Coverage for python/lsst/pipe/base/pipeline.py: 19%
Shortcuts on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
Shortcuts on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
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"""Module defining Pipeline class and related methods.
24"""
26__all__ = ["Pipeline", "TaskDef", "TaskDatasetTypes", "PipelineDatasetTypes", "LabelSpecifier"]
28import copy
29import logging
30import os
31import re
32import urllib.parse
33import warnings
35# -------------------------------
36# Imports of standard modules --
37# -------------------------------
38from dataclasses import dataclass
39from types import MappingProxyType
40from typing import (
41 TYPE_CHECKING,
42 ClassVar,
43 Dict,
44 Generator,
45 Iterable,
46 Iterator,
47 Mapping,
48 Optional,
49 Set,
50 Tuple,
51 Union,
52)
54# -----------------------------
55# Imports for other modules --
56from lsst.daf.butler import DatasetType, NamedValueSet, Registry, SkyPixDimension
57from lsst.resources import ResourcePath, ResourcePathExpression
58from lsst.utils import doImport
60from . import pipelineIR, pipeTools
61from ._task_metadata import TaskMetadata
62from .configOverrides import ConfigOverrides
63from .connections import iterConnections
64from .pipelineTask import PipelineTask
65from .task import _TASK_METADATA_TYPE
67if TYPE_CHECKING: # Imports needed only for type annotations; may be circular. 67 ↛ 68line 67 didn't jump to line 68, because the condition on line 67 was never true
68 from lsst.obs.base import Instrument
70# ----------------------------------
71# Local non-exported definitions --
72# ----------------------------------
74_LOG = logging.getLogger(__name__)
76# ------------------------
77# Exported definitions --
78# ------------------------
81@dataclass
82class LabelSpecifier:
83 """A structure to specify a subset of labels to load
85 This structure may contain a set of labels to be used in subsetting a
86 pipeline, or a beginning and end point. Beginning or end may be empty,
87 in which case the range will be a half open interval. Unlike python
88 iteration bounds, end bounds are *INCLUDED*. Note that range based
89 selection is not well defined for pipelines that are not linear in nature,
90 and correct behavior is not guaranteed, or may vary from run to run.
91 """
93 labels: Optional[Set[str]] = None
94 begin: Optional[str] = None
95 end: Optional[str] = None
97 def __post_init__(self):
98 if self.labels is not None and (self.begin or self.end):
99 raise ValueError(
100 "This struct can only be initialized with a labels set or a begin (and/or) end specifier"
101 )
104class TaskDef:
105 """TaskDef is a collection of information about task needed by Pipeline.
107 The information includes task name, configuration object and optional
108 task class. This class is just a collection of attributes and it exposes
109 all of them so that attributes could potentially be modified in place
110 (e.g. if configuration needs extra overrides).
112 Attributes
113 ----------
114 taskName : `str`, optional
115 `PipelineTask` class name, currently it is not specified whether this
116 is a fully-qualified name or partial name (e.g. ``module.TaskClass``).
117 Framework should be prepared to handle all cases. If not provided,
118 ``taskClass`` must be, and ``taskClass.__name__`` is used.
119 config : `lsst.pex.config.Config`, optional
120 Instance of the configuration class corresponding to this task class,
121 usually with all overrides applied. This config will be frozen. If
122 not provided, ``taskClass`` must be provided and
123 ``taskClass.ConfigClass()`` will be used.
124 taskClass : `type`, optional
125 `PipelineTask` class object, can be ``None``. If ``None`` then
126 framework will have to locate and load class.
127 label : `str`, optional
128 Task label, usually a short string unique in a pipeline. If not
129 provided, ``taskClass`` must be, and ``taskClass._DefaultName`` will
130 be used.
131 """
133 def __init__(self, taskName=None, config=None, taskClass=None, label=None):
134 if taskName is None:
135 if taskClass is None:
136 raise ValueError("At least one of `taskName` and `taskClass` must be provided.")
137 taskName = taskClass.__name__
138 if config is None:
139 if taskClass is None:
140 raise ValueError("`taskClass` must be provided if `config` is not.")
141 config = taskClass.ConfigClass()
142 if label is None:
143 if taskClass is None:
144 raise ValueError("`taskClass` must be provided if `label` is not.")
145 label = taskClass._DefaultName
146 self.taskName = taskName
147 try:
148 config.validate()
149 except Exception:
150 _LOG.error("Configuration validation failed for task %s (%s)", label, taskName)
151 raise
152 config.freeze()
153 self.config = config
154 self.taskClass = taskClass
155 self.label = label
156 self.connections = config.connections.ConnectionsClass(config=config)
158 @property
159 def configDatasetName(self) -> str:
160 """Name of a dataset type for configuration of this task (`str`)"""
161 return self.label + "_config"
163 @property
164 def metadataDatasetName(self) -> Optional[str]:
165 """Name of a dataset type for metadata of this task, `None` if
166 metadata is not to be saved (`str`)
167 """
168 if self.config.saveMetadata:
169 return self.label + "_metadata"
170 else:
171 return None
173 @property
174 def logOutputDatasetName(self) -> Optional[str]:
175 """Name of a dataset type for log output from this task, `None` if
176 logs are not to be saved (`str`)
177 """
178 if self.config.saveLogOutput:
179 return self.label + "_log"
180 else:
181 return None
183 def __str__(self):
184 rep = "TaskDef(" + self.taskName
185 if self.label:
186 rep += ", label=" + self.label
187 rep += ")"
188 return rep
190 def __eq__(self, other: object) -> bool:
191 if not isinstance(other, TaskDef):
192 return False
193 # This does not consider equality of configs when determining equality
194 # as config equality is a difficult thing to define. Should be updated
195 # after DM-27847
196 return self.taskClass == other.taskClass and self.label == other.label
198 def __hash__(self):
199 return hash((self.taskClass, self.label))
202class Pipeline:
203 """A `Pipeline` is a representation of a series of tasks to run, and the
204 configuration for those tasks.
206 Parameters
207 ----------
208 description : `str`
209 A description of that this pipeline does.
210 """
212 def __init__(self, description: str):
213 pipeline_dict = {"description": description, "tasks": {}}
214 self._pipelineIR = pipelineIR.PipelineIR(pipeline_dict)
216 @classmethod
217 def fromFile(cls, filename: str) -> Pipeline:
218 """Load a pipeline defined in a pipeline yaml file.
220 Parameters
221 ----------
222 filename: `str`
223 A path that points to a pipeline defined in yaml format. This
224 filename may also supply additional labels to be used in
225 subsetting the loaded Pipeline. These labels are separated from
226 the path by a \\#, and may be specified as a comma separated
227 list, or a range denoted as beginning..end. Beginning or end may
228 be empty, in which case the range will be a half open interval.
229 Unlike python iteration bounds, end bounds are *INCLUDED*. Note
230 that range based selection is not well defined for pipelines that
231 are not linear in nature, and correct behavior is not guaranteed,
232 or may vary from run to run.
234 Returns
235 -------
236 pipeline: `Pipeline`
237 The pipeline loaded from specified location with appropriate (if
238 any) subsetting
240 Notes
241 -----
242 This method attempts to prune any contracts that contain labels which
243 are not in the declared subset of labels. This pruning is done using a
244 string based matching due to the nature of contracts and may prune more
245 than it should.
246 """
247 return cls.from_uri(filename)
249 @classmethod
250 def from_uri(cls, uri: ResourcePathExpression) -> Pipeline:
251 """Load a pipeline defined in a pipeline yaml file at a location
252 specified by a URI.
254 Parameters
255 ----------
256 uri: convertible to `ResourcePath`
257 If a string is supplied this should be a URI path that points to a
258 pipeline defined in yaml format, either as a direct path to the
259 yaml file, or as a directory containing a "pipeline.yaml" file (the
260 form used by `write_to_uri` with ``expand=True``). This uri may
261 also supply additional labels to be used in subsetting the loaded
262 Pipeline. These labels are separated from the path by a \\#, and
263 may be specified as a comma separated list, or a range denoted as
264 beginning..end. Beginning or end may be empty, in which case the
265 range will be a half open interval. Unlike python iteration bounds,
266 end bounds are *INCLUDED*. Note that range based selection is not
267 well defined for pipelines that are not linear in nature, and
268 correct behavior is not guaranteed, or may vary from run to run.
