HI, Wenjin We believe that the interfaces proposed by FLIP are more flexible and offer greater freedom to the user. For example, users can decide for themselves how to process data and in what form it is stored in the Window. In cases where pre-aggregation is required, the user can choose to store only the aggregated results in the Window, instead of all input data. In addition, we have provided a default implementation that allows users to use the Window API proposed in FLIP in a similar way to DataStream V1, thus providing greater flexibility to users. Best, Xu Huang
wenjin <wenjin...@gmail.com> 于2025年1月10日周五 10:41写道： > Hi, Xu > > I've noticed that the Window API provided in the FLIP differs > significantly from the one on DataStream V1. For example, it introduces the > definition of the Window lifecycle and allows users to decide whether to > store input records within the Window. > What are the advantages of designing the interface this way? > > Best, > Wenjin Xie > > On 2025/01/07 16:10:59 Xu Huang wrote: > > Hi Devs, > > > > > > Weijie Guo and I would like to initiate a discussion about FLIP-501: > > Support Window Extension in DataStream V2 API [1]. > > > > Window are at the heart of processing infinite streams. Window split the > > stream into "buckets" of finite size, over which we can apply > computations. > > For example, time windows are used to aggregate data within specific time > > intervals. > > > > The goal of this FLIP is to provide Window extension for DataStream V2, > > thereby enhancing its data processing capabilities and usability. > > > > For more details, please refer to FLIP-501 [1]. We look forward to your > > feedback. > > > > > > Best, > > > > Xu Huang > > > > > > [1] https://cwiki.apache.org/confluence/x/z4kgF > >
