Skip to main content

19. Multi-Output Tagged Branch RoutingMedium

Custom DoFn & Advanced⏱️ ~18 mins

19. Multi-Output Tagged Branch Routing

Enterprise Architecture Context

In production stream-processing architectures (Google Cloud Dataflow / Flink), pipeline stages must handle parallel transformations without data loss, managing schema mutations and aggregations across distributed worker workers.

Problem Statement

### Business Context Real-time routing pipelines split high-priority alerts from standard informational events into isolated destination queues in a single compute pass. ### Problem Statement Implement a `DoFn` subclass `SplitOddsFn` that routes **even integers** to the main output, and **odd integers** to a tagged side output `"odds"` using `beam.pvalue.TaggedOutput`.

Key Learning Objectives

  • Understand distributed Apache Beam execution DAG stages and pipeline lifecycle.
  • Apply idiomatic functional Python transforms using the pipe operator |.
  • Ensure data consistency and idempotency across distributed stream workers.

Sample Data Fixtures

Sample Example 1
Input Stream:
[1, 2, 3, 4]
Expected Output:
Evens: [2, 4], Odds: [1, 3]
Sample Example 2
Input Stream:
[10, 20]
Expected Output:
Evens: [10, 20], Odds: []
Topics:#ParDo#Tagged Outputs
Support