Coverage for python/lsst/pipe/base/graph/graph.py: 19%
402 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-08-06 02:28 +0000
« prev ^ index » next coverage.py v7.2.7, created at 2023-08-06 02:28 +0000
1# This file is part of pipe_base.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <http://www.gnu.org/licenses/>.
21from __future__ import annotations
23__all__ = ("QuantumGraph", "IncompatibleGraphError")
25import io
26import json
27import lzma
28import os
29import struct
30import time
31import uuid
32from collections import defaultdict, deque
33from collections.abc import Generator, Iterable, Mapping, MutableMapping
34from itertools import chain
35from types import MappingProxyType
36from typing import Any, BinaryIO, TypeVar
38import networkx as nx
39from lsst.daf.butler import DatasetRef, DatasetType, DimensionRecordsAccumulator, DimensionUniverse, Quantum
40from lsst.resources import ResourcePath, ResourcePathExpression
41from lsst.utils.introspection import get_full_type_name
42from networkx.drawing.nx_agraph import write_dot
44from ..connections import iterConnections
45from ..pipeline import TaskDef
46from ._implDetails import DatasetTypeName, _DatasetTracker, _pruner
47from ._loadHelpers import LoadHelper
48from ._versionDeserializers import DESERIALIZER_MAP
49from .quantumNode import BuildId, QuantumNode
51_T = TypeVar("_T", bound="QuantumGraph")
53# modify this constant any time the on disk representation of the save file
54# changes, and update the load helpers to behave properly for each version.
55SAVE_VERSION = 3
57# Strings used to describe the format for the preamble bytes in a file save
58# The base is a big endian encoded unsigned short that is used to hold the
59# file format version. This allows reading version bytes and determine which
60# loading code should be used for the rest of the file
61STRUCT_FMT_BASE = ">H"
62#
63# Version 1
64# This marks a big endian encoded format with an unsigned short, an unsigned
65# long long, and an unsigned long long in the byte stream
66# Version 2
67# A big endian encoded format with an unsigned long long byte stream used to
68# indicate the total length of the entire header.
69STRUCT_FMT_STRING = {1: ">QQ", 2: ">Q"}
71# magic bytes that help determine this is a graph save
72MAGIC_BYTES = b"qgraph4\xf6\xe8\xa9"
75class IncompatibleGraphError(Exception):
76 """Exception class to indicate that a lookup by NodeId is impossible due
77 to incompatibilities
78 """
80 pass
83class QuantumGraph:
84 """QuantumGraph is a directed acyclic graph of `QuantumNode` objects
86 This data structure represents a concrete workflow generated from a
87 `Pipeline`.
89 Parameters
90 ----------
91 quanta : `~collections.abc.Mapping` [ `TaskDef`, \
92 `set` [ `~lsst.daf.butler.Quantum` ] ]
93 This maps tasks (and their configs) to the sets of data they are to
94 process.
95 metadata : Optional `~collections.abc.Mapping` of `str` to primitives
96 This is an optional parameter of extra data to carry with the graph.
97 Entries in this mapping should be able to be serialized in JSON.
98 pruneRefs : iterable [ `~lsst.daf.butler.DatasetRef` ], optional
99 Set of dataset refs to exclude from a graph.
100 universe : `~lsst.daf.butler.DimensionUniverse`, optional
101 The dimensions in which quanta can be defined. Need only be provided if
102 no quanta have data IDs.
103 initInputs : `~collections.abc.Mapping`, optional
104 Maps tasks to their InitInput dataset refs. Dataset refs can be either
105 resolved or non-resolved. Presently the same dataset refs are included
106 in each `~lsst.daf.butler.Quantum` for the same task.
107 initOutputs : `~collections.abc.Mapping`, optional
108 Maps tasks to their InitOutput dataset refs. Dataset refs can be either
109 resolved or non-resolved. For intermediate resolved refs their dataset
110 ID must match ``initInputs`` and Quantum ``initInputs``.
111 globalInitOutputs : iterable [ `~lsst.daf.butler.DatasetRef` ], optional
112 Dataset refs for some global objects produced by pipeline. These
113 objects include task configurations and package versions. Typically
114 they have an empty DataId, but there is no real restriction on what
115 can appear here.
116 registryDatasetTypes : iterable [ `~lsst.daf.butler.DatasetType` ], \
117 optional
118 Dataset types which are used by this graph, their definitions must
119 match registry. If registry does not define dataset type yet, then
120 it should match one that will be created later.
122 Raises
123 ------
124 ValueError
125 Raised if the graph is pruned such that some tasks no longer have nodes
126 associated with them.
127 """
129 def __init__(
130 self,
131 quanta: Mapping[TaskDef, set[Quantum]],
132 metadata: Mapping[str, Any] | None = None,
133 pruneRefs: Iterable[DatasetRef] | None = None,
134 universe: DimensionUniverse | None = None,
135 initInputs: Mapping[TaskDef, Iterable[DatasetRef]] | None = None,
136 initOutputs: Mapping[TaskDef, Iterable[DatasetRef]] | None = None,
137 globalInitOutputs: Iterable[DatasetRef] | None = None,
138 registryDatasetTypes: Iterable[DatasetType] | None = None,
139 ):
140 self._buildGraphs(
141 quanta,
142 metadata=metadata,
143 pruneRefs=pruneRefs,
144 universe=universe,
145 initInputs=initInputs,
146 initOutputs=initOutputs,
147 globalInitOutputs=globalInitOutputs,
148 registryDatasetTypes=registryDatasetTypes,
149 )
151 def _buildGraphs(
152 self,
153 quanta: Mapping[TaskDef, set[Quantum]],
154 *,
155 _quantumToNodeId: Mapping[Quantum, uuid.UUID] | None = None,
156 _buildId: BuildId | None = None,
157 metadata: Mapping[str, Any] | None = None,
158 pruneRefs: Iterable[DatasetRef] | None = None,
159 universe: DimensionUniverse | None = None,
160 initInputs: Mapping[TaskDef, Iterable[DatasetRef]] | None = None,
161 initOutputs: Mapping[TaskDef, Iterable[DatasetRef]] | None = None,
162 globalInitOutputs: Iterable[DatasetRef] | None = None,
163 registryDatasetTypes: Iterable[DatasetType] | None = None,
164 ) -> None:
165 """Build the graph that is used to store the relation between tasks,
166 and the graph that holds the relations between quanta
167 """
168 self._metadata = metadata
169 self._buildId = _buildId if _buildId is not None else BuildId(f"{time.time()}-{os.getpid()}")
170 # Data structures used to identify relations between components;
171 # DatasetTypeName -> TaskDef for task,
172 # and DatasetRef -> QuantumNode for the quanta
173 self._datasetDict = _DatasetTracker[DatasetTypeName, TaskDef](createInverse=True)
174 self._datasetRefDict = _DatasetTracker[DatasetRef, QuantumNode]()
176 self._nodeIdMap: dict[uuid.UUID, QuantumNode] = {}
177 self._taskToQuantumNode: MutableMapping[TaskDef, set[QuantumNode]] = defaultdict(set)
178 for taskDef, quantumSet in quanta.items():
179 connections = taskDef.connections
181 # For each type of connection in the task, add a key to the
182 # `_DatasetTracker` for the connections name, with a value of
183 # the TaskDef in the appropriate field
184 for inpt in iterConnections(connections, ("inputs", "prerequisiteInputs", "initInputs")):
185 # Have to handle components in inputs.
