Skip to content

prefect.utilities.engine

collect_task_run_inputs(expr, max_depth=-1) async

This function recurses through an expression to generate a set of any discernible task run inputs it finds in the data structure. It produces a set of all inputs found.

Examples:

>>> task_inputs = {
>>>    k: await collect_task_run_inputs(v) for k, v in parameters.items()
>>> }
Source code in src/prefect/utilities/engine.py
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
async def collect_task_run_inputs(expr: Any, max_depth: int = -1) -> Set[TaskRunInput]:
    """
    This function recurses through an expression to generate a set of any discernible
    task run inputs it finds in the data structure. It produces a set of all inputs
    found.

    Examples:
        >>> task_inputs = {
        >>>    k: await collect_task_run_inputs(v) for k, v in parameters.items()
        >>> }
    """
    # TODO: This function needs to be updated to detect parameters and constants

    inputs = set()
    futures = set()

    def add_futures_and_states_to_inputs(obj):
        if isinstance(obj, PrefectFuture):
            # We need to wait for futures to be submitted before we can get the task
            # run id but we want to do so asynchronously
            futures.add(obj)
        elif isinstance(obj, State):
            if obj.state_details.task_run_id:
                inputs.add(TaskRunResult(id=obj.state_details.task_run_id))
        # Expressions inside quotes should not be traversed
        elif isinstance(obj, quote):
            raise StopVisiting
        else:
            state = get_state_for_result(obj)
            if state and state.state_details.task_run_id:
                inputs.add(TaskRunResult(id=state.state_details.task_run_id))

    visit_collection(
        expr,
        visit_fn=add_futures_and_states_to_inputs,
        return_data=False,
        max_depth=max_depth,
    )

    await asyncio.gather(*[future._wait_for_submission() for future in futures])
    for future in futures:
        inputs.add(TaskRunResult(id=future.task_run.id))

    return inputs

collect_task_run_inputs_sync(expr, future_cls=PrefectFuture, max_depth=-1)

This function recurses through an expression to generate a set of any discernible task run inputs it finds in the data structure. It produces a set of all inputs found.

Examples:

>>> task_inputs = {
>>>    k: collect_task_run_inputs_sync(v) for k, v in parameters.items()
>>> }
Source code in src/prefect/utilities/engine.py
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
def collect_task_run_inputs_sync(
    expr: Any, future_cls: Any = PrefectFuture, max_depth: int = -1
) -> Set[TaskRunInput]:
    """
    This function recurses through an expression to generate a set of any discernible
    task run inputs it finds in the data structure. It produces a set of all inputs
    found.

    Examples:
        >>> task_inputs = {
        >>>    k: collect_task_run_inputs_sync(v) for k, v in parameters.items()
        >>> }
    """
    # TODO: This function needs to be updated to detect parameters and constants

    inputs = set()

    def add_futures_and_states_to_inputs(obj):
        if isinstance(obj, future_cls) and hasattr(obj, "task_run_id"):
            inputs.add(TaskRunResult(id=obj.task_run_id))
        elif isinstance(obj, State):
            if obj.state_details.task_run_id:
                inputs.add(TaskRunResult(id=obj.state_details.task_run_id))
        # Expressions inside quotes should not be traversed
        elif isinstance(obj, quote):
            raise StopVisiting
        else:
            state = get_state_for_result(obj)
            if state and state.state_details.task_run_id:
                inputs.add(TaskRunResult(id=state.state_details.task_run_id))

    visit_collection(
        expr,
        visit_fn=add_futures_and_states_to_inputs,
        return_data=False,
        max_depth=max_depth,
    )

    return inputs

get_state_for_result(obj)

Get the state related to a result object.

link_state_to_result must have been called first.

Source code in src/prefect/utilities/engine.py
584
585
586
587
588
589
590
591
592
def get_state_for_result(obj: Any) -> Optional[State]:
    """
    Get the state related to a result object.

    `link_state_to_result` must have been called first.
    """
    flow_run_context = FlowRunContext.get()
    if flow_run_context:
        return flow_run_context.task_run_results.get(id(obj))

Caches a link between a state and a result and its components using the id of the components to map to the state. The cache is persisted to the current flow run context since task relationships are limited to within a flow run.

This allows dependency tracking to occur when results are passed around. Note: Because id is used, we cannot cache links between singleton objects.

We only cache the relationship between components 1-layer deep. Example: Given the result [1, ["a","b"], ("c",)], the following elements will be mapped to the state: - [1, ["a","b"], ("c",)] - ["a","b"] - ("c",)

Note: the int `1` will not be mapped to the state because it is a singleton.

