Hi Timo, Sergey, Yunfeng, Jingsong, Ron, XuYang, Shengkai, and others, The FLIP has been updated based on the feedback from the DISCUSS thread:
- Switched to Calcite lambda syntax - Added an optional config parameter as an extension point - Clarified the implementation approach The updated proposal is available on the FLIP page. If there are no further concerns, I would appreciate your feedback and vote. Thanks for your time and review. Best regards, Feat Zhang On Thu, May 28, 2026 at 9:34 PM FeatZhang <[email protected]> wrote: > Hi Timo, > > Thanks for the follow-up and for the last remarks in the DISCUSS thread. > > I have updated the FLIP accordingly: > > 1. The public SQL surface now uses Calcite lambda expression syntax for the > watermark strategy, e.g. > > APPLY_WATERMARK( > orders, > DESCRIPTOR(order_time), > (order_time) -> order_time - INTERVAL '5' SECOND) > > 2. I added the optional function-level config parameter as an extension > point: > > config => MAP<STRING, STRING> > > The default is an empty config. This FLIP does not define concrete > runtime > config keys yet, and unsupported keys should fail validation rather > than be > silently ignored. > > 3. I also clarified that the implementation does not require Calcite parser > changes. The initial implementation can use Calcite's SqlOperator stack, > while still lowering to the dedicated LogicalWatermarkAssigner / > StreamExecWatermarkAssigner path as described in the FLIP. > > The FLIP document has been updated to reflect these changes. > > With this update, I would like to continue the VOTE thread. Please continue > voting. > > Thanks again for the review and suggestions. > > Best regards, > FeatZhang > > On Tue, May 26, 2026 at 4:21 PM Timo Walther <[email protected]> wrote: > >> Hi FeatZhang, >> >> sorry for the late reply but I had some last remarks in the DISCUSS >> thread before I cast my vote. >> >> Thanks, >> Timo >> >> On 25.05.26 14:25, 熊饶饶 wrote: >> > +1 >> > >> > FeatZhang <[email protected]> 于2026年5月25日周一 18:46写道： >> > >> >> Hi everyone, I would like to start the vote to approve FLIP-581: >> Support >> >> Flexible Watermark Assignment via Built-in Function [1]. The discussion >> >> thread [2] has been active for a while and received feedback, and the >> >> proposal is ready for a formal vote. The vote will be open for at >> least 72 >> >> hours. [1] >> >> >> >> >> https://cwiki.apache.org/confluence/display/FLINK/FLIP-581%3A+Support+Flexible+Watermark+Assignment+via+Built-in+Function >> >> >> >> < >> >> >> https://cwiki.apache.org/confluence/display/FLINK/FLIP-581%3A+Support+Flexible+Watermark+Assignment+via+Built-in+Function >> >>> >> >> [2] https://lists.apache.org/thread/oonylk4h8dnsom40g8rr5k52zf3tz64v >> >> Best regards, Feat Zhang >> >> >> > >> >>
