Details
-
Improvement
-
Status: Open
-
Not a Priority
-
Resolution: Unresolved
-
None
-
None
Description
It would be better if fig/program_dataflow.svg used a ProcessWindowFunction, rather than a WindowFunction.
It also uses a BucketingSink, which sets a bad example.
Note that this is not a trivial edit, since it doesn't work to simply replace new BucketingSink with new StreamingFileSink. Something like this would be better:
final StreamingFileSink<GenericRecord> sink = StreamingFileSink
.forBulkFormat(...)
.build();
stats.addSink(sink);
Note: This figure is only used once, in the Training Overview page.