使用ValueProvider格式化数据流中的BigQuery

Question

pipeline_options = PipelineOptions()
p = beam.Pipeline(options=pipeline_options)

lines = (p
             | 'read_big_query' >> beam.io.Read(beam.io.BigQuerySource(query=QUERY, use_standard_sql=True))
             | "doing stuff" >> beam.Map(do_some_stuff)
             )

class TemplateOption(PipelineOptions):
    @classmethod
    def _add_argparse_args(cls, parser):
        parser.add_value_provider_argument('--template_do_stuff_param',
                                           default=45,
                                           type=int)
class MyDoStuffFn(beam.DoFn):
    def __init__(self, template_do_stuff_param):
      self.template_do_stuff_param = template_do_stuff_param

    def process(self, *_):
      yield do_some_stuff(self.template_do_stuff_param.get())


pipeline_options = PipelineOptions()
p = beam.Pipeline(options=pipeline_options)
template_option = pipeline_options.view_as(TemplateOption)

lines = (p
             | 'read_big_query' >> beam.io.Read(beam.io.BigQuerySource(query=QUERY),
                                                                                     use_standard_sql=True))
             | "doing stuff" >> beam.ParDo(MyDoStuffFn(template_option.template_do_stuff_param))
             )

class TemplateOption(PipelineOptions):
    @classmethod
    def _add_argparse_args(cls, parser):
        parser.add_value_provider_argument('--template_nb_users',
                                           default=100,
                                           type=int)
        parser.add_value_provider_argument('--template_do_stuff_param',
                                           default=45,
                                           type=int)
class MyDoStuffFn(beam.DoFn):
    def __init__(self, template_do_stuff_param):
      self.template_do_stuff_param = template_do_stuff_param

    def process(self, *_):
      yield do_some_stuff(self.template_do_stuff_param.get())


pipeline_options = PipelineOptions()
p = beam.Pipeline(options=pipeline_options)
template_option = pipeline_options.view_as(TemplateOption)

lines = (p
             | 'read_big_query' >> beam.io.Read(beam.io.BigQuerySource(query=QUERY.format(nb_users=template_option.template_nb_users.get()),
                                                                                     use_standard_sql=True))
             | "doing stuff" >> beam.ParDo(MyDoStuffFn(template_option.template_do_stuff_param))
             )

...这不起作用，因为我在管道执行之前调用get（）。到目前为止，我还没有将为do_some_stuff函数所做的调整到“read”行

任何关于如何进行的建议或解决方案都将不胜感激。谢了！