Thanks for the confirmation :) Regards, Alexis.
On Sun, 9 Oct 2022, 10:37 Hangxiang Yu, <master...@gmail.com> wrote: > Hi, Alexis. > I think you are right. It also applies for a global window with a custom > trigger. > If you apply a ReduceFunction or AggregateFunction, the window state size > usually is smaller than applying ProcessWindowFunction due to the > aggregated value. It also works for global windows. > Of course, the state size of a global window also depends on how you > implement your trigger. > BTW, we often use TTL to reduce the state size of the global window. > Hope these can help you. > > > On Sat, Oct 8, 2022 at 4:49 AM Alexis Sarda-Espinosa < > sarda.espin...@gmail.com> wrote: > >> Hello, >> >> I found an SO thread that clarifies some details of window state size >> [1]. I would just like to confirm that this also applies when using a >> global window with a custom trigger. >> >> The reason I ask is that the TriggerResult API is meant to cover all >> supported scenarios, so FIRE vs FIRE_AND_PURGE is relevant, for example, >> for a ProcessWindowFunction that holds all input records until it fires. >> However, I assume there would be no distinction if I use a >> (Rich)AggregateFunction, regardless of window type (global vs timed), but >> I'd like to be sure. >> >> Regards, >> Alexis. >> >> [1] >> https://stackoverflow.com/questions/55247668/flink-window-state-size-and-state-management >> >> > > -- > Best, > Hangxiang. >
