This document provides additional information regarding the Beam Transform service. For primary document regarding the Transform serviceThis guide includes a list of transforms available in the Docker Composed based transform service provided with Beam. For more details regarding the Transform Service, please see the Beam programming guide.

Transforms included in the Transform service

This includes following informationsection includes a list of Beam transforms currently included in the Transform service. Following information regarding the transforms are provided.

• Transform - a short description about the transform.
• SDK implemented - SDK where the transform is implemented in.
• Unique ID - a unique ID that can be used to access the transform. There are three types of unique IDs supported by the transform service.
  • schema-aware: a schema-aware transform ID (only available for transforms implemented using Java). We recommend using schema-aware transforms over direct URN based transforms whenever available.
  • urn: a URN that is used to register the transform in an expansion service
  • python-name: Fully qualified class name or a callable (only available for transforms implemented using Python).

Note that Beam transform service can be used to expand all portable transforms provided with Python SDK, hence what's mentioned here for Python transforms is not a comprehensive list.

Upgrade transforms without upgrading the pipeline

Transform service allows Beam pipeline authors to upgrade specific transforms within their pipelines to a newer Beam version without upgrading the full pipeline. Please see the Beam programming guide for more details regarding this feature and supported SDKs and Beam versions.

Here we provide an example that upgrades a Beam Java BigQuery transforms of a pipeline to a newer Beam version without upgrading the pipeline.

• Setup the environment according to the Beam Java quickstart guide.
• Install Docker if not already available in the system.
• Checkout the Beam examples Maven archetype for the relevant Beam version.
  
  

export BEAM_VERSION=<Beam version> # Needs to be Beam 2.53.0 or later.

mvn archetype:generate \
    -DarchetypeGroupId=org.apache.beam \
    -DarchetypeArtifactId=beam-sdks-java-maven-archetypes-examples \
    -DarchetypeVersion=$BEAM_VERSION \
    -DgroupId=org.example \
    -DartifactId=beam-transform-upgrade \
    -Dversion="0.1" \
    -Dpackage=org.apache.beam.examples \
    -DinteractiveMode=false

cd beam-transform-upgrade

• Add the Beam examples Java dependency to the <dependencies> section of the Maven pom.xml file.
  
  

  Code Block
  <dependency> <groupId>org.apache.beam</groupId> <artifactId>beam-examples-java</artifactId> <version>${beam.version}</version> </dependency>

  
  

• Execute the pipeline using a portable Beam runner. Following example uses Dataflow Runner v2.
  
  

export GCP_PROJECT=<GCP project>
export GCP_BUCKET=<GCP bucket>
export GCP_REGION=<GCP region>
export OUTPUT_BIGQUERY_TABLE=<A BigQuery table to write the output to>
export TRANSFORM_BEAM_VERSION=<Beam version to upgrade transforms to> # Needs to be Beam 2.53.0 or later.

mvn compile exec:java -Dexec.mainClass=org.apache.beam.examples.cookbook.BigQueryTornadoes -Dexec.args="--runner=DataflowRunner --project=$GCP_PROJECT \
                 --region=$GCP_REGION \
                 --tempLocation=gs://$GCP_BUCKET/transform-upgrade/tmp \
                 --experiments=use_runner_v2 --output=$OUTPUT_BIGQUERY_TABLE \
                 --transformsToOverride=beam:transform:org.apache.beam:bigquery_read:v1,beam:transform:org.apache.beam:bigquery_write:v1 \
                 --transformServiceBeamVersion=$TRANSFORM_BEAM_VERSION" -Pdataflow-runner