solprop.driver.session_loop#

The session loop: Python’s side of an execution session (SEP 47 section 5).

Not to be confused with solprop.service, the front end’s service. A session (solprop.backends.session) owns the run’s numerical state and its schedule; serve() advances it and services what it reports (saved planes and their output transactions, plane observers and progress, snapshot persistence, warnings and the terminal record), then acknowledges each batch, with what it needs besides the session and the output in a ServiceContext. It holds no numerical state of its own: it never rediscovers the next plane, rebuilds a statistic or replays the attempts to reconstruct a controller. _Exports shares one export per pending ticket among its consumers, _Chunks gathers the attempts and rows since the last plane, _MarchKeeper keeps the Adams march’s state beside each save, and _Status and MONITOR_RECONSTRUCTED feed the execution record. Ordering follows SEP 47 section 5.3: a plane is committed with its method history and its execution record in one transaction (solprop._interrupts.committing); a snapshot is serviced after that commit; the plane observer and the reporter follow. A service failure leaves the batch unacknowledged, so the field it names can still be exported for the terminal checkpoint, and the primary failure is preserved.

Module Attributes

MONITOR_RECONSTRUCTED

How a resumed execution's boundary monitor was restored (SEP 47 section 5.5).

Functions

serve(session, output, context)

Advance a session to its end, servicing every event it reports.

Classes

ServiceContext(problem, request, metadata, ...)

What the service loop needs besides the session and the output.

solprop.driver.session_loop.MONITOR_RECONSTRUCTED = {'reason': 'persisted_records', 'restarted': ['check_cadence', 'transverse_cap'], 'restored': ['cumulative_totals', 'quiet_state', 'transverse_record'], 'state': 'reconstructed'}#

How a resumed execution’s boundary monitor was restored (SEP 47 section 5.5).

Today’s reconstruction from the persisted records: the cumulative totals, the quiet state and the transverse record are restored, and the check cadence and the transverse cap restart from that restored state; persisting them is not claimed.

class solprop.driver.session_loop.ServiceContext(problem, request, metadata, reporter, retention, resume_history, snapshots, tdse_binding, tdse_archive, recipe, start_plane, stop=None, restoration=None, interpolated_resume=False)[source]#

Bases: object

What the service loop needs besides the session and the output.

Attributes:
problemsolprop.translate.problem.PropagationProblem

The translated problem.

requestsolprop.translate.run_request.RunRequest

The run’s policies.

metadatasolprop.output.metadata.RunMetadata

The run’s record.

reportersolprop.log.ProgressReporter

Where progress goes.

retentionsolprop.output.store.FieldRetention

Which planes keep their field.

resume_historybool or None

The public method-history retention choice.

snapshotssolprop.output.snapshots.SnapshotController or None

Snapshot requests, when enabled.

tdse_bindingsolprop.translate.tdse.TDSEBinding or None

A live atomic runtime.

tdse_archivesolprop.output.tdse_archive.TDSEArchiveWriter or None

A live archive to suspend beside the optical writer.

recipesolprop.output.records.RecipeRecord or None

The recipe the run was built from.

start_planeint

The first save plane beyond the starting state.

stopsolprop.log.StopFlag or None

The run’s stop flag, polled between batches as well as inside the session.

restorationcollections.abc.Mapping or None, optional

What a resumed execution restored of the method history (solprop.driver.resume._continue_march), for its execution record; None for a fresh run or a caller that says nothing.

interpolated_resumebool, optional

Whether a resumed run continues afresh from a plane interpolated inside a step (resume="interpolated"), whose position a plane store then records before the first step.

interpolated_resume: bool = False#
metadata: RunMetadata#
problem: PropagationProblem#
recipe: RecipeRecord | None#
reporter: ProgressReporter#
request: RunRequest#
restoration: Mapping[str, object] | None = None#
resume_history: bool | None#
retention: FieldRetention#
snapshots: SnapshotController | None#
start_plane: int#
stop: StopFlag | None = None#
tdse_archive: TDSEArchiveWriter | None#
tdse_binding: TDSEBinding | None#
solprop.driver.session_loop.serve(session, output, context)[source]#

Advance a session to its end, servicing every event it reports.

Any RunSession is served – the reference’s HostSession or the native backend’s – through its protocol alone (SessionFacts).

Parameters:
sessionsolprop.backends.session.RunSession

The started session.

outputsolprop.output.store.Output

Where the planes go; closed when the loop ends, whether or not it succeeded.

contextServiceContext

Everything else the service needs.

Raises:
BaseException

The run’s primary failure, a stop request or a signal, after its terminal record and any checkpoint are written.