cost of making this information available in
the general case seems pretty high.
What is the use case that needs that feature - can that be implemented in a
Github user aljoscha commented on the issue:
https://github.com/apache/flink/pull/5236
I'm sorry because my earlier comment on the Jira issue wasn't very clear.
We might have to take a step back here and think about what we want to achieve.
When I mentioned that an extra bit
Github user bowenli86 commented on the issue:
https://github.com/apache/flink/pull/5236
@GJL Thanks for the feedback, will address them. I thought about this
overnight, and realized this is actually not all parts I want to build. What I
want is number of timers **per key** in a