Other Notes: We do not hash the result because: - If changes are made to the object in the flow between task calls, we can still track that they are related. - Hashing can be expensive. - Not all objects are hashable.

We do not set an attribute, e.g. __prefect_state__, on the result because:

  • Mutating user's objects is dangerous.
  • Unrelated equality comparisons can break unexpectedly.
  • The field can be preserved on copy.
  • We cannot set this attribute on Python built-ins.
Source code in src/prefect/utilities/engine.py
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
def link_state_to_result(state: State, result: Any) -> None:
    """
    Caches a link between a state and a result and its components using
    the `id` of the components to map to the state. The cache is persisted to the
    current flow run context since task relationships are limited to within a flow run.

    This allows dependency tracking to occur when results are passed around.
    Note: Because `id` is used, we cannot cache links between singleton objects.

    We only cache the relationship between components 1-layer deep.
    Example:
        Given the result [1, ["a","b"], ("c",)], the following elements will be
        mapped to the state:
        - [1, ["a","b"], ("c",)]
        - ["a","b"]
        - ("c",)

        Note: the int `1` will not be mapped to the state because it is a singleton.

    Other Notes:
    We do not hash the result because:
    - If changes are made to the object in the flow between task calls, we can still
      track that they are related.
    - Hashing can be expensive.
    - Not all objects are hashable.

    We do not set an attribute, e.g. `__prefect_state__`, on the result because:

    - Mutating user's objects is dangerous.
    - Unrelated equality comparisons can break unexpectedly.
    - The field can be preserved on copy.
    - We cannot set this attribute on Python built-ins.
    """

    flow_run_context = FlowRunContext.get()
    # Drop the data field to avoid holding a strong reference to the result
    # Holding large user objects in memory can cause memory bloat
    linked_state = state.model_copy(update={"data": None})

    def link_if_trackable(obj: Any) -> None:
        """Track connection between a task run result and its associated state if it has a unique ID.

        We cannot track booleans, Ellipsis, None, NotImplemented, or the integers from -5 to 256
        because they are singletons.

        This function will mutate the State if the object is an untrackable type by setting the value
        for `State.state_details.untrackable_result` to `True`.

        """
        if (type(obj) in UNTRACKABLE_TYPES) or (
            isinstance(obj, int) and (-5 <= obj <= 256)
        ):
            state.state_details.untrackable_result = True
            return
        flow_run_context.task_run_results[id(obj)] = linked_state

    if flow_run_context:
        visit_collection(expr=result, visit_fn=link_if_trackable, max_depth=1)

propose_state(client, state, force=False, task_run_id=None, flow_run_id=None) async

Propose a new state for a flow run or task run, invoking Prefect orchestration logic.

If the proposed state is accepted, the provided state will be augmented with details and returned.

If the proposed state is rejected, a new state returned by the Prefect API will be returned.

If the proposed state results in a WAIT instruction from the Prefect API, the function will sleep and attempt to propose the state again.

If the proposed state results in an ABORT instruction from the Prefect API, an error will be raised.

Parameters:

Name Type Description Default
state State[object]

a new state for the task or flow run

required
task_run_id Optional[UUID]

an optional task run id, used when proposing task run states

None
flow_run_id Optional[UUID]

an optional flow run id, used when proposing flow run states

None

Returns:

Type Description
State[object]

a State model representation of the flow or task run state

Raises:

Type Description
ValueError

if neither task_run_id or flow_run_id is provided

Abort

if an ABORT instruction is received from the Prefect API

Source code in src/prefect/utilities/engine.py
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
async def propose_state(
    client: "PrefectClient",
    state: State[object],
    force: bool = False,
    task_run_id: Optional[UUID] = None,
    flow_run_id: Optional[UUID] = None,
) -> State[object]:
    """
    Propose a new state for a flow run or task run, invoking Prefect orchestration logic.

    If the proposed state is accepted, the provided `state` will be augmented with
     details and returned.

    If the proposed state is rejected, a new state returned by the Prefect API will be
    returned.

    If the proposed state results in a WAIT instruction from the Prefect API, the
    function will sleep and attempt to propose the state again.

    If the proposed state results in an ABORT instruction from the Prefect API, an
    error will be raised.

