apache_beam.runners.runner module

PipelineRunner, an abstract base runner object.

class apache_beam.runners.runner.PipelineRunner

Bases: object

A runner of a pipeline object.

The base runner provides a run() method for visiting every node in the pipeline’s DAG and executing the transforms computing the PValue in the node.

A custom runner will typically provide implementations for some of the transform methods (ParDo, GroupByKey, Create, etc.). It may also provide a new implementation for clear_pvalue(), which is used to wipe out materialized values in order to reduce footprint.

run(pipeline)

Execute the entire pipeline or the sub-DAG reachable from a node.

apply(transform, input)

Runner callback for a pipeline.apply call.

Parameters:

• transform – the transform to apply.
• input – transform’s input (typically a PCollection).

A concrete implementation of the Runner class may want to do custom pipeline construction for a given transform. To override the behavior for a transform class Xyz, implement an apply_Xyz method with this same signature.

apply_PTransform(transform, input)

run_transform(transform_node)

Runner callback for a pipeline.run call.

Parameters:transform_node – transform node for the transform to run.

A concrete implementation of the Runner class must implement run_Abc for some class Abc in the method resolution order for every non-composite transform Xyz in the pipeline.

class apache_beam.runners.runner.PipelineState

Bases: object

State of the Pipeline, as returned by PipelineResult.state.

This is meant to be the union of all the states any runner can put a pipeline in. Currently, it represents the values of the dataflow API JobState enum.

UNKNOWN = 'UNKNOWN'

STARTING = 'STARTING'

STOPPED = 'STOPPED'

RUNNING = 'RUNNING'

DONE = 'DONE'

FAILED = 'FAILED'

CANCELLED = 'CANCELLED'

UPDATED = 'UPDATED'

DRAINING = 'DRAINING'

DRAINED = 'DRAINED'

PENDING = 'PENDING'

CANCELLING = 'CANCELLING'

class apache_beam.runners.runner.PipelineResult(state)

Bases: object

A PipelineResult provides access to info about a pipeline.

state

Return the current state of the pipeline execution.

wait_until_finish(duration=None)

Waits until the pipeline finishes and returns the final status.

Parameters:

duration (int) – The time to wait (in milliseconds) for job to finish. If it is set to None, it will wait indefinitely until the job is finished.

Raises:

• IOError – If there is a persistent problem getting job information.
• NotImplementedError – If the runner does not support this operation.

Returns:

The final state of the pipeline, or None on timeout.

cancel()

Cancels the pipeline execution.

Raises:

• IOError – If there is a persistent problem getting job information.
• NotImplementedError – If the runner does not support this operation.

Returns:

The final state of the pipeline.

metrics()

Returns MetricResults object to query metrics from the runner.

Raises:NotImplementedError – If the runner does not support this operation.

aggregated_values(aggregator_or_name)

Return a dict of step names to values of the Aggregator.