186 dataset_name, _, _ = inpt.name.partition(".")
187 self._datasetDict.addConsumer(DatasetTypeName(dataset_name), taskDef)
189 for output in iterConnections(connections, ("outputs",)):
190 # Have to handle possible components in outputs.
191 dataset_name, _, _ = output.name.partition(".")
192 self._datasetDict.addProducer(DatasetTypeName(dataset_name), taskDef)
194 # For each `Quantum` in the set of all `Quantum` for this task,
195 # add a key to the `_DatasetTracker` that is a `DatasetRef` for one
196 # of the individual datasets inside the `Quantum`, with a value of
197 # a newly created QuantumNode to the appropriate input/output
198 # field.
199 for quantum in quantumSet:
200 if quantum.dataId is not None:
201 if universe is None:
202 universe = quantum.dataId.universe
203 elif universe != quantum.dataId.universe:
204 raise RuntimeError(
205 "Mismatched dimension universes in QuantumGraph construction: "
206 f"{universe} != {quantum.dataId.universe}. "
207 )
209 if _quantumToNodeId:
210 if (nodeId := _quantumToNodeId.get(quantum)) is None:
211 raise ValueError(
212 "If _quantuMToNodeNumber is not None, all quanta must have an "
213 "associated value in the mapping"
214 )
215 else:
216 nodeId = uuid.uuid4()
218 inits = quantum.initInputs.values()
219 inputs = quantum.inputs.values()
220 value = QuantumNode(quantum, taskDef, nodeId)
221 self._taskToQuantumNode[taskDef].add(value)
222 self._nodeIdMap[nodeId] = value
224 for dsRef in chain(inits, inputs):
225 # unfortunately, `Quantum` allows inits to be individual
226 # `DatasetRef`s or an Iterable of such, so there must
227 # be an instance check here
228 if isinstance(dsRef, Iterable):
229 for sub in dsRef:
230 if sub.isComponent():
231 sub = sub.makeCompositeRef()
232 self._datasetRefDict.addConsumer(sub, value)
233 else:
234 assert isinstance(dsRef, DatasetRef)
235 if dsRef.isComponent():
236 dsRef = dsRef.makeCompositeRef()
237 self._datasetRefDict.addConsumer(dsRef, value)
238 for dsRef in chain.from_iterable(quantum.outputs.values()):
239 self._datasetRefDict.addProducer(dsRef, value)
241 if pruneRefs is not None:
242 # track what refs were pruned and prune the graph
243 prunes: set[QuantumNode] = set()
244 _pruner(self._datasetRefDict, pruneRefs, alreadyPruned=prunes)
246 # recreate the taskToQuantumNode dict removing nodes that have been
247 # pruned. Keep track of task defs that now have no QuantumNodes
248 emptyTasks: set[str] = set()
249 newTaskToQuantumNode: defaultdict[TaskDef, set[QuantumNode]] = defaultdict(set)
250 # accumulate all types
251 types_ = set()
252 # tracker for any pruneRefs that have caused tasks to have no nodes
253 # This helps the user find out what caused the issues seen.
254 culprits = set()
255 # Find all the types from the refs to prune
256 for r in pruneRefs:
257 types_.add(r.datasetType)
259 # For each of the tasks, and their associated nodes, remove any
260 # any nodes that were pruned. If there are no nodes associated
261 # with a task, record that task, and find out if that was due to
262 # a type from an input ref to prune.
263 for td, taskNodes in self._taskToQuantumNode.items():
264 diff = taskNodes.difference(prunes)
265 if len(diff) == 0:
266 if len(taskNodes) != 0:
267 tp: DatasetType
268 for tp in types_:
269 if (tmpRefs := next(iter(taskNodes)).quantum.inputs.get(tp)) and not set(
270 tmpRefs
271 ).difference(pruneRefs):
272 culprits.add(tp.name)
273 emptyTasks.add(td.label)
274 newTaskToQuantumNode[td] = diff
276 # update the internal dict
277 self._taskToQuantumNode = newTaskToQuantumNode
279 if emptyTasks:
280 raise ValueError(
281 f"{', '.join(emptyTasks)} task(s) have no nodes associated with them "
282 f"after graph pruning; {', '.join(culprits)} caused over-pruning"
283 )
285 # Dimension universe
286 if universe is None:
287 raise RuntimeError(
288 "Dimension universe or at least one quantum with a data ID "
289 "must be provided when constructing a QuantumGraph."
