1. Looks good. Thanks for the explanation. We can add some comments in the
javadoc that this is normally not needed for app programmers, but only for
the framework developers.
2. Thanks!
On Mon, Dec 5, 2016 at 2:21 AM, Damian Guy wrote:
> Hi Guozhang,
>
> Thanks for the feedback.
>
> 1. If we w
Hi Guozhang,
Thanks for the feedback.
1. If we want developers to be able to provide custom SessionStore
implementations then we need this method to be on the public interface. We
are currently allowing this for other operations, so it seems to make
sense, for the sake of consistency, to allow a
Thanks for the KIP Damian, it looks great to me overall. A couple of minor
comments:
1. findSessionsToMerge(): does this need to be a public interface? Does
users ever need to call it from the processor API or even DSL?
2. SessionMerger: I am wondering if we could make it more general by
renaming
Thanks Matthias.
1) Yes good suggestion will update.
2) As it is consistent with Aggregator and a developer may want to use the
key. So why not?
3) Thanks. I'll update the KIP
Cheers,
Damian
On Tue, 29 Nov 2016 at 23:47 Matthias J. Sax wrote:
> Very nice KIP!
>
>
> Couple of comments:
>
> 1) C
Very nice KIP!
Couple of comments:
1) Current window API is as follows:
> JoinWindows.of(timeDifference).before(timeDifference).after(timeDifference)
> TimeWindows.of(size).advanceBy(interval)
> UnlimitedWindow.of().startOn(start)
To align with this scheme, I think it would be better to use th
