Andrei Zhabinski

> I made the suggested change. But my code is in a Pluto notebook. Do you think it will rebuild automatically if I restart the notebook? Spark is not available...

As I mentioned above, I don't have proper environment to test it myself, but your code looks correct to me. If you could post the error you encounter, I may...

I have an idea what may have gone wrong. Could you please try `streaming-fixes` branch, i.e.: ```julia ] add Spark#streaming-fixes ``` and then run your code again?

Can you show complete Julia and Scala code? Spark.jl is nothing more than a shallow wrapper around Spark API. For instance, the `format` function is [implemented](https://github.com/dfdx/Spark.jl/blob/streaming-fixes/src/streaming.jl#L67-L70) as: ```julia function format(writer::DataStreamWriter,...

You use latest offset in the Julia version and earliest in the Scala version. Is it intended? How many messages do you add to the topic during the test?

[Here](https://github.com/dfdx/Spark.jl/blob/streaming-fixes/Dockerfile) I'm trying to set up a dev environment to check this issue, but today I've run out of time. I'll try to allocate more time in the coming days.

By the way, what version of Java are you running?

Yes, it's a known issue and a notable historical artifact. Initially, Yota was designed to work with its custom rules that required only adding new operations onto the tape and...

True! I experimented with a single-pass AD for some time, but eventually decided to keep them separate because of corner cases. For example, Julia has a pretty special way to...

I'd say there are 2 main steps that need to be done before we can freely interchange models between Julia and Python: 1. Add missing operations. 2. Implement full-featured conversion...