290 )
291 self._universe = universe
293 # Graph of quanta relations
294 self._connectedQuanta = self._datasetRefDict.makeNetworkXGraph()
295 self._count = len(self._connectedQuanta)
297 # Graph of task relations, used in various methods
298 self._taskGraph = self._datasetDict.makeNetworkXGraph()
300 # convert default dict into a regular to prevent accidental key
301 # insertion
302 self._taskToQuantumNode = dict(self._taskToQuantumNode.items())
304 self._initInputRefs: dict[TaskDef, list[DatasetRef]] = {}
305 self._initOutputRefs: dict[TaskDef, list[DatasetRef]] = {}
306 self._globalInitOutputRefs: list[DatasetRef] = []
307 self._registryDatasetTypes: list[DatasetType] = []
308 if initInputs is not None:
309 self._initInputRefs = {taskDef: list(refs) for taskDef, refs in initInputs.items()}
310 if initOutputs is not None:
311 self._initOutputRefs = {taskDef: list(refs) for taskDef, refs in initOutputs.items()}
312 if globalInitOutputs is not None:
313 self._globalInitOutputRefs = list(globalInitOutputs)
314 if registryDatasetTypes is not None:
315 self._registryDatasetTypes = list(registryDatasetTypes)
317 @property
318 def taskGraph(self) -> nx.DiGraph:
319 """A graph representing the relations between the tasks inside
320 the quantum graph (`networkx.DiGraph`).
321 """
322 return self._taskGraph
324 @property
325 def graph(self) -> nx.DiGraph:
326 """A graph representing the relations between all the `QuantumNode`
327 objects (`networkx.DiGraph`).
329 The graph should usually be iterated over, or passed to methods of this
330 class, but sometimes direct access to the ``networkx`` object may be
331 helpful.
332 """
333 return self._connectedQuanta
335 @property
336 def inputQuanta(self) -> Iterable[QuantumNode]:
337 """The nodes that are inputs to the graph (iterable [`QuantumNode`]).
339 These are the nodes that do not depend on any other nodes in the
340 graph.
341 """
342 return (q for q, n in self._connectedQuanta.in_degree if n == 0)
344 @property
345 def outputQuanta(self) -> Iterable[QuantumNode]:
346 """The nodes that are outputs of the graph (iterable [`QuantumNode`]).
348 These are the nodes that have no nodes that depend on them in the
349 graph.
350 """
351 return [q for q, n in self._connectedQuanta.out_degree if n == 0]
353 @property
354 def allDatasetTypes(self) -> tuple[DatasetTypeName, ...]:
355 """All the data set type names that are present in the graph
356 (`tuple` [`str`]).
358 These types do not include global init-outputs.
359 """
360 return tuple(self._datasetDict.keys())
362 @property
363 def isConnected(self) -> bool:
364 """Whether all of the nodes in the graph are connected, ignoring
365 directionality of connections (`bool`).
366 """
367 return nx.is_weakly_connected(self._connectedQuanta)
369 def pruneGraphFromRefs(self: _T, refs: Iterable[DatasetRef]) -> _T:
370 r"""Return a graph pruned of input `~lsst.daf.butler.DatasetRef`\ s
371 and nodes which depend on them.
373 Parameters
374 ----------
375 refs : `~collections.abc.Iterable` of `~lsst.daf.butler.DatasetRef`
376 Refs which should be removed from resulting graph
378 Returns
379 -------
380 graph : `QuantumGraph`
381 A graph that has been pruned of specified refs and the nodes that
382 depend on them.
383 """
384 newInst = object.__new__(type(self))
385 quantumMap = defaultdict(set)
386 for node in self:
387 quantumMap[node.taskDef].add(node.quantum)
389 # convert to standard dict to prevent accidental key insertion
390 quantumDict: dict[TaskDef, set[Quantum]] = dict(quantumMap.items())
392 # This should not change set of tasks in a graph, so we can keep the
393 # same registryDatasetTypes as in the original graph.
394 # TODO: Do we need to copy initInputs/initOutputs?
395 newInst._buildGraphs(
396 quantumDict,
397 _quantumToNodeId={n.quantum: n.nodeId for n in self},
398 metadata=self._metadata,
399 pruneRefs=refs,
400 universe=self._universe,
401 globalInitOutputs=self._globalInitOutputRefs,
402 registryDatasetTypes=self._registryDatasetTypes,
403 )
404 return newInst
406 def getQuantumNodeByNodeId(self, nodeId: uuid.UUID) -> QuantumNode:
407 """Lookup a `QuantumNode` from an id associated with the node.
409 Parameters
410 ----------
411 nodeId : `NodeId`
412 The number associated with a node
414 Returns
415 -------
416 node : `QuantumNode`
417 The node corresponding with input number
419 Raises
420 ------
421 KeyError
422 Raised if the requested nodeId is not in the graph.
423 """
424 return self._nodeIdMap[nodeId]
426 def getQuantaForTask(self, taskDef: TaskDef) -> frozenset[Quantum]:
427 """Return all the `~lsst.daf.butler.Quantum` associated with a
428 `TaskDef`.
430 Parameters
431 ----------
432 taskDef : `TaskDef`
433 The `TaskDef` for which `~lsst.daf.butler.Quantum` are to be
434 queried.
436 Returns
437 -------
438 quanta : `frozenset` of `~lsst.daf.butler.Quantum`
439 The `set` of `~lsst.daf.butler.Quantum` that is associated with the
440 specified `TaskDef`.
441 """
442 return frozenset(node.quantum for node in self._taskToQuantumNode.get(taskDef, ()))
444 def getNumberOfQuantaForTask(self, taskDef: TaskDef) -> int:
445 """Return the number of `~lsst.daf.butler.Quantum` associated with
446 a `TaskDef`.
448 Parameters
449 ----------
450 taskDef : `TaskDef`
451 The `TaskDef` for which `~lsst.daf.butler.Quantum` are to be
452 queried.
454 Returns
455 -------
456 count : `int`
457 The number of `~lsst.daf.butler.Quantum` that are associated with
458 the specified `TaskDef`.
459 """
460 return len(self._taskToQuantumNode.get(taskDef, ()))
462 def getNodesForTask(self, taskDef: TaskDef) -> frozenset[QuantumNode]:
463 r"""Return all the `QuantumNode`\s associated with a `TaskDef`.
465 Parameters
466 ----------
467 taskDef : `TaskDef`
468 The `TaskDef` for which `~lsst.daf.butler.Quantum` are to be
469 queried.
471 Returns
472 -------
473 nodes : `frozenset` [ `QuantumNode` ]
474 A `frozenset` of `QuantumNode` that is associated with the
475 specified `TaskDef`.
476 """
477 return frozenset(self._taskToQuantumNode[taskDef])
479 def findTasksWithInput(self, datasetTypeName: DatasetTypeName) -> Iterable[TaskDef]:
480 """Find all tasks that have the specified dataset type name as an
481 input.
483 Parameters
484 ----------
485 datasetTypeName : `str`
486 A string representing the name of a dataset type to be queried,
487 can also accept a `DatasetTypeName` which is a `~typing.NewType` of
488 `str` for type safety in static type checking.
490 Returns
491 -------
492 tasks : iterable of `TaskDef`
493 `TaskDef` objects that have the specified `DatasetTypeName` as an
494 input, list will be empty if no tasks use specified
495 `DatasetTypeName` as an input.
497 Raises
498 ------
499 KeyError
500 Raised if the `DatasetTypeName` is not part of the `QuantumGraph`.
501 """
502 return (c for c in self._datasetDict.getConsumers(datasetTypeName))
504 def findTaskWithOutput(self, datasetTypeName: DatasetTypeName) -> TaskDef | None:
505 """Find all tasks that have the specified dataset type name as an
506 output.
508 Parameters
509 ----------
510 datasetTypeName : `str`
511 A string representing the name of a dataset type to be queried,
512 can also accept a `DatasetTypeName` which is a `~typing.NewType` of
513 `str` for type safety in static type checking.
515 Returns
516 -------
517 result : `TaskDef` or `None`
518 `TaskDef` that outputs `DatasetTypeName` as an output or `None` if
519 none of the tasks produce this `DatasetTypeName`.
521 Raises
522 ------
523 KeyError
524 Raised if the `DatasetTypeName` is not part of the `QuantumGraph`.
525 """
526 return self._datasetDict.getProducer(datasetTypeName)
528 def tasksWithDSType(self, datasetTypeName: DatasetTypeName) -> Iterable[TaskDef]:
529 """Find all tasks that are associated with the specified dataset type
530 name.
532 Parameters
533 ----------
534 datasetTypeName : `str`
535 A string representing the name of a dataset type to be queried,
536 can also accept a `DatasetTypeName` which is a `~typing.NewType` of
537 `str` for type safety in static type checking.
539 Returns
540 -------
541 result : iterable of `TaskDef`
542 `TaskDef` objects that are associated with the specified
543 `DatasetTypeName`.
545 Raises
546 ------
547 KeyError
548 Raised if the `DatasetTypeName` is not part of the `QuantumGraph`.
549 """
550 return self._datasetDict.getAll(datasetTypeName)
552 def findTaskDefByName(self, taskName: str) -> list[TaskDef]:
553 """Determine which `TaskDef` objects in this graph are associated
554 with a `str` representing a task name (looks at the ``taskName``
555 property of `TaskDef` objects).
557 Returns a list of `TaskDef` objects as a `PipelineTask` may appear
558 multiple times in a graph with different labels.
560 Parameters
561 ----------
562 taskName : `str`
563 Name of a task to search for.
565 Returns
566 -------
567 result : `list` of `TaskDef`
568 List of the `TaskDef` objects that have the name specified.
569 Multiple values are returned in the case that a task is used
570 multiple times with different labels.
571 """
572 results = []
573 for task in self._taskToQuantumNode:
574 split = task.taskName.split(".")
575 if split[-1] == taskName:
576 results.append(task)
577 return results
579 def findTaskDefByLabel(self, label: str) -> TaskDef | None:
580 """Determine which `TaskDef` objects in this graph are associated
581 with a `str` representing a tasks label.
583 Parameters
584 ----------
585 taskName : `str`
586 Name of a task to search for
588 Returns
589 -------
590 result : `TaskDef`
591 `TaskDef` objects that has the specified label.
592 """
593 for task in self._taskToQuantumNode:
594 if label == task.label:
595 return task
596 return None
598 def findQuantaWithDSType(self, datasetTypeName: DatasetTypeName) -> set[Quantum]:
599 r"""Return all the `~lsst.daf.butler.Quantum` that contain a specified
600 `DatasetTypeName`.
602 Parameters
603 ----------
604 datasetTypeName : `str`
605 The name of the dataset type to search for as a string,
606 can also accept a `DatasetTypeName` which is a `~typing.NewType` of
607 `str` for type safety in static type checking.
609 Returns
610 -------
611 result : `set` of `QuantumNode` objects
612 A `set` of `QuantumNode`\s that contain specified
613 `DatasetTypeName`.
615 Raises
616 ------
617 KeyError
618 Raised if the `DatasetTypeName` is not part of the `QuantumGraph`
620 """
621 tasks = self._datasetDict.getAll(datasetTypeName)
622 result: set[Quantum] = set()
623 result = result.union(quantum for task in tasks for quantum in self.getQuantaForTask(task))
624 return result
626 def checkQuantumInGraph(self, quantum: Quantum) -> bool:
627 """Check if specified quantum appears in the graph as part of a node.
629 Parameters
630 ----------
631 quantum : `lsst.daf.butler.Quantum`
632 The quantum to search for.
634 Returns
635 -------
636 in_graph : `bool`
637 The result of searching for the quantum.
638 """
639 return any(quantum == node.quantum for node in self)
641 def writeDotGraph(self, output: str | io.BufferedIOBase) -> None:
642 """Write out the graph as a dot graph.
644 Parameters
645 ----------
646 output : `str` or `io.BufferedIOBase`
647 Either a filesystem path to write to, or a file handle object.
648 """
649 write_dot(self._connectedQuanta, output)
651 def subset(self: _T, nodes: QuantumNode | Iterable[QuantumNode]) -> _T:
652 """Create a new graph object that contains the subset of the nodes
653 specified as input. Node number is preserved.
655 Parameters
656 ----------
657 nodes : `QuantumNode` or iterable of `QuantumNode`
658 Nodes from which to create subset.
660 Returns
661 -------
662 graph : instance of graph type
663 An instance of the type from which the subset was created.
664 """
665 if not isinstance(nodes, Iterable):
666 nodes = (nodes,)
667 quantumSubgraph = self._connectedQuanta.subgraph(nodes).nodes
668 quantumMap = defaultdict(set)
670 dataset_type_names: set[str] = set()
671 node: QuantumNode
672 for node in quantumSubgraph:
673 quantumMap[node.taskDef].add(node.quantum)
674 dataset_type_names.update(
675 dstype.name
676 for dstype in chain(
677 node.quantum.inputs.keys(), node.quantum.outputs.keys(), node.quantum.initInputs.keys()
678 )
679 )
681 # May need to trim dataset types from registryDatasetTypes.