269 The same specifiers can be used with a `ResourcePath` object, by
270 being the sole contents in the fragments attribute.
272 Returns
273 -------
274 pipeline: `Pipeline`
275 The pipeline loaded from specified location with appropriate (if
276 any) subsetting
278 Notes
279 -----
280 This method attempts to prune any contracts that contain labels which
281 are not in the declared subset of labels. This pruning is done using a
282 string based matching due to the nature of contracts and may prune more
283 than it should.
284 """
285 # Split up the uri and any labels that were supplied
286 uri, label_specifier = cls._parse_file_specifier(uri)
287 pipeline: Pipeline = cls.fromIR(pipelineIR.PipelineIR.from_uri(uri))
289 # If there are labels supplied, only keep those
290 if label_specifier is not None:
291 pipeline = pipeline.subsetFromLabels(label_specifier)
292 return pipeline
294 def subsetFromLabels(self, labelSpecifier: LabelSpecifier) -> Pipeline:
295 """Subset a pipeline to contain only labels specified in labelSpecifier
297 Parameters
298 ----------
299 labelSpecifier : `labelSpecifier`
300 Object containing labels that describes how to subset a pipeline.
302 Returns
303 -------
304 pipeline : `Pipeline`
305 A new pipeline object that is a subset of the old pipeline
307 Raises
308 ------
309 ValueError
310 Raised if there is an issue with specified labels
312 Notes
313 -----
314 This method attempts to prune any contracts that contain labels which
315 are not in the declared subset of labels. This pruning is done using a
316 string based matching due to the nature of contracts and may prune more
317 than it should.
318 """
319 # Labels supplied as a set
320 if labelSpecifier.labels:
321 labelSet = labelSpecifier.labels
322 # Labels supplied as a range, first create a list of all the labels
323 # in the pipeline sorted according to task dependency. Then only
324 # keep labels that lie between the supplied bounds
325 else:
326 # Create a copy of the pipeline to use when assessing the label
327 # ordering. Use a dict for fast searching while preserving order.
328 # Remove contracts so they do not fail in the expansion step. This
329 # is needed because a user may only configure the tasks they intend
330 # to run, which may cause some contracts to fail if they will later
331 # be dropped
332 pipeline = copy.deepcopy(self)
333 pipeline._pipelineIR.contracts = []
334 labels = {taskdef.label: True for taskdef in pipeline.toExpandedPipeline()}
336 # Verify the bounds are in the labels
337 if labelSpecifier.begin is not None:
338 if labelSpecifier.begin not in labels:
339 raise ValueError(
340 f"Beginning of range subset, {labelSpecifier.begin}, not found in "
341 "pipeline definition"
342 )
343 if labelSpecifier.end is not None:
344 if labelSpecifier.end not in labels:
345 raise ValueError(
346 f"End of range subset, {labelSpecifier.end}, not found in pipeline definition"
347 )
349 labelSet = set()
350 for label in labels:
351 if labelSpecifier.begin is not None:
352 if label != labelSpecifier.begin:
353 continue
354 else:
355 labelSpecifier.begin = None
356 labelSet.add(label)
357 if labelSpecifier.end is not None and label == labelSpecifier.end:
358 break
359 return Pipeline.fromIR(self._pipelineIR.subset_from_labels(labelSet))
361 @staticmethod
362 def _parse_file_specifier(uri: ResourcePathExpression) -> Tuple[ResourcePath, Optional[LabelSpecifier]]:
363 """Split appart a uri and any possible label subsets"""
364 if isinstance(uri, str):
365 # This is to support legacy pipelines during transition
366 uri, num_replace = re.subn("[:](?!\\/\\/)", "#", uri)
367 if num_replace:
368 warnings.warn(
369 f"The pipeline file {uri} seems to use the legacy : to separate "
370 "labels, this is deprecated and will be removed after June 2021, please use "
371 "# instead.",
372 category=FutureWarning,
373 )
374 if uri.count("#") > 1:
375 raise ValueError("Only one set of labels is allowed when specifying a pipeline to load")
376 # Everything else can be converted directly to ResourcePath.
