Hi all, I’m Nikita Grover (nikitagrover19), currently working on ordering key support for Python PubSubIO (Issue #36201). I implemented the XLang approach suggested in the thread and opened a PR: https://github.com/apache/beam/pull/37345.
I’m applying for the Streaming Transforms GSoC project and had a few quick questions: 1. For Read.expand(), would you prefer handling both Bounded and Unbounded sources in one entry point, or keeping a separate transform for Unbounded? 2. For Watch, should it live alongside periodicsequence.py, or as a separate module? 3. For defer_remainder, is the Flink checkpoint growth issue (#27648) something new implementations should work around? Would really appreciate any guidance. I’d also be happy to share my current proposal if that’s helpful. Thanks, Nikita
