Re: Picking up the thread on emit-on-change and at-least-once

2025-03-01 Thread Steven Schlansker
Thank you for your response Matthias. > On Feb 26, 2025, at 8:32 PM, Matthias J. Sax wrote: > > Hey, > > the problem you describe it not totally fundamental, but there is no > workaround in KS currently... So right now, only EOS can fix it. > > In the end, what you would need is a way to firs

Re: Picking up the thread on emit-on-change and at-least-once

2025-02-28 Thread Matthias J. Sax
We did split out KIP-1035 (already accepted), that is a requirement/blocker for KIP-892. Progress is slow, but there is WIP to implement KIP-1035. After it's done, we can continue/pick-up KIP-892 implementation. -Matthias On 2/28/25 11:20 AM, Steven Schlansker wrote: Thank you for your res

Re: Picking up the thread on emit-on-change and at-least-once

2025-02-26 Thread Matthias J. Sax
Hey, the problem you describe it not totally fundamental, but there is no workaround in KS currently... So right now, only EOS can fix it. In the end, what you would need is a way to first write to the output topic (and flush the write), before you update the state store. Using the PAPI you

Picking up the thread on emit-on-change and at-least-once

2025-02-25 Thread Steven Schlansker
Hi kafka-users, We are implementing a Kafka Streams app that computes various streaming statistics over a corpus of data stored in Kafka topics. While some aggregates update often, others like 'min', 'max', or histogram buckets could have relatively few distinct updates relative to the input data.