Github user tdas commented on the issue:
https://github.com/apache/spark/pull/21673
Thanks @arunmahadevan for making this PR. However, I dont like the solution
of adding another field as a workaround thus making the control flow harder to
reason about. I think the fundamental problem is that the original design of
the ProgressReport that sees all the internal details of StreamExecution (e.g.
availableOffsets and committedOffsets) and its very reason what informatio is
read when. I want to refactor this a little bit towards improving this
underlying problem. I am working on a PR myself for that.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
