Depending on the runner that may be an error, for example for Dataflow that would give you an error when you tried to build your pipeline since Dataflow does not yet support having both bounded and unbounded PCollections within a single pipeline.
There are usecases where having both is valid, for example you may want to initialize your streaming pipeline with some static data to use as a side input into a DoFn which works over an unbounded PCollection. On Thu, Aug 11, 2016 at 12:54 PM, amir bahmanyari <[email protected]> wrote: > Hi Colleagues, > Is *bounded *KafkaIO().withMaxNumRecords(finite-number) essentially the > same as TextIO().read_from_a_finite_records data file? > In neither case we are Streaming, right? No on-going streaming data mode > in both since "reading" will halt at some point by design: *bounded*. > What difference is setting options.setStreaming(*true*); to true making > here? > Thanks >
