Github user fhueske commented on the issue:
https://github.com/apache/flink/pull/4183
Hi @wuchong, that's an interesting idea but I think it has the drawback
that it might add latency.
A timestamp extractor is only called with records and doesn't see
watermarks. Since an operator might emit multiple records for the same
timestamp, a timestamp extractor would always have to emit a watermark of last
timestamp - 1 (we can be sure that the records are emitted in timestamp order)
because it does not know which record is the last for a timestamp. So, we would
add a latency of one window length (until the next window is processed).
Custom operators are a low level interface but it shouldn't be too hard to
implement one that holds watermarks back.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