    Args:
        state: a new state for the task or flow run
        task_run_id: an optional task run id, used when proposing task run states
        flow_run_id: an optional flow run id, used when proposing flow run states

    Returns:
        a [State model][prefect.client.schemas.objects.State] representation of the
            flow or task run state

    Raises:
        ValueError: if neither task_run_id or flow_run_id is provided
        prefect.exceptions.Abort: if an ABORT instruction is received from
            the Prefect API
    """

    # Determine if working with a task run or flow run
    if not task_run_id and not flow_run_id:
        raise ValueError("You must provide either a `task_run_id` or `flow_run_id`")

    # Handle task and sub-flow tracing
    if state.is_final():
        if isinstance(state.data, BaseResult) and state.data.has_cached_object():
            # Avoid fetching the result unless it is cached, otherwise we defeat
            # the purpose of disabling `cache_result_in_memory`
            result = await state.result(raise_on_failure=False, fetch=True)
        elif isinstance(state.data, ResultRecord):
            result = state.data.result
        else:
            result = state.data

        link_state_to_result(state, result)

    # Handle repeated WAITs in a loop instead of recursively, to avoid
    # reaching max recursion depth in extreme cases.
    async def set_state_and_handle_waits(set_state_func) -> OrchestrationResult:
        response = await set_state_func()
        while response.status == SetStateStatus.WAIT:
            engine_logger.debug(
                f"Received wait instruction for {response.details.delay_seconds}s: "
                f"{response.details.reason}"
            )
            await anyio.sleep(response.details.delay_seconds)
            response = await set_state_func()
        return response

    # Attempt to set the state
    if task_run_id:
        set_state = partial(client.set_task_run_state, task_run_id, state, force=force)
        response = await set_state_and_handle_waits(set_state)
    elif flow_run_id:
        set_state = partial(client.set_flow_run_state, flow_run_id, state, force=force)
        response = await set_state_and_handle_waits(set_state)
    else:
        raise ValueError(
            "Neither flow run id or task run id were provided. At least one must "
            "be given."
        )

    # Parse the response to return the new state
    if response.status == SetStateStatus.ACCEPT:
        # Update the state with the details if provided
        state.id = response.state.id
        state.timestamp = response.state.timestamp
        if response.state.state_details:
            state.state_details = response.state.state_details
        return state

    elif response.status == SetStateStatus.ABORT:
        raise prefect.exceptions.Abort(response.details.reason)

    elif response.status == SetStateStatus.REJECT:
        if response.state.is_paused():
            raise Pause(response.details.reason, state=response.state)
        return response.state

    else:
        raise ValueError(
            f"Received unexpected `SetStateStatus` from server: {response.status!r}"
        )

propose_state_sync(client, state, force=False, task_run_id=None, flow_run_id=None)

Propose a new state for a flow run or task run, invoking Prefect orchestration logic.

If the proposed state is accepted, the provided state will be augmented with details and returned.

If the proposed state is rejected, a new state returned by the Prefect API will be returned.

If the proposed state results in a WAIT instruction from the Prefect API, the function will sleep and attempt to propose the state again.

If the proposed state results in an ABORT instruction from the Prefect API, an error will be raised.

Parameters:

Name Type Description Default
state State[object]

a new state for the task or flow run

required
task_run_id Optional[UUID]

an optional task run id, used when proposing task run states

None
flow_run_id Optional[UUID]

an optional flow run id, used when proposing flow run states

None

Returns:

Type Description
State[object]

a State model representation of the flow or task run state

Raises:

Type Description
ValueError

if neither task_run_id or flow_run_id is provided

Abort

if an ABORT instruction is received from the Prefect API

Source code in src/prefect/utilities/engine.py
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
def propose_state_sync(
    client: "SyncPrefectClient",
    state: State[object],
    force: bool = False,
    task_run_id: Optional[UUID] = None,
    flow_run_id: Optional[UUID] = None,
) -> State[object]:
    """
    Propose a new state for a flow run or task run, invoking Prefect orchestration logic.

    If the proposed state is accepted, the provided `state` will be augmented with
     details and returned.

    If the proposed state is rejected, a new state returned by the Prefect API will be
    returned.

    If the proposed state results in a WAIT instruction from the Prefect API, the
    function will sleep and attempt to propose the state again.

    If the proposed state results in an ABORT instruction from the Prefect API, an
    error will be raised.

