Apache Beam Dataflow运行报错：NameError: name 'beam' is not defined

云上老周

2026-6-27

NameError: name 'beam' is not defined when creating beam.Row in Apache Beam Dataflow

Below are the most common causes and corresponding fixes for your issue:

1. Import is Shadowed by a Local Variable

Check if any local variable, function parameter, or class in your code shares the name beam, which would overwrite the imported Apache Beam module. For example:

import apache_beam as beam

def transform_data(beam):  # Parameter shadows the imported module
    return beam.Row(id=1, value=element)  # Throws NameError here

Fix: Rename the conflicting local identifier (e.g., change beam to input_element).

2. Import Not Accessible in Serialized Code

When using beam.Row inside a DoFn, lambda, or other code that gets serialized for Dataflow workers, ensure the apache_beam as beam import is present in the same module where the code is defined. If your DoFn lives in a separate file, add the import there too—workers need access to the import at runtime to resolve beam.Row.

3. Outdated Apache Beam Version

The beam.Row class was introduced in Apache Beam 2.16.0. If you're using an older version, the class won't exist, and you might see this error. Fix: Upgrade to a recent version:

pip install --upgrade apache-beam[gcp]

4. Lambda Serialization Issues

Lambdas can sometimes fail to capture imports properly during serialization. If you're using a lambda to create beam.Row, replace it with a named DoFn instead:
Before:

pipeline | beam.Map(lambda elem: beam.Row(key=elem[0], count=elem[1]))

After:

class CreateAggregatedRow(beam.DoFn):
    def process(self, element):
        key, count = element
        return [beam.Row(key=key, count=count)]

pipeline | beam.ParDo(CreateAggregatedRow())

5. Incorrect Environment Setup

Ensure you're running your code in the virtual environment where Apache Beam is installed. If you have multiple Python environments, the import may not be available in the one you're using. Verify with: