Github user fhueske commented on the issue:
https://github.com/apache/flink/pull/3675
Hi @haohui, I updated the PR and refactored the rule into an abstract class
that is extended by a rule for datastream and another one for dataset.
**Please note**: I also removed support for `GROUP BY FLOOR/CEIL` on
streams. My motivation is the following:
- it can be expressed with `TUMBLE` (which is even more expressive)
- We will add support for "non-windowed" aggregations (see PR #3646 by
@shaoxuan-wang ) which can include the `GROUP BY FLOOR/CEIL` case. These
aggregations will compute the same result but behave differently at runtime.
Instead of collecting data and emitting a final result, they will be
incrementally computed and send out more updates. I think the behavior of
system is more predictable if we only translate Calcite's group window
functions into DataStream group windows and all other grouped aggregates into
non-windowed aggregates. We target the non-windowed aggregates for the next
release.
- It will make the translation easier. The pattern matching of `FLOOR/CEIL`
seems a bit fragile to me. I like Calcite's group window functions more.
Please let me know what you think about this change and if you have
concerns.
Best, Fabian
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