682 for taskDef in quantumMap:
683 if refs := self.initOutputRefs(taskDef):
684 dataset_type_names.update(ref.datasetType.name for ref in refs)
685 dataset_type_names.update(ref.datasetType.name for ref in self._globalInitOutputRefs)
686 registryDatasetTypes = [
687 dstype for dstype in self._registryDatasetTypes if dstype.name in dataset_type_names
688 ]
690 # convert to standard dict to prevent accidental key insertion
691 quantumDict: dict[TaskDef, set[Quantum]] = dict(quantumMap.items())
692 # Create an empty graph, and then populate it with custom mapping
693 newInst = type(self)({}, universe=self._universe)
694 # TODO: Do we need to copy initInputs/initOutputs?
695 newInst._buildGraphs(
696 quantumDict,
697 _quantumToNodeId={n.quantum: n.nodeId for n in nodes},
698 _buildId=self._buildId,
699 metadata=self._metadata,
700 universe=self._universe,
701 globalInitOutputs=self._globalInitOutputRefs,
702 registryDatasetTypes=registryDatasetTypes,
703 )
704 return newInst
706 def subsetToConnected(self: _T) -> tuple[_T, ...]:
707 """Generate a list of subgraphs where each is connected.
709 Returns
710 -------
711 result : `list` of `QuantumGraph`
712 A list of graphs that are each connected.
713 """
714 return tuple(
715 self.subset(connectedSet)
716 for connectedSet in nx.weakly_connected_components(self._connectedQuanta)
717 )
719 def determineInputsToQuantumNode(self, node: QuantumNode) -> set[QuantumNode]:
720 """Return a set of `QuantumNode` that are direct inputs to a specified
721 node.
723 Parameters
724 ----------
725 node : `QuantumNode`
726 The node of the graph for which inputs are to be determined.
728 Returns
729 -------
730 inputs : `set` of `QuantumNode`
731 All the nodes that are direct inputs to specified node.
732 """
733 return set(self._connectedQuanta.predecessors(node))
735 def determineOutputsOfQuantumNode(self, node: QuantumNode) -> set[QuantumNode]:
736 """Return a set of `QuantumNode` that are direct outputs of a specified
737 node.
739 Parameters
740 ----------
741 node : `QuantumNode`
742 The node of the graph for which outputs are to be determined.
744 Returns
745 -------
746 outputs : `set` of `QuantumNode`
747 All the nodes that are direct outputs to specified node.
748 """
749 return set(self._connectedQuanta.successors(node))
751 def determineConnectionsOfQuantumNode(self: _T, node: QuantumNode) -> _T:
752 """Return a graph of `QuantumNode` that are direct inputs and outputs
753 of a specified node.
755 Parameters
756 ----------
757 node : `QuantumNode`
758 The node of the graph for which connected nodes are to be
759 determined.
761 Returns
762 -------
763 graph : graph of `QuantumNode`
764 All the nodes that are directly connected to specified node.
765 """
766 nodes = self.determineInputsToQuantumNode(node).union(self.determineOutputsOfQuantumNode(node))
767 nodes.add(node)
768 return self.subset(nodes)
770 def determineAncestorsOfQuantumNode(self: _T, node: QuantumNode) -> _T:
771 """Return a graph of the specified node and all the ancestor nodes
772 directly reachable by walking edges.
774 Parameters
775 ----------
776 node : `QuantumNode`
777 The node for which all ancestors are to be determined
779 Returns
780 -------
781 ancestors : graph of `QuantumNode`
782 Graph of node and all of its ancestors.
783 """
784 predecessorNodes = nx.ancestors(self._connectedQuanta, node)
785 predecessorNodes.add(node)
786 return self.subset(predecessorNodes)
788 def findCycle(self) -> list[tuple[QuantumNode, QuantumNode]]:
789 """Check a graph for the presense of cycles and returns the edges of
790 any cycles found, or an empty list if there is no cycle.
792 Returns
793 -------
794 result : `list` of `tuple` of [ `QuantumNode`, `QuantumNode` ]
795 A list of any graph edges that form a cycle, or an empty list if
796 there is no cycle. Empty list to so support if graph.find_cycle()
797 syntax as an empty list is falsy.
798 """
799 try:
800 return nx.find_cycle(self._connectedQuanta)
801 except nx.NetworkXNoCycle:
802 return []
804 def saveUri(self, uri: ResourcePathExpression) -> None:
805 """Save `QuantumGraph` to the specified URI.
807 Parameters
808 ----------
809 uri : convertible to `~lsst.resources.ResourcePath`
810 URI to where the graph should be saved.
811 """
812 buffer = self._buildSaveObject()
813 path = ResourcePath(uri)
814 if path.getExtension() not in (".qgraph"):
815 raise TypeError(f"Can currently only save a graph in qgraph format not {uri}")
816 path.write(buffer) # type: ignore # Ignore because bytearray is safe to use in place of bytes
818 @property
819 def metadata(self) -> MappingProxyType[str, Any] | None:
820 """Extra data carried with the graph (mapping [`str`] or `None`).
822 The mapping is a dynamic view of this object's metadata. Values should
823 be able to be serialized in JSON.
824 """
825 if self._metadata is None:
826 return None
827 return MappingProxyType(self._metadata)
829 def initInputRefs(self, taskDef: TaskDef) -> list[DatasetRef] | None:
830 """Return DatasetRefs for a given task InitInputs.
832 Parameters
833 ----------
834 taskDef : `TaskDef`
835 Task definition structure.
837 Returns
838 -------
839 refs : `list` [ `~lsst.daf.butler.DatasetRef` ] or `None`
840 DatasetRef for the task InitInput, can be `None`. This can return
841 either resolved or non-resolved reference.
842 """
843 return self._initInputRefs.get(taskDef)
845 def initOutputRefs(self, taskDef: TaskDef) -> list[DatasetRef] | None:
846 """Return DatasetRefs for a given task InitOutputs.
848 Parameters
849 ----------
850 taskDef : `TaskDef`
851 Task definition structure.
853 Returns
854 -------
855 refs : `list` [ `~lsst.daf.butler.DatasetRef` ] or `None`
856 DatasetRefs for the task InitOutput, can be `None`. This can return
857 either resolved or non-resolved reference. Resolved reference will
858 match Quantum's initInputs if this is an intermediate dataset type.
