[BEAM-12955] Add support for inferring Beam Schemas from Python protobuf types

• Type: Improvement
• Status: Open
• Priority: P3
• Resolution: Unresolved
• Component/s: sdk-py-core
• Assignee: Unassigned
• Reporter: Brian Hulette
• Created: 24/Sep/21 18:51
• Updated: 04/Jun/22 21:46

Description

Just as we can infer a Beam Schema from a NamedTuple type (code), we should have support for inferring a schema from a protobuf-generated Python type.

This should integrate well with the rest of the schema infrastructure. For example it should be possible to use schema-aware transforms like SqlTransform, Select, or beam.dataframe.convert.to_dataframe on a PCollection that is annotated with a protobuf type. For example (using the addressbook_pb2 example from the tutorial):

import adressbook_pb2

import apache_beam as beam
from apache_beam.dataframe.convert import to_dataframe

pc = (input_pc | beam.Map(create_person).with_output_type(addressbook_pb2.Person))

df = to_dataframe(pc) # deferred dataframe with fields id, name, email, ...

# OR

pc | beam.transforms.SqlTransform("SELECT name WHERE email = '[e-mail]' FROM PCOLLECTION")

Issue Links

• relates to: BEAM-8732 Add support for mapping additional structured types to Python Schemas
• split to: BEAM-13150 Integrate TFRecord/tf.train.Example with Beam Schemas and the DataFrame API