schrodinger.seam.reporting.pipeline_snapshot module

Load and query recorded Seam pipelines and their output counts.

class schrodinger.seam.reporting.pipeline_snapshot.PipelineSnapshot(pipeline_proto: Pipeline, output_counts: dict[str, int])

Bases: object

Indexes a recorded pipeline’s transforms, leaf producers, and counts.

__init__(pipeline_proto: Pipeline, output_counts: dict[str, int])

classmethod fromSeamDir(seam_dir: schrodinger.seam.runners._seamdir.SeamDir | pathlib.Path | str) → Self

Load a snapshot from a completed Seam run.

Parameters:

seam_dir – completed Seam run directory

Raises:

ValueError – if the run did not complete successfully

walkTransforms() → Iterator[PTransform]

Yield transforms in canonical root and subtransform order.

getTransform(unique_name: str) → org.apache.beam.model.pipeline.v1.beam_runner_api_pb2.PTransform | None

static getDisplayData(transform: PTransform, key: str) → str | None

Return a labelled display-data string value by key.

resolveCountedOutput(unique_name: str, output_tag: str | None = None) → str

Resolve one transform output to its unique counted leaf producer.

An explicit output_tag selects that output. Otherwise, the sole output or Beam’s untagged default output is selected. A transform with multiple tagged outputs and no untagged default requires an explicit tag. Within the selected transform’s subtree, the PCollection must have exactly one leaf producer in the snapshot’s output counts. Leaves belonging to downstream transforms are not candidates.

Raises:

ValueError – if the transform or output does not exist, output selection is ambiguous, or the selected PCollection does not have exactly one counted leaf producer

getPCollectionCount(pcoll_id: str) → int | None

Return a PCollection’s count when it has one counted leaf producer.