    Args:
        state: a new state for the task or flow run
        task_run_id: an optional task run id, used when proposing task run states
        flow_run_id: an optional flow run id, used when proposing flow run states

    Returns:
        a [State model][prefect.client.schemas.objects.State] representation of the
            flow or task run state

    Raises:
        ValueError: if neither task_run_id or flow_run_id is provided
        prefect.exceptions.Abort: if an ABORT instruction is received from
            the Prefect API
    """

    # Determine if working with a task run or flow run
    if not task_run_id and not flow_run_id:
        raise ValueError("You must provide either a `task_run_id` or `flow_run_id`")

    # Handle task and sub-flow tracing
    if state.is_final():
        if isinstance(state.data, BaseResult) and state.data.has_cached_object():
            # Avoid fetching the result unless it is cached, otherwise we defeat
            # the purpose of disabling `cache_result_in_memory`
            result = state.result(raise_on_failure=False, fetch=True)
            if asyncio.iscoroutine(result):
                result = run_coro_as_sync(result)
        elif isinstance(state.data, ResultRecord):
            result = state.data.result
        else:
            result = state.data

        link_state_to_result(state, result)

    # Handle repeated WAITs in a loop instead of recursively, to avoid
    # reaching max recursion depth in extreme cases.
    def set_state_and_handle_waits(set_state_func) -> OrchestrationResult:
        response = set_state_func()
        while response.status == SetStateStatus.WAIT:
            engine_logger.debug(
                f"Received wait instruction for {response.details.delay_seconds}s: "
                f"{response.details.reason}"
            )
            time.sleep(response.details.delay_seconds)
            response = set_state_func()
        return response

    # Attempt to set the state
    if task_run_id:
        set_state = partial(client.set_task_run_state, task_run_id, state, force=force)
        response = set_state_and_handle_waits(set_state)
    elif flow_run_id:
        set_state = partial(client.set_flow_run_state, flow_run_id, state, force=force)
        response = set_state_and_handle_waits(set_state)
    else:
        raise ValueError(
            "Neither flow run id or task run id were provided. At least one must "
            "be given."
        )

    # Parse the response to return the new state
    if response.status == SetStateStatus.ACCEPT:
        # Update the state with the details if provided
        state.id = response.state.id
        state.timestamp = response.state.timestamp
        if response.state.state_details:
            state.state_details = response.state.state_details
        return state

    elif response.status == SetStateStatus.ABORT:
        raise prefect.exceptions.Abort(response.details.reason)

    elif response.status == SetStateStatus.REJECT:
        if response.state.is_paused():
            raise Pause(response.details.reason, state=response.state)
        return response.state

    else:
        raise ValueError(
            f"Received unexpected `SetStateStatus` from server: {response.status!r}"
        )

resolve_inputs(parameters, return_data=True, max_depth=-1) async

Resolve any Quote, PrefectFuture, or State types nested in parameters into data.

Returns:

Type Description
Dict[str, Any]

A copy of the parameters with resolved data

Raises:

Type Description
UpstreamTaskError

If any of the upstream states are not COMPLETED

Source code in src/prefect/utilities/engine.py
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
async def resolve_inputs(
    parameters: Dict[str, Any], return_data: bool = True, max_depth: int = -1
) -> Dict[str, Any]:
    """
    Resolve any `Quote`, `PrefectFuture`, or `State` types nested in parameters into
    data.

    Returns:
        A copy of the parameters with resolved data

    Raises:
        UpstreamTaskError: If any of the upstream states are not `COMPLETED`
    """

    futures = set()
    states = set()
    result_by_state = {}

    if not parameters:
        return {}

    def collect_futures_and_states(expr, context):
        # Expressions inside quotes should not be traversed
        if isinstance(context.get("annotation"), quote):
            raise StopVisiting()

        if isinstance(expr, PrefectFuture):
            futures.add(expr)
        if isinstance(expr, State):
            states.add(expr)

        return expr

    visit_collection(
        parameters,
        visit_fn=collect_futures_and_states,
        return_data=False,
        max_depth=max_depth,
        context={},
    )

    # Wait for all futures so we do not block when we retrieve the state in `resolve_input`
    states.update(await asyncio.gather(*[future._wait() for future in futures]))

    # Only retrieve the result if requested as it may be expensive
    if return_data:
        finished_states = [state for state in states if state.is_final()]

        state_results = await asyncio.gather(
            *[
                state.result(raise_on_failure=False, fetch=True)
                for state in finished_states
            ]
        )

        for state, result in zip(finished_states, state_results):
            result_by_state[state] = result

    def resolve_input(expr, context):
        state = None

        # Expressions inside quotes should not be modified
        if isinstance(context.get("annotation"), quote):
            raise StopVisiting()

        if isinstance(expr, PrefectFuture):
            state = expr._final_state
        elif isinstance(expr, State):
            state = expr
        else:
            return expr

