Flink wrappers and helper libraries should integrate with Table API. We should allow injection of UDFs (ideally cross language).
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Ottomata | T324689 [EPIC] Streaming and event driven Python services | |||
| Declined | None | T324953 [NEEDS GROOMING] Integrate Flink Table API in eventutils-python |
Event Timeline
Comment Actions
Spent a little bit of time thinking about this today, and I'm not sure how it will work. You've been able to workaround some of the annoying input and output type issues by using the Python DataStream API, working with dicts only. But, from our experiments with udfs, udfs need the explicit result_type, which has to be a Flink DataType. I couldn't figure out how to pass the full row to the UDF. It feels like there should be a way?
I guess this would be doing like you did with the DataStream row.to_dict(recursive=True) stuff, but wrapping the user provided func in a udf that specifies the result_type as the same as the output sink?
Comment Actions
I'm going to be bold and decline this one. If/when we decide to really really support Flink Table API, we can revisit or reopen.