859 """
860 return self._initOutputRefs.get(taskDef)
862 def globalInitOutputRefs(self) -> list[DatasetRef]:
863 """Return DatasetRefs for global InitOutputs.
865 Returns
866 -------
867 refs : `list` [ `~lsst.daf.butler.DatasetRef` ]
868 DatasetRefs for global InitOutputs.
869 """
870 return self._globalInitOutputRefs
872 def registryDatasetTypes(self) -> list[DatasetType]:
873 """Return dataset types used by this graph, their definitions match
874 dataset types from registry.
876 Returns
877 -------
878 refs : `list` [ `~lsst.daf.butler.DatasetType` ]
879 Dataset types for this graph.
880 """
881 return self._registryDatasetTypes
883 @classmethod
884 def loadUri(
885 cls,
886 uri: ResourcePathExpression,
887 universe: DimensionUniverse | None = None,
888 nodes: Iterable[uuid.UUID] | None = None,
889 graphID: BuildId | None = None,
890 minimumVersion: int = 3,
891 ) -> QuantumGraph:
892 """Read `QuantumGraph` from a URI.
894 Parameters
895 ----------
896 uri : convertible to `~lsst.resources.ResourcePath`
897 URI from where to load the graph.
898 universe : `~lsst.daf.butler.DimensionUniverse`, optional
899 If `None` it is loaded from the `QuantumGraph`
900 saved structure. If supplied, the
901 `~lsst.daf.butler.DimensionUniverse` from the loaded `QuantumGraph`
902 will be validated against the supplied argument for compatibility.
903 nodes : iterable of `uuid.UUID` or `None`
904 UUIDs that correspond to nodes in the graph. If specified, only
905 these nodes will be loaded. Defaults to None, in which case all
906 nodes will be loaded.
907 graphID : `str` or `None`
908 If specified this ID is verified against the loaded graph prior to
909 loading any Nodes. This defaults to None in which case no
910 validation is done.
911 minimumVersion : `int`
912 Minimum version of a save file to load. Set to -1 to load all
913 versions. Older versions may need to be loaded, and re-saved
914 to upgrade them to the latest format before they can be used in
915 production.
917 Returns
918 -------
919 graph : `QuantumGraph`
920 Resulting QuantumGraph instance.
922 Raises
923 ------
924 TypeError
925 Raised if file contains instance of a type other than
926 `QuantumGraph`.
927 ValueError
928 Raised if one or more of the nodes requested is not in the
929 `QuantumGraph` or if graphID parameter does not match the graph
930 being loaded or if the supplied uri does not point at a valid
931 `QuantumGraph` save file.
932 RuntimeError
933 Raise if Supplied `~lsst.daf.butler.DimensionUniverse` is not
934 compatible with the `~lsst.daf.butler.DimensionUniverse` saved in
935 the graph.
936 """
937 uri = ResourcePath(uri)
938 if uri.getExtension() in {".qgraph"}:
939 with LoadHelper(uri, minimumVersion) as loader:
940 qgraph = loader.load(universe, nodes, graphID)
941 else:
942 raise ValueError(f"Only know how to handle files saved as `.qgraph`, not {uri}")
943 if not isinstance(qgraph, QuantumGraph):
944 raise TypeError(f"QuantumGraph file {uri} contains unexpected object type: {type(qgraph)}")
945 return qgraph
947 @classmethod
948 def readHeader(cls, uri: ResourcePathExpression, minimumVersion: int = 3) -> str | None:
949 """Read the header of a `QuantumGraph` pointed to by the uri parameter
950 and return it as a string.
952 Parameters
953 ----------
954 uri : convertible to `~lsst.resources.ResourcePath`
955 The location of the `QuantumGraph` to load. If the argument is a
956 string, it must correspond to a valid
957 `~lsst.resources.ResourcePath` path.
958 minimumVersion : `int`
959 Minimum version of a save file to load. Set to -1 to load all
960 versions. Older versions may need to be loaded, and re-saved
961 to upgrade them to the latest format before they can be used in
962 production.
964 Returns
965 -------
966 header : `str` or `None`
967 The header associated with the specified `QuantumGraph` it there is
968 one, else `None`.
970 Raises
971 ------
972 ValueError
973 Raised if the extension of the file specified by uri is not a
974 `QuantumGraph` extension.
975 """
976 uri = ResourcePath(uri)
977 if uri.getExtension() in {".qgraph"}:
978 return LoadHelper(uri, minimumVersion).readHeader()
979 else:
980 raise ValueError("Only know how to handle files saved as `.qgraph`")
982 def buildAndPrintHeader(self) -> None:
983 """Create a header that would be used in a save of this object and
984 prints it out to standard out.
985 """
986 _, header = self._buildSaveObject(returnHeader=True)
987 print(json.dumps(header))
989 def save(self, file: BinaryIO) -> None:
990 """Save QuantumGraph to a file.
992 Parameters
993 ----------
994 file : `io.BufferedIOBase`
995 File to write data open in binary mode.
996 """
997 buffer = self._buildSaveObject()
998 file.write(buffer) # type: ignore # Ignore because bytearray is safe to use in place of bytes
1000 def _buildSaveObject(self, returnHeader: bool = False) -> bytearray | tuple[bytearray, dict]:
1001 # make some containers
1002 jsonData: deque[bytes] = deque()
1003 # node map is a list because json does not accept mapping keys that
1004 # are not strings, so we store a list of key, value pairs that will
1005 # be converted to a mapping on load
1006 nodeMap = []
1007 taskDefMap = {}
1008 headerData: dict[str, Any] = {}
1010 # Store the QauntumGraph BuildId, this will allow validating BuildIds
1011 # at load time, prior to loading any QuantumNodes. Name chosen for
1012 # unlikely conflicts.
1013 headerData["GraphBuildID"] = self.graphID
1014 headerData["Metadata"] = self._metadata
1016 # Store the universe this graph was created with
1017 universeConfig = self._universe.dimensionConfig
1018 headerData["universe"] = universeConfig.toDict()
1020 # counter for the number of bytes processed thus far
1021 count = 0
1022 # serialize out the task Defs recording the start and end bytes of each
1023 # taskDef
1024 inverseLookup = self._datasetDict.inverse
1025 taskDef: TaskDef
1026 # sort by task label to ensure serialization happens in the same order
1027 for taskDef in self.taskGraph:
1028 # compressing has very little impact on saving or load time, but
1029 # a large impact on on disk size, so it is worth doing
1030 taskDescription: dict[str, Any] = {}
1031 # save the fully qualified name.
