Thanks Danny. In current enumerable implementation, explicit watermark is not needed because data is already known so we can think watermark already advances to positive infinity, and every RelNode can close all windows immediately. Thus [1] make tumbling works for batch workload.
I created [2] to track work related to watermark. After [1] is done, I can start to think about watermark in calcite. [1]: https://github.com/apache/calcite/pull/1587 [2]: https://issues.apache.org/jira/browse/CALCITE-3509 -Rui On Fri, Nov 15, 2019 at 9:40 PM Danny Chan <[email protected]> wrote: > No, there is no watermark syntax or implementation in Calcite, watermark > is needed for the google data stream model, and was implemented well in > Apache Flink, you can take a reference.(The syntax, the watermark > strategies translation). > > > Best, > Danny Chan > 在 2019年11月16日 +0800 AM1:44，Rui Wang <[email protected]>，写道： > > Is watermark is ever discussed in calcite? In order to make enumerable > > implementation useful, we will need watermark support. Watermark will be > > just a value that represents a time and keep advancing to +inf. It > > generates from sources and propagate to all RelNodes. Emit syntax will > > require such tool to determine when window will close and then emit data. > > > > > > -Rui > > > > On Tue, Aug 20, 2019 at 11:51 PM Danny Chan <[email protected]> > wrote: > > > > > Sorry, I have no permission to grant you the JIRA contributor > permission, > > > you can send another mail asking for this. > > > > > > Best, > > > Danny Chan > > > 在 2019年8月21日 +0800 AM11:58，Rui Wang <[email protected]>，写道： > > > > Thanks Danny. I have filed a sub-task under it: CALCITE-3272. > > > > > > > > Can you grant me contributor permission so I can assign it to myself? > > > > > > > > > > > > -Rui > > > > > > > > On Tue, Aug 20, 2019 at 8:24 PM Danny Chan <[email protected]> > wrote: > > > > > > > > > Thanks for bring up this topic, Rui Wang ~ > > > > > > > > > > I have fired an Umbrella issue [1] about this topic where we can > create > > > > > sub-tasks. > > > > > > > > > > [1] https://issues.apache.org/jira/browse/CALCITE-3271 > > > > > > > > > > Best, > > > > > Danny Chan > > > > > 在 2019年8月20日 +0800 AM9:59，Rui Wang <[email protected]>，写道： > > > > > > Hi Calcite community, > > > > > > > > > > > > I have seen [1] is published, which proposes TVF style > > > TUMBLE/HOP/SESSION > > > > > > windowing and EMIT syntax for materialization control. > > > > > > > > > > > > I tried to search dev@calcite and Calcite JIRA but didn't find > those > > > > > > support above. So is Calcite already support TVF windowing and > EMIT > > > > > syntax > > > > > > somehow, or it's a future work? > > > > > > > > > > > > > > > > > > > > > > > > [1]: https://arxiv.org/pdf/1905.12133.pdf > > > > > > > > > > > > > > > > > > -Rui > > > > > > > > >