        # Do not allow uncompleted upstreams except failures when `allow_failure` has
        # been used
        if not state.is_completed() and not (
            # TODO: Note that the contextual annotation here is only at the current level
            #       if `allow_failure` is used then another annotation is used, this will
            #       incorrectly evaluate to false — to resolve this, we must track all
            #       annotations wrapping the current expression but this is not yet
            #       implemented.
            isinstance(context.get("annotation"), allow_failure) and state.is_failed()
        ):
            raise UpstreamTaskError(
                f"Upstream task run '{state.state_details.task_run_id}' did not reach a"
                " 'COMPLETED' state."
            )

        return result_by_state.get(state)

    resolved_parameters = {}
    for parameter, value in parameters.items():
        try:
            resolved_parameters[parameter] = visit_collection(
                value,
                visit_fn=resolve_input,
                return_data=return_data,
                # we're manually going 1 layer deeper here
                max_depth=max_depth - 1,
                remove_annotations=True,
                context={},
            )
        except UpstreamTaskError:
            raise
        except Exception as exc:
            raise PrefectException(
                f"Failed to resolve inputs in parameter {parameter!r}. If your"
                " parameter type is not supported, consider using the `quote`"
                " annotation to skip resolution of inputs."
            ) from exc

    return resolved_parameters

resolve_inputs_sync(parameters, return_data=True, max_depth=-1)

Resolve any Quote, PrefectFuture, or State types nested in parameters into data.

Returns:

Type Description
Dict[str, Any]

A copy of the parameters with resolved data

Raises:

Type Description
UpstreamTaskError

If any of the upstream states are not COMPLETED

Source code in src/prefect/utilities/engine.py
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
def resolve_inputs_sync(
    parameters: Dict[str, Any], return_data: bool = True, max_depth: int = -1
) -> Dict[str, Any]:
    """
    Resolve any `Quote`, `PrefectFuture`, or `State` types nested in parameters into
    data.

    Returns:
        A copy of the parameters with resolved data

    Raises:
        UpstreamTaskError: If any of the upstream states are not `COMPLETED`
    """

    if not parameters:
        return {}

    resolved_parameters = {}
    for parameter, value in parameters.items():
        try:
            resolved_parameters[parameter] = visit_collection(
                value,
                visit_fn=resolve_to_final_result,
                return_data=return_data,
                max_depth=max_depth,
                remove_annotations=True,
                context={},
            )
        except UpstreamTaskError:
            raise
        except Exception as exc:
            raise PrefectException(
                f"Failed to resolve inputs in parameter {parameter!r}. If your"
                " parameter type is not supported, consider using the `quote`"
                " annotation to skip resolution of inputs."
            ) from exc

    return resolved_parameters

resolve_to_final_result(expr, context)

Resolve any PrefectFuture, or State types nested in parameters into data. Designed to be use with visit_collection.

Source code in src/prefect/utilities/engine.py
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
def resolve_to_final_result(expr, context):
    """
    Resolve any `PrefectFuture`, or `State` types nested in parameters into
    data. Designed to be use with `visit_collection`.
    """
    state = None

    # Expressions inside quotes should not be modified
    if isinstance(context.get("annotation"), quote):
        raise StopVisiting()

    if isinstance(expr, PrefectFuture):
        upstream_task_run = context.get("current_task_run")
        upstream_task = context.get("current_task")
        if (
            upstream_task
            and upstream_task_run
            and expr.task_run_id == upstream_task_run.id
        ):
            raise ValueError(
                f"Discovered a task depending on itself. Raising to avoid a deadlock. Please inspect the inputs and dependencies of {upstream_task.name}."
            )

        expr.wait()
        state = expr.state
    elif isinstance(expr, State):
        state = expr
    else:
        return expr

    assert state

    # Do not allow uncompleted upstreams except failures when `allow_failure` has
    # been used
    if not state.is_completed() and not (
        # TODO: Note that the contextual annotation here is only at the current level
        #       if `allow_failure` is used then another annotation is used, this will
        #       incorrectly evaluate to false — to resolve this, we must track all
        #       annotations wrapping the current expression but this is not yet
        #       implemented.
        isinstance(context.get("annotation"), allow_failure) and state.is_failed()
    ):
        raise UpstreamTaskError(
            f"Upstream task run '{state.state_details.task_run_id}' did not reach a"
            " 'COMPLETED' state."
        )

    _result = state.result(raise_on_failure=False, fetch=True)
    if asyncio.iscoroutine(_result):
        _result = run_coro_as_sync(_result)
    return _result