1032 taskDescription["taskName"] = get_full_type_name(taskDef.taskClass)
1033 # save the config as a text stream that will be un-persisted on the
1034 # other end
1035 stream = io.StringIO()
1036 taskDef.config.saveToStream(stream)
1037 taskDescription["config"] = stream.getvalue()
1038 taskDescription["label"] = taskDef.label
1039 if (refs := self._initInputRefs.get(taskDef)) is not None:
1040 taskDescription["initInputRefs"] = [ref.to_json() for ref in refs]
1041 if (refs := self._initOutputRefs.get(taskDef)) is not None:
1042 taskDescription["initOutputRefs"] = [ref.to_json() for ref in refs]
1044 inputs = []
1045 outputs = []
1047 # Determine the connection between all of tasks and save that in
1048 # the header as a list of connections and edges in each task
1049 # this will help in un-persisting, and possibly in a "quick view"
1050 # method that does not require everything to be un-persisted
1051 #
1052 # Typing returns can't be parameter dependent
1053 for connection in inverseLookup[taskDef]: # type: ignore
1054 consumers = self._datasetDict.getConsumers(connection)
1055 producer = self._datasetDict.getProducer(connection)
1056 if taskDef in consumers:
1057 # This checks if the task consumes the connection directly
1058 # from the datastore or it is produced by another task
1059 producerLabel = producer.label if producer is not None else "datastore"
1060 inputs.append((producerLabel, connection))
1061 elif taskDef not in consumers and producer is taskDef:
1062 # If there are no consumers for this tasks produced
1063 # connection, the output will be said to be the datastore
1064 # in which case the for loop will be a zero length loop
1065 if not consumers:
1066 outputs.append(("datastore", connection))
1067 for td in consumers:
1068 outputs.append((td.label, connection))
1070 # dump to json string, and encode that string to bytes and then
1071 # conpress those bytes
1072 dump = lzma.compress(json.dumps(taskDescription).encode())
1073 # record the sizing and relation information
1074 taskDefMap[taskDef.label] = {
1075 "bytes": (count, count + len(dump)),
1076 "inputs": inputs,
1077 "outputs": outputs,
1078 }
1079 count += len(dump)
1080 jsonData.append(dump)
1082 headerData["TaskDefs"] = taskDefMap
1084 # serialize the nodes, recording the start and end bytes of each node
1085 dimAccumulator = DimensionRecordsAccumulator()
1086 for node in self:
1087 # compressing has very little impact on saving or load time, but
1088 # a large impact on on disk size, so it is worth doing
1089 simpleNode = node.to_simple(accumulator=dimAccumulator)
1091 dump = lzma.compress(simpleNode.json().encode())
1092 jsonData.append(dump)
1093 nodeMap.append(
1094 (
1095 str(node.nodeId),
1096 {
1097 "bytes": (count, count + len(dump)),
1098 "inputs": [str(n.nodeId) for n in self.determineInputsToQuantumNode(node)],
1099 "outputs": [str(n.nodeId) for n in self.determineOutputsOfQuantumNode(node)],
1100 },
1101 )
1102 )
1103 count += len(dump)
1105 headerData["DimensionRecords"] = {
1106 key: value.model_dump()
1107 for key, value in dimAccumulator.makeSerializedDimensionRecordMapping().items()
1108 }
1110 # need to serialize this as a series of key,value tuples because of
1111 # a limitation on how json cant do anything but strings as keys
1112 headerData["Nodes"] = nodeMap
1114 if self._globalInitOutputRefs:
1115 headerData["GlobalInitOutputRefs"] = [ref.to_json() for ref in self._globalInitOutputRefs]
1117 if self._registryDatasetTypes:
1118 headerData["RegistryDatasetTypes"] = [dstype.to_json() for dstype in self._registryDatasetTypes]
1120 # dump the headerData to json
1121 header_encode = lzma.compress(json.dumps(headerData).encode())
1123 # record the sizes as 2 unsigned long long numbers for a total of 16
1124 # bytes
1125 save_bytes = struct.pack(STRUCT_FMT_BASE, SAVE_VERSION)
1127 fmt_string = DESERIALIZER_MAP[SAVE_VERSION].FMT_STRING()
1128 map_lengths = struct.pack(fmt_string, len(header_encode))
1130 # write each component of the save out in a deterministic order
1131 buffer = bytearray()
1132 buffer.extend(MAGIC_BYTES)
1133 buffer.extend(save_bytes)
1134 buffer.extend(map_lengths)
1135 buffer.extend(header_encode)
1136 # Iterate over the length of jsonData, and for each element pop the
1137 # leftmost element off the deque and write it out. This is to save
1138 # memory, as the memory is added to the buffer object, it is removed
1139 # from from the container.
1140 #
1141 # Only this section needs to worry about memory pressure because
1142 # everything else written to the buffer prior to this data is
1143 # only on the order of kilobytes to low numbers of megabytes.
1144 while jsonData:
1145 buffer.extend(jsonData.popleft())
1146 if returnHeader:
1147 return buffer, headerData
1148 else:
1149 return buffer
1151 @classmethod
1152 def load(
1153 cls,
1154 file: BinaryIO,
1155 universe: DimensionUniverse | None = None,
1156 nodes: Iterable[uuid.UUID] | None = None,
1157 graphID: BuildId | None = None,
1158 minimumVersion: int = 3,
1159 ) -> QuantumGraph:
1160 """Read `QuantumGraph` from a file that was made by `save`.
1162 Parameters
1163 ----------
1164 file : `io.IO` of bytes
1165 File with data open in binary mode.
1166 universe : `~lsst.daf.butler.DimensionUniverse`, optional
1167 If `None` it is loaded from the `QuantumGraph`
1168 saved structure. If supplied, the
1169 `~lsst.daf.butler.DimensionUniverse` from the loaded `QuantumGraph`
1170 will be validated against the supplied argument for compatibility.
1171 nodes : iterable of `uuid.UUID` or `None`
1172 UUIDs that correspond to nodes in the graph. If specified, only
1173 these nodes will be loaded. Defaults to None, in which case all
1174 nodes will be loaded.
