Martin Krasser

@lowchinwei I first need to check whether a more recent version of the async producer API supports atomic writes of user-defined batches (see also #20). Although this is independent of...

This would require a `persistenceId` -> (topic, partition) mapping, for example, using a naming convention for `persistenceId`s such as `my_topic_0`, `my_topic_1`and so on. Shouldn't be a big deal to implement....

I see, thanks for clarifying. I'm fine with updating the API docs of `EventLog` describing the degrees of freedom for implementations and then merge your PR.

@odd it would be great if you could do that. I actually meant: *I'm fine if you update the API docs ..* 😃. Thanks in advance!

Thanks for sharing your ideas. The main purpose of the low-level snapshot API is to provide a foundation for implementing what you proposed (among other possible solution). I'd like to...

@2beaucoup the low level API will provide sources that associate (i.e. tag) emitted snapshots with offsets (= version numbers). Do you see any issues?

Here's the idea. Given a `Snapshot` type, an `eventSource` starting `fromSequenceNr` and an `eventHandler`: ```scala import com.github.krasserm.ases.Durable import com.github.krasserm.ases.EventSourcing.EventHandler trait Snapshot[S] { def state: S def sequenceNr: Long // state...

@t3hnar it will be. The above only explains the concept i.e. no need to cover implementation details.

Unfortunately, the Keras backend API doesn't provide probability distributions needed here such as `tf.distributions.Normal`. I could implement them low-level but this wouldn't really scale to more complex distributions. OTOH, `tf.random.normal`,...

Thank you very much Stefan for your kind feedback. I'll leave this ticket open if you don't mind :wink: