-
Notifications
You must be signed in to change notification settings - Fork 1.6k
/
flow.py
1741 lines (1483 loc) · 67.2 KB
/
flow.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
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
121
122
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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
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
354
355
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
457
458
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
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
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
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
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
884
885
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
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import collections
import collections.abc
import copy
import functools
import hashlib
import inspect
import itertools
import json
import os
import tempfile
import time
import uuid
import warnings
from contextlib import contextmanager
from pathlib import Path
from typing import (
Any,
Callable,
Dict,
Iterable,
Iterator,
List,
Mapping,
Optional,
Set,
Tuple,
cast,
Union,
)
import pendulum
from mypy_extensions import TypedDict
from slugify import slugify
import prefect
import prefect.schedules
from prefect.core.edge import Edge
from prefect.core.parameter import Parameter
from prefect.core.task import Task
from prefect.executors import Executor
from prefect.engine.result import Result
from prefect.engine.state import State
from prefect.storage import Storage, get_default_storage_class
from prefect.run_configs import RunConfig, UniversalRun
from prefect.utilities import diagnostics, logging
from prefect.utilities.configuration import set_temporary_config
from prefect.utilities.notifications import callback_factory
from prefect.utilities.tasks import as_task
ParameterDetails = TypedDict("ParameterDetails", {"default": Any, "required": bool})
def cache(method: Callable) -> Callable:
"""
Decorator for caching Flow methods.
Each Flow has a _cache dict that can be used to memoize expensive functions. This
decorator automatically compares a hash of the Flow's current tasks, edges, and reference_tasks
to a cached hash; if the hash is the same, it attempts to retrieve a value from the cache.
If the hash is different, it invalidates the cache.
"""
@functools.wraps(method)
def wrapper(self, *args, **kwargs): # type: ignore
cache_check = dict(
tasks=self.tasks.copy(),
edges=self.edges.copy(),
reference_tasks=copy.copy(self._reference_tasks),
)
if any(self._cache.get(k) != v for k, v in cache_check.items()):
self._cache.clear()
self._cache.update(cache_check)
callargs = inspect.signature(method).bind(self, *args, **kwargs).arguments
key = (method.__name__, tuple(callargs.items())[1:])
if key not in self._cache:
self._cache[key] = method(self, *args, **kwargs)
return self._cache[key]
return wrapper
class Flow:
"""
The Flow class is used as the representation of a collection of dependent Tasks. Flows
track Task dependencies, parameters and provide the main API for constructing and managing
workflows.
Initializing Flow example:
```python
class MyTask(Task):
def run(self):
return "hello"
task_1 = MyTask()
flow = Flow(name="my_flow", tasks=[task_1])
flow.run()
```
Initializing Flow as context manager example:
```python
@task
def my_task():
return "hello"
with Flow("my_flow") as flow:
task_1 = my_task()
flow.run()
```
Args:
- name (str): The name of the flow. Cannot be `None` or an empty string
- schedule (prefect.schedules.Schedule, optional): A default schedule for the flow
- executor (prefect.executors.Executor, optional): The executor that the flow
should use. If `None`, the default executor configured in the runtime environment
will be used.
- run_config (prefect.run_configs.RunConfig, optional): The runtime
configuration to use when deploying this flow.
- storage (prefect.storage.Storage, optional): The unit of storage
that the flow will be written into.
- tasks ([Task], optional): If provided, a list of tasks that will initialize the flow
- edges ([Edge], optional): A list of edges between tasks
- reference_tasks ([Task], optional): A list of tasks that determine the final
state of a flow
- result (Result, optional): the result instance used to retrieve and store
task results during execution
- state_handlers (Iterable[Callable], optional): A list of state change handlers
that will be called whenever the flow changes state, providing an
opportunity to inspect or modify the new state. The handler
will be passed the flow instance, the old (prior) state, and the new
(current) state, with the following signature:
`state_handler(flow: Flow, old_state: State, new_state: State) -> Optional[State]`
If multiple functions are passed, then the `new_state` argument will be the
result of the previous handler.
- on_failure (Callable, optional): A function with signature `fn(flow: Flow, state:
State) -> None` which will be called anytime this Flow enters a failure state
- validate (bool, optional): Whether or not to check the validity of
the flow (e.g., presence of cycles and illegal keys) after adding the edges passed
in the `edges` argument. Defaults to the value of `eager_edge_validation` in
your prefect configuration file.
- terminal_state_handler (Callable, optional): A state handler that can be used to
inspect or modify the final state of the flow run. Expects a callable with signature
`handler(flow: Flow, state: State, reference_task_states: Set[State]) -> Optional[State]`,
where `flow` is the current Flow, `state` is the current state of the Flow run, and
`reference_task_states` is set of states for all reference tasks in the flow. It should
return either a new state for the flow run, or `None` (in which case the existing
state will be used).
"""
def __init__(
self,
name: str,
schedule: prefect.schedules.Schedule = None,
executor: Executor = None,
run_config: RunConfig = None,
storage: Storage = None,
tasks: Iterable[Task] = None,
edges: Iterable[Edge] = None,
reference_tasks: Iterable[Task] = None,
state_handlers: List[Callable] = None,
on_failure: Callable = None,
validate: bool = None,
result: Optional[Result] = None,
terminal_state_handler: Optional[
Callable[["Flow", State, Set[State]], Optional[State]]
] = None,
):
self._cache = {} # type: dict
if not name:
raise ValueError("A name must be provided for the flow.")
self.name = name
self.logger = logging.get_logger(self.name)
self.schedule = schedule
self.executor = executor
self.run_config = run_config
self.storage = storage
self.result = result
self.terminal_state_handler = terminal_state_handler
self.tasks = set() # type: Set[Task]
self.edges = set() # type: Set[Edge]
self._slug_counters = collections.defaultdict(
cast(Callable, functools.partial(itertools.count, 1))
) # type: Dict[str, Iterator[int]]
self.slugs = {} # type: Dict[Task, str]
self.constants = collections.defaultdict(
dict
) # type: Dict[Task, Dict[str, Any]]
for t in tasks or []:
self.add_task(t)
self.set_reference_tasks(reference_tasks or [])
for e in edges or []:
self.add_edge(
upstream_task=e.upstream_task,
downstream_task=e.downstream_task,
key=e.key,
mapped=e.mapped,
flattened=e.flattened,
validate=validate,
)
self._prefect_version = prefect.__version__
if state_handlers and not isinstance(state_handlers, collections.abc.Sequence):
raise TypeError("state_handlers should be iterable.")
self.state_handlers = state_handlers or []
if on_failure is not None:
self.state_handlers.append(
callback_factory(on_failure, check=lambda s: s.is_failed())
)
super().__init__()
def __eq__(self, other: Any) -> bool:
if type(self) == type(other):
s = (self.name, self.tasks, self.edges, self.reference_tasks())
o = (other.name, other.tasks, other.edges, other.reference_tasks())
return s == o
return False
def __repr__(self) -> str:
template = '<{cls}: name="{self.name}">'
return template.format(cls=type(self).__name__, self=self)
def __iter__(self) -> Iterable[Task]:
yield from self.sorted_tasks()
def copy(self) -> "Flow":
"""
Create and returns a copy of the current Flow.
"""
new = copy.copy(self)
# create a new cache
new._cache = dict()
new.constants = self.constants.copy()
new.tasks = self.tasks.copy()
new.edges = self.edges.copy()
new.slugs = self.slugs.copy()
new.set_reference_tasks(self._reference_tasks)
return new
# Identification -----------------------------------------------------------
def get_tasks(
self,
name: str = None,
slug: str = None,
tags: Iterable[str] = None,
task_type: type = None,
) -> List[Task]:
"""
Helper method for retrieving tasks from this flow based on certain attributes.
The _intersection_ of all provided attributes is taken, i.e., only those tasks
which match _all_ provided conditions are returned.
Args:
- name (str, optional): the name of the task
- slug (str, optional): the slug of the task
- tags ([str], optional): an iterable of task tags
- task_type (type, optional): a possible task class type
Returns:
- [Task]: a list of tasks that meet the required conditions
"""
def sieve(t: Task) -> bool:
keep = True
if name is not None:
keep &= t.name == name
if slug is not None:
keep &= t.slug == slug
if tags is not None:
keep &= t.tags.issuperset(tags)
if task_type is not None:
keep &= isinstance(t, task_type)
return keep
keep_tasks = filter(sieve, self.tasks)
return list(keep_tasks)
def replace(self, old: Task, new: Task, validate: bool = True) -> None:
"""
Performs an inplace replacement of the old task with the provided new task.
Args:
- old (Task): the old task to replace
- new (Task): the new task to replace the old with; if not a Prefect
Task, Prefect will attempt to convert it to one
- validate (boolean, optional): whether to validate the Flow after
the replace has been completed; defaults to `True`
Raises:
- ValueError: if the `old` task is not a part of this flow
"""
if old not in self.tasks:
raise ValueError("Task {t} was not found in Flow {f}".format(t=old, f=self))
new = as_task(new, flow=self)
# update tasks
self.tasks.remove(old)
self.slugs.pop(old)
self.add_task(new)
self._cache.clear()
affected_edges = {e for e in self.edges if old in e.tasks}
# remove old edges
for edge in affected_edges:
self.edges.remove(edge)
# replace with new edges
for edge in affected_edges:
upstream = new if edge.upstream_task == old else edge.upstream_task
downstream = new if edge.downstream_task == old else edge.downstream_task
self.add_edge(
upstream_task=upstream,
downstream_task=downstream,
key=edge.key,
mapped=edge.mapped,
flattened=edge.flattened,
validate=False,
)
if self._reference_tasks:
# update auxiliary task collections
ref_tasks = self.reference_tasks()
new_refs = [t for t in ref_tasks if t != old] + (
[new] if old in ref_tasks else []
)
self.set_reference_tasks(new_refs)
if validate:
self.validate()
# Context Manager ----------------------------------------------------------
@contextmanager
def _flow_context(self) -> Iterator["Flow"]:
"""
When entering a flow context, the Prefect context is modified to include:
- `flow`: the flow itself
- `_unused_task_tracker`: a set of all tasks created while the context is
open, in order to provide user friendly warnings if they aren't added
to the flow itself. This is purely for user experience.
"""
unused_task_tracker = set() # type: Set[Task]
with prefect.context(flow=self, _unused_task_tracker=unused_task_tracker):
yield self
if unused_task_tracker.difference(self.tasks):
warnings.warn(
"Tasks were created but not added to the flow: "
f"{unused_task_tracker.difference(self.tasks)}. This can occur "
"when `Task` classes, including `Parameters`, are instantiated "
"inside a `with flow:` block but not added to the flow either "
"explicitly or as the input to another task. For more information, see "
"https://docs.prefect.io/core/advanced_tutorials/"
"task-guide.html#adding-tasks-to-flows.",
stacklevel=2,
)
def __enter__(self) -> "Flow":
self._ctx = self._flow_context()
return self._ctx.__enter__()
def __exit__(self, exc_type, exc_value, traceback) -> None: # type: ignore
self._ctx.__exit__(exc_type, exc_value, traceback)
# delete _ctx because it's an active generator, which prevents pickling
del self._ctx
# Introspection ------------------------------------------------------------
@cache
def root_tasks(self) -> Set[Task]:
"""
Get the tasks in the flow that have no upstream dependencies; these are
the tasks that, by default, flow execution begins with.
Returns:
- set of Task objects that have no upstream dependencies
"""
return set(t for t in self.tasks if not self.edges_to(t))
@cache
def terminal_tasks(self) -> Set[Task]:
"""
Get the tasks in the flow that have no downstream dependencies
Returns:
- set of Task objects that have no downstream dependencies
"""
return set(t for t in self.tasks if not self.edges_from(t))
def parameters(self) -> Set[Parameter]:
"""
Returns any parameters of the flow.
Returns:
- set: a set of any Parameters in this flow
"""
return {p for p in self.tasks if isinstance(p, Parameter)}
@cache
def _default_reference_tasks(self) -> Set[Task]:
from prefect.tasks.core.resource_manager import (
ResourceInitTask,
ResourceCleanupTask,
)
# Select all tasks that aren't a ResourceInitTask/ResourceCleanupTask
# and have no downstream dependencies that aren't ResourceCleanupTasks
#
# Note: this feels a bit gross, since it special cases a certain
# subclass inside the flow runner. If this behavior expands to other
# classes we should think of a general way of indicating this on a Task
# class instead of special casing.
return {
t
for t in self.tasks
if not isinstance(t, (ResourceInitTask, ResourceCleanupTask))
and not any(
t
for t in self.downstream_tasks(t)
if not isinstance(t, ResourceCleanupTask)
)
}
def reference_tasks(self) -> Set[Task]:
"""
A flow's "reference tasks" are used to determine its state when it runs. If all the
reference tasks are successful, then the flow run is considered successful. However, if
any of the reference tasks fail, the flow is considered to fail. (Note that skips are
counted as successes; see [the state documentation](../engine/state.html) for a full
description of what is considered failure, success, etc.)
By default, a flow's reference tasks are its terminal tasks. This means the state of a
flow is determined by those tasks that have no downstream dependencies.
In some situations, users may want to customize this behavior; for example, if a flow's
terminal tasks are "clean up" tasks for the rest of the flow that only run if certain
(more relevant) tasks fail, we might not want them determining the overall state of the
flow run. The `flow.set_reference_tasks()` method can be used to set such custom
`reference_tasks`.
Please note that even if `reference_tasks` are provided that are not terminal tasks,
the flow will not be considered "finished" until all terminal tasks have completed.
Only then will state be determined, using the reference tasks.
Returns:
- set of Task objects which are the reference tasks in the flow
"""
if self._reference_tasks:
return set(self._reference_tasks)
else:
return self._default_reference_tasks()
def set_reference_tasks(self, tasks: Iterable[Task]) -> None:
"""
Sets the `reference_tasks` for the flow. See `flow.reference_tasks` for more details.
Args:
- tasks ([Task]): the tasks that should be set as a flow's reference tasks
Returns:
- None
"""
self._cache.clear()
reference_tasks = set(tasks)
if any(t not in self.tasks for t in reference_tasks):
raise ValueError("reference tasks must be part of the flow.")
self._reference_tasks = reference_tasks
# Graph --------------------------------------------------------------------
def _generate_task_slug(self, task: Task) -> str:
"""
Given a Task, generates the corresponding slug for this Flow. Slugs are the unique IDs
the Prefect API uses to track state updates for each task within a Flow.
The logic for slug generation within a Flow is essentially "name-tags-#count". Having
slugs be properties of (Task, Flow) instead of just Task alone allows Prefect to ensure
that every time you run the same build script for a Flow, the Task slugs remain
constant.
Args:
- task (Task): the task to generate a slug for
Returns:
- str: the corresponding slug
"""
parts = [task.name]
parts.extend(sorted(task.tags))
prefix = "-".join(parts)
while True:
ind = next(self._slug_counters[prefix])
slug = f"{prefix}-{ind}"
if slug not in self.slugs.values():
return slug
def add_task(self, task: Task) -> Task:
"""
Add a task to the flow if the task does not already exist. The tasks are
uniquely identified by their `slug`.
Args:
- task (Task): the new Task to be added to the flow
Returns:
- Task: the `Task` object passed in if the task was successfully added
Raises:
- TypeError: if the `task` is not of type `Task`
- ValueError: if the `task.slug` matches that of a task already in the flow
"""
if not isinstance(task, Task):
raise TypeError(
"Tasks must be Task instances (received {})".format(type(task))
)
elif task not in self.tasks:
if task.slug and task.slug in self.slugs.values():
raise ValueError(
'A task with the slug "{}" already exists in this '
"flow.".format(task.slug)
)
self.slugs[task] = task.slug or self._generate_task_slug(task)
self.tasks.add(task)
self._cache.clear()
# Parameters and constants must be root tasks
# All other new tasks should be added to the current case/resource (if any)
if not isinstance(task, (Parameter, prefect.tasks.core.constants.Constant)):
case = prefect.context.get("case", None)
if case is not None:
case.add_task(task, self)
resource = prefect.context.get("resource", None)
if resource is not None:
resource.add_task(task, self)
return task
def add_edge(
self,
upstream_task: Any,
downstream_task: Any,
key: str = None,
mapped: bool = False,
flattened: bool = False,
validate: bool = None,
) -> Edge:
"""
Add an edge in the flow between two tasks. All edges are directed beginning with
an upstream task and ending with a downstream task.
Args:
- upstream_task (Any): The task that the edge should start from. If
it is not a `Task`, it will be converted into one.
- downstream_task (Any): The task that the edge should end with. If
it is not a `Task`, it will be converted into one.
- key (str, optional): The key to be set for the new edge; the result of the
upstream task will be passed to the downstream task's `run()` method under this
keyword argument
- mapped (bool, optional): Whether this edge represents a call to `Task.map()`;
defaults to `False`
- flattened (bool, optional): Whether the upstream task result is flattened
- validate (bool, optional): Whether or not to check the validity of the flow
(e.g., presence of cycles and illegal keys). Defaults to the value of
`eager_edge_validation` in your prefect configuration file.
Returns:
- prefect.core.edge.Edge: The `Edge` object that was successfully added to the flow
Raises:
- ValueError: if the `downstream_task` is of type `Parameter`
- ValueError: if the edge exists with this `key` and `downstream_task`
"""
if validate is None:
validate = cast(bool, prefect.config.flows.eager_edge_validation)
if mapped and prefect.context.get("mapped", False):
raise ValueError(
"Cannot set `mapped=True` when running from inside a mapped context"
)
if isinstance(downstream_task, Parameter):
raise ValueError(
"Parameters must be root tasks and can not have upstream dependencies."
)
edge = Edge(
upstream_task=upstream_task,
downstream_task=downstream_task,
key=key,
mapped=mapped,
flattened=flattened,
flow=self,
)
# if the edge represents a keyed, unmapped constant, then we can optimize it
# out of the graph and into the special `constants` dict. We still return the edge
# object as a description of the relationship.
if (
isinstance(edge.upstream_task, prefect.tasks.core.constants.Constant)
and edge.key
and not edge.mapped
and not edge.flattened
):
self.constants[edge.downstream_task].update(
{edge.key: edge.upstream_task.value}
)
self.add_task(edge.downstream_task)
return edge
# add the edge
self.edges.add(edge)
# add the tasks from the edge (note they may be different than the passed tasks)
# due to calling `as_task()` inside the Edge constructor
self.add_task(edge.upstream_task)
self.add_task(edge.downstream_task)
# we can only check the downstream task's edges once it has been added to the
# flow, so we need to perform this check here and not earlier.
if (
validate
and key is not None
and key in {e.key for e in self.edges_to(downstream_task) if e is not edge}
):
raise ValueError(
'Argument "{a}" for task {t} has already been assigned in '
"this flow. If you are trying to call the task again with "
"new arguments, call Task.copy() before adding the result "
"to this flow.".format(a=key, t=downstream_task)
)
# check that the edges are valid keywords by binding them
if validate and key is not None:
edge_keys = {
e.key: None for e in self.edges_to(downstream_task) if e.key is not None
}
inspect.signature(downstream_task.run).bind_partial(**edge_keys)
self._cache.clear()
# check for cycles
if validate:
self.validate()
return edge
def chain(self, *tasks: Task, validate: bool = None) -> List[Edge]:
"""
Adds a sequence of dependent tasks to the flow; each task should be provided
as an argument (or splatted from a list).
Args:
- *tasks (list): A list of tasks to chain together
- validate (bool, optional): Whether or not to check the validity of
the flow (e.g., presence of cycles). Defaults to the value of
`eager_edge_validation` in your prefect configuration file.
Returns:
- A list of Edge objects added to the flow
"""
edges = []
for u_task, d_task in zip(tasks, tasks[1:]):
edges.append(
self.add_edge(
upstream_task=u_task, downstream_task=d_task, validate=validate
)
)
return edges
def update(
self,
flow: "Flow",
merge_parameters: bool = False,
validate: bool = None,
merge_reference_tasks: bool = False,
) -> None:
"""
Take all tasks and edges in another flow and add it to this flow.
When `merge_parameters` is set to`True` -- Duplicate parameters in the input `flow`
are replaced with those in the flow being updated.
Args:
- flow (Flow): A flow which is used to update this flow.
- merge_parameters (bool, False): If `True`, duplicate parameters are replaced
with parameters from the provided flow. Defaults to `False`.
If `True`, validate will also be set to `True`.
- validate (bool, optional): Whether or not to check the validity of the flow.
- merge_reference_tasks(bool, False): If `True`, add reference tasks from the provided
flow to the current flow reference tasks set.
Returns:
- None
"""
if merge_parameters:
validate = True
new_parameters = {p.name: p for p in flow.parameters()}
for p in self.parameters():
if p.name in new_parameters:
self.replace(p, new_parameters[p.name])
for task in flow.tasks:
if task not in self.tasks:
self.add_task(task)
for edge in flow.edges:
if edge not in self.edges:
self.add_edge(
upstream_task=edge.upstream_task,
downstream_task=edge.downstream_task,
key=edge.key,
mapped=edge.mapped,
flattened=edge.flattened,
validate=validate,
)
if merge_reference_tasks:
self.set_reference_tasks(
self.reference_tasks().union(flow.reference_tasks())
)
self.constants.update(flow.constants or {})
@cache
def all_upstream_edges(self) -> Dict[Task, Set[Edge]]:
"""
Returns a dictionary relating each task in the Flow to the set of
all _upstream_ edges for the task
Returns:
- dict with the key as tasks and the value as a set of upstream edges
"""
edges = {t: set() for t in self.tasks} # type: Dict[Task, Set[Edge]]
for edge in self.edges:
edges[edge.downstream_task].add(edge)
return edges
@cache
def all_downstream_edges(self) -> Dict[Task, Set[Edge]]:
"""
Returns a dictionary relating each task in the Flow to the set of
all _downstream_ edges for the task
Returns:
- dict with the key as tasks and the value as a set of downstream edges
"""
edges = {t: set() for t in self.tasks} # type: Dict[Task, Set[Edge]]
for edge in self.edges:
edges[edge.upstream_task].add(edge)
return edges
def edges_to(self, task: Task) -> Set[Edge]:
"""
Get all of the edges leading to a task (i.e., the upstream edges)
Args:
- task (Task): The task that we want to find edges leading to
Returns:
- Set: set of all edges leading from that task
Raises:
- ValueError: if `task` is not found in this flow
"""
if task not in self.tasks:
raise ValueError(
"Task {t} was not found in Flow {f}".format(t=task, f=self)
)
return self.all_upstream_edges()[task]
def edges_from(self, task: Task) -> Set[Edge]:
"""
Get all of the edges leading from a task (i.e., the downstream edges)
Args:
- task (Task): The task that we want to find edges leading from
Returns:
- Set: set of all edges leading from that task
Raises:
- ValueError: if `task` is not found in this flow
"""
if task not in self.tasks:
raise ValueError(
"Task {t} was not found in Flow {f}".format(t=task, f=self)
)
return self.all_downstream_edges()[task]
def upstream_tasks(self, task: Task) -> Set[Task]:
"""
Get all of the tasks upstream of a task
Args:
- task (Task): The task that we want to find upstream tasks of
Returns:
- set of Task objects which are upstream of `task`
"""
return set(e.upstream_task for e in self.edges_to(task))
def downstream_tasks(self, task: Task) -> Set[Task]:
"""
Get all of the tasks downstream of a task
Args:
- task (Task): The task that we want to find downstream tasks from
Returns:
- set of Task objects which are downstream of `task`
"""
return set(e.downstream_task for e in self.edges_from(task))
def validate(self) -> None:
"""
Checks that the flow is valid.
Returns:
- None
Raises:
- ValueError: if edges refer to tasks that are not in this flow
- ValueError: if specified reference tasks are not in this flow
- ValueError: if any tasks do not have assigned IDs
"""
self._cache.clear()
if any(e.upstream_task not in self.tasks for e in self.edges) or any(
e.downstream_task not in self.tasks for e in self.edges
):
raise ValueError("Some edges refer to tasks not contained in this flow.")
self.sorted_tasks()
if any(t not in self.tasks for t in self.reference_tasks()):
raise ValueError("Some reference tasks are not contained in this flow.")
def sorted_tasks(self, root_tasks: Iterable[Task] = None) -> Tuple[Task, ...]:
"""
Get the tasks in this flow in a sorted manner. This allows us to find if any
cycles exist in this flow's DAG.
Args:
- root_tasks ([Tasks], optional): an `Iterable` of `Task` objects to
start the sorting from
Returns:
- tuple of task objects that were sorted
Raises:
- ValueError: if a cycle is found in the flow's DAG
"""
return self._sorted_tasks(root_tasks=tuple(root_tasks or []))
@cache
def _sorted_tasks(self, root_tasks: Tuple[Task, ...] = None) -> Tuple[Task, ...]:
"""
Computes a topological sort of the flow's tasks.
Flow.sorted_tasks() can accept non-hashable arguments and therefore can't be
cached, so this private method is called and cached instead.
"""
# begin by getting all tasks under consideration (root tasks and all
# downstream tasks)
if root_tasks:
# double check all root tasks exist in the flow
for task in root_tasks:
if task not in self.tasks:
raise ValueError(
"Task {t} was not found in Flow {f}".format(t=task, f=self)
)
tasks = set(root_tasks)
seen = set() # type: Set[Task]
# compute the downstream edges dict once, this method uses
# @cached but validation is expensive for large flows
downstream_edges = self.all_downstream_edges()
# while the set of tasks is different from the seen tasks...
while tasks.difference(seen):
# iterate over the new tasks...
for t in list(tasks.difference(seen)):
# add its downstream tasks to the task list
tasks.update({e.downstream_task for e in downstream_edges[t]})
# mark it as seen
seen.add(t)
else:
tasks = self.tasks
# build the list of sorted tasks
remaining_tasks = list(tasks)
sorted_tasks = []
# compute the upstream edges dict once, this method uses
# @cached but validation is expensive for large flows
upstream_edges = self.all_upstream_edges()
while remaining_tasks:
# mark the flow as cyclic unless we prove otherwise
cyclic = True
# iterate over each remaining task
for task in remaining_tasks.copy():
# check all the upstream tasks of that task
upstream_tasks = {e.upstream_task for e in upstream_edges[task]}
for upstream_task in upstream_tasks:
# if the upstream task is also remaining, it means it
# hasn't been sorted, so we can't sort this task either
if upstream_task in remaining_tasks:
break
else:
# but if all upstream tasks have been sorted, we can sort
# this one too. We note that we found no cycle this time.
cyclic = False
remaining_tasks.remove(task)
sorted_tasks.append(task)
# if we were unable to match any upstream tasks, we have a cycle
if cyclic:
raise ValueError("Cycle found; flows must be acyclic!")
return tuple(sorted_tasks)
# Dependencies ------------------------------------------------------------
def set_dependencies(
self,
task: Any,
upstream_tasks: Iterable[Any] = None,
downstream_tasks: Iterable[Any] = None,
keyword_tasks: Mapping[str, Any] = None,
mapped: bool = False,
validate: bool = None,
) -> None:
"""
Convenience function for adding task dependencies.
Args:
- task (Any): a Task that will become part of the Flow. If the task is not a
Task subclass, Prefect will attempt to convert it to one.
- upstream_tasks ([Any], optional): Tasks that will run before the task runs. If
any task is not a Task subclass, Prefect will attempt to convert it to one.
- downstream_tasks ([Any], optional): Tasks that will run after the task runs.
If any task is not a Task subclass, Prefect will attempt to convert it to one.
- keyword_tasks ({key: Any}, optional): The results of these tasks
will be provided to the task under the specified keyword
arguments. If any task is not a Task subclass, Prefect will attempt to
convert it to one.
- mapped (bool, optional): Whether the upstream tasks (both keyed
and non-keyed) should be mapped over; defaults to `False`. If `True`, any
tasks wrapped in the `prefect.utilities.edges.unmapped` container will
_not_ be mapped over.
- validate (bool, optional): Whether or not to check the validity of the flow
(e.g., presence of cycles). Defaults to the value of `eager_edge_validation`
in your Prefect configuration file.
Returns:
- None
"""
if mapped and prefect.context.get("mapped", False):
raise ValueError(
"Cannot set `mapped=True` when running from inside a mapped context"
)
task = as_task(task, flow=self)
assert isinstance(task, Task) # mypy assert
# add the main task (in case it was called with no arguments)
self.add_task(task)
# add upstream tasks
for t in upstream_tasks or []:
self.add_edge(
upstream_task=t, downstream_task=task, mapped=mapped, validate=validate
)
# add downstream tasks
for t in downstream_tasks or []:
self.add_edge(upstream_task=task, downstream_task=t, validate=validate)
# add data edges to upstream tasks
for key, t in (keyword_tasks or {}).items():
self.add_edge(
upstream_task=t,