1175 graphID : `str` or `None`
1176 If specified this ID is verified against the loaded graph prior to
1177 loading any Nodes. This defaults to None in which case no
1178 validation is done.
1179 minimumVersion : `int`
1180 Minimum version of a save file to load. Set to -1 to load all
1181 versions. Older versions may need to be loaded, and re-saved
1182 to upgrade them to the latest format before they can be used in
1183 production.
1185 Returns
1186 -------
1187 graph : `QuantumGraph`
1188 Resulting QuantumGraph instance.
1190 Raises
1191 ------
1192 TypeError
1193 Raised if data contains instance of a type other than
1194 `QuantumGraph`.
1195 ValueError
1196 Raised if one or more of the nodes requested is not in the
1197 `QuantumGraph` or if graphID parameter does not match the graph
1198 being loaded or if the supplied uri does not point at a valid
1199 `QuantumGraph` save file.
1200 """
1201 with LoadHelper(file, minimumVersion) as loader:
1202 qgraph = loader.load(universe, nodes, graphID)
1203 if not isinstance(qgraph, QuantumGraph):
1204 raise TypeError(f"QuantumGraph file contains unexpected object type: {type(qgraph)}")
1205 return qgraph
1207 def iterTaskGraph(self) -> Generator[TaskDef, None, None]:
1208 """Iterate over the `taskGraph` attribute in topological order
1210 Yields
1211 ------
1212 taskDef : `TaskDef`
1213 `TaskDef` objects in topological order
1214 """
1215 yield from nx.topological_sort(self.taskGraph)
1217 def updateRun(self, run: str, *, metadata_key: str | None = None, update_graph_id: bool = False) -> None:
1218 """Change output run and dataset ID for each output dataset.
1220 Parameters
1221 ----------
1222 run : `str`
1223 New output run name.
1224 metadata_key : `str` or `None`
1225 Specifies matadata key corresponding to output run name to update
1226 with new run name. If `None` or if metadata is missing it is not
1227 updated. If metadata is present but key is missing, it will be
1228 added.
1229 update_graph_id : `bool`, optional
1230 If `True` then also update graph ID with a new unique value.
1231 """
1233 def _update_refs_in_place(refs: list[DatasetRef], run: str) -> None:
1234 """Update list of `~lsst.daf.butler.DatasetRef` with new run and
1235 dataset IDs.
1236 """
1237 for ref in refs:
1238 # hack the run to be replaced explicitly
1239 object.__setattr__(ref, "run", run)
1241 # Loop through all outputs and update their datasets.
1242 for node in self._connectedQuanta:
1243 for refs in node.quantum.outputs.values():
1244 _update_refs_in_place(refs, run)
1246 for refs in self._initOutputRefs.values():
1247 _update_refs_in_place(refs, run)
1249 _update_refs_in_place(self._globalInitOutputRefs, run)
1251 # Update all intermediates from their matching outputs.
1252 for node in self._connectedQuanta:
1253 for refs in node.quantum.inputs.values():
1254 _update_refs_in_place(refs, run)
1256 for refs in self._initInputRefs.values():
1257 _update_refs_in_place(refs, run)
1259 if update_graph_id:
1260 self._buildId = BuildId(f"{time.time()}-{os.getpid()}")
1262 # Update metadata if present.
1263 if self._metadata is not None and metadata_key is not None:
1264 metadata = dict(self._metadata)
1265 metadata[metadata_key] = run
1266 self._metadata = metadata
1268 @property
1269 def graphID(self) -> BuildId:
1270 """The ID generated by the graph at construction time (`str`)."""
1271 return self._buildId
1273 @property
1274 def universe(self) -> DimensionUniverse:
1275 """Dimension universe associated with this graph
1276 (`~lsst.daf.butler.DimensionUniverse`).
1277 """
1278 return self._universe
1280 def __iter__(self) -> Generator[QuantumNode, None, None]:
1281 yield from nx.topological_sort(self._connectedQuanta)
1283 def __len__(self) -> int:
1284 return self._count
1286 def __contains__(self, node: QuantumNode) -> bool:
1287 return self._connectedQuanta.has_node(node)
1289 def __getstate__(self) -> dict:
1290 """Store a compact form of the graph as a list of graph nodes, and a
1291 tuple of task labels and task configs. The full graph can be
1292 reconstructed with this information, and it preserves the ordering of
1293 the graph nodes.
1294 """
1295 universe: DimensionUniverse | None = None
1296 for node in self:
1297 dId = node.quantum.dataId
1298 if dId is None:
1299 continue
1300 universe = dId.graph.universe
1301 return {"reduced": self._buildSaveObject(), "graphId": self._buildId, "universe": universe}
1303 def __setstate__(self, state: dict) -> None:
1304 """Reconstructs the state of the graph from the information persisted
1305 in getstate.
1306 """
1307 buffer = io.BytesIO(state["reduced"])
1308 with LoadHelper(buffer, minimumVersion=3) as loader:
1309 qgraph = loader.load(state["universe"], graphID=state["graphId"])
1311 self._metadata = qgraph._metadata
1312 self._buildId = qgraph._buildId
1313 self._datasetDict = qgraph._datasetDict
1314 self._nodeIdMap = qgraph._nodeIdMap
1315 self._count = len(qgraph)
1316 self._taskToQuantumNode = qgraph._taskToQuantumNode
1317 self._taskGraph = qgraph._taskGraph
1318 self._connectedQuanta = qgraph._connectedQuanta
1319 self._initInputRefs = qgraph._initInputRefs
1320 self._initOutputRefs = qgraph._initOutputRefs
1322 def __eq__(self, other: object) -> bool:
1323 if not isinstance(other, QuantumGraph):
1324 return False
1325 if len(self) != len(other):
1326 return False
1327 for node in self:
1328 if node not in other:
1329 return False
1330 if self.determineInputsToQuantumNode(node) != other.determineInputsToQuantumNode(node):
1331 return False
1332 if self.determineOutputsOfQuantumNode(node) != other.determineOutputsOfQuantumNode(node):
1333 return False
1334 if set(self.allDatasetTypes) != set(other.allDatasetTypes):
1335 return False
1336 return set(self.taskGraph) == set(other.taskGraph)