Ways of using value provider parameter in Python Apache Beam

Question

Right now I'm just able to grab the RunTime value inside a class using a ParDo, is there another way to get to use the runtime parameter like in my functions?

This is the code I got right now:

class UserOptions(PipelineOptions):
    @classmethod
    def _add_argparse_args(cls, parser):
        parser.add_value_provider_argument('--firestore_document',default='')

def run(argv=None):

    parser = argparse.ArgumentParser()

    pipeline_options = PipelineOptions()

    user_options = pipeline_options.view_as(UserOptions)

    pipeline_options.view_as(SetupOptions).save_main_session = True

    with beam.Pipeline(options=pipeline_options) as p:

        rows = (p 
        | 'Create inputs' >> beam.Create(['']) 
        | 'Call Firestore' >> beam.ParDo(
                CallFirestore(user_options.firestore_document)) 
        | 'Read DB2' >> beam.Map(ReadDB2))

I want the user_options.firestore_document to be usable in other functions without having to do a ParDo

Darg

Hola! What do you mean by RunTime value? Do you mean this user_options.firestore_document variable? – Chatty

What do you mean by user_options.firestore_document to be usable in other functions without having to do a ParDo ? – Crucify

@Chatty Hola! Yes, that is my RunTime value, that value will only be filled when the template is executed and not when the template is build, right now if I try to read that value inside any function that isn't a ParDo function I will get errors about it being empty – Darg

@guillaumeblaquiere What I mean is that if I try to use that variable in any function, which is only passed when executing the template and not when creating it, the template won't generate and I will start getting errors about my variable being empty – Darg

What other functions are you thinking of? Can you give an example of these other functions? This should work for any pardo - so in your code, it should work fine. – Chatty

@Chatty Yes it works inside the ParDo, but for example if we want to use it inside a Create how can we do it? Is it possible to use a RunTimeValue Provider outside a ParDo? What I am trying to achieve is to have one template and from the runtime parameters read from one table or another, and then write to one table or another. – Darg

It can only be used from a ParDo - not from Create, but a ParDo should be enough for what you need. Are you trying to use it within CallFirestore? – Chatty

@Chatty Oh so those can only be used inside a ParDo :( I was able to achieve what I was trying to do with the ParDo, but thought there was another way to do it, Thanks! – Darg

I'll add an answer to see if it helps others. – Chatty

This is my exact point. I would like to use the value of the argument in functions inside dataflow which isn't a ParDo.Did you get any luck on this? – Nils

Answer 1

The only way in which you can use value providers are in ParDos, and Combines. It is not possible to pass a value provider in a create, but you can define a DoFn that returns the value provider you pass to it in the constructor:

class OutputValueProviderFn(beam.DoFn):
  def __init__(self, vp):
    self.vp = vp

  def process(self, unused_elm):
    yield self.vp.get()

And in your pipeline, you would do the following:

user_options = pipeline_options.view_as(UserOptions)

with beam.Pipeline(options=pipeline_options) as p:
  my_value_provided_pcoll = (
      p
      | beam.Create([None])
      | beam.ParDo(OutputValueProviderFn(user_options.firestore_document))

That way you wouldn't use it in a Create, as it's not possible, but you could still get it in a PCollection.

Chatty

Hey @pablo, that was really helpful to me today, thanks!. But where was I supposed to find that information? Is it in the docs? I tried to find it but couldn't. Any recommendations for sources? – Beggs