Stephen Parente

I don't think it's been implemented yet, so I'll classify this as a feature and try to get it into the next release. Thanks for the report.

Added it here: d8ddd8e4d3fabf94b123261b41fbf8ebc126e86d

I definitely see the appeal and use case for using Kafka this way. There is even a utility in the Java client called "KafkaStreams" which are meant to transform data...

Anything I can do to make consuming and producing data this way easier is something I'd be happy to help with. I think one of our Kafka use-cases is pretty...

If I'm understanding correctly it looks like this may make it into `librdkafka`: https://github.com/edenhill/librdkafka/pull/1191

You need to handle the potential error case in `consume` through its callback. There is a chance, especially directly after a consume, that the consume will fail because a timeout...

Hi there, Firstly, `librdkafka` internally manages connection state, so it's okay to get disconnected as long as it is recoverable. Are you talking about a situation where you get disconnected...

Is this part of a test suite you're writing or does this actually happen on occasion? I'm currently using `KafkaConsumer::commit` which is a blocking call. Perhaps if I change it...

I haven't forgotten about this! There are a few issues related to disconnects that I'm trying to clean up in case they fix this issue. Then I will investigate it...

I think query watermark offset will do this for you, and that was added in a recent commit?