377 uri = ResourcePath(uri)
378 label_subset = uri.fragment or None
380 specifier: Optional[LabelSpecifier]
381 if label_subset is not None:
382 label_subset = urllib.parse.unquote(label_subset)
383 args: Dict[str, Union[Set[str], str, None]]
384 # labels supplied as a list
385 if "," in label_subset:
386 if ".." in label_subset:
387 raise ValueError(
388 "Can only specify a list of labels or a rangewhen loading a Pipline not both"
389 )
390 args = {"labels": set(label_subset.split(","))}
391 # labels supplied as a range
392 elif ".." in label_subset:
393 # Try to de-structure the labelSubset, this will fail if more
394 # than one range is specified
395 begin, end, *rest = label_subset.split("..")
396 if rest:
397 raise ValueError("Only one range can be specified when loading a pipeline")
398 args = {"begin": begin if begin else None, "end": end if end else None}
399 # Assume anything else is a single label
400 else:
401 args = {"labels": {label_subset}}
403 specifier = LabelSpecifier(**args)
404 else:
405 specifier = None
407 return uri, specifier
409 @classmethod
410 def fromString(cls, pipeline_string: str) -> Pipeline:
411 """Create a pipeline from string formatted as a pipeline document.
413 Parameters
414 ----------
415 pipeline_string : `str`
416 A string that is formatted according like a pipeline document
418 Returns
419 -------
420 pipeline: `Pipeline`
421 """
422 pipeline = cls.fromIR(pipelineIR.PipelineIR.from_string(pipeline_string))
423 return pipeline
425 @classmethod
426 def fromIR(cls, deserialized_pipeline: pipelineIR.PipelineIR) -> Pipeline:
427 """Create a pipeline from an already created `PipelineIR` object.
429 Parameters
430 ----------
431 deserialized_pipeline: `PipelineIR`
432 An already created pipeline intermediate representation object
434 Returns
435 -------
436 pipeline: `Pipeline`
437 """
438 pipeline = cls.__new__(cls)
439 pipeline._pipelineIR = deserialized_pipeline
440 return pipeline
442 @classmethod
443 def fromPipeline(cls, pipeline: pipelineIR.PipelineIR) -> Pipeline:
444 """Create a new pipeline by copying an already existing `Pipeline`.
446 Parameters
447 ----------
448 pipeline: `Pipeline`
449 An already created pipeline intermediate representation object
451 Returns
452 -------
453 pipeline: `Pipeline`
454 """
455 return cls.fromIR(copy.deepcopy(pipeline._pipelineIR))
457 def __str__(self) -> str:
458 # tasks need sorted each call because someone might have added or
459 # removed task, and caching changes does not seem worth the small
460 # overhead
461 labels = [td.label for td in self._toExpandedPipelineImpl(checkContracts=False)]
462 self._pipelineIR.reorder_tasks(labels)
463 return str(self._pipelineIR)
465 def addInstrument(self, instrument: Union[Instrument, str]) -> None:
466 """Add an instrument to the pipeline, or replace an instrument that is
467 already defined.
469 Parameters
470 ----------
471 instrument : `~lsst.daf.butler.instrument.Instrument` or `str`
472 Either a derived class object of a `lsst.daf.butler.instrument` or
473 a string corresponding to a fully qualified
474 `lsst.daf.butler.instrument` name.
475 """
476 if isinstance(instrument, str):
477 pass
478 else:
479 # TODO: assume that this is a subclass of Instrument, no type
480 # checking
481 instrument = f"{instrument.__module__}.{instrument.__qualname__}"
482 self._pipelineIR.instrument = instrument
484 def getInstrument(self) -> Instrument:
485 """Get the instrument from the pipeline.
487 Returns
488 -------
489 instrument : `~lsst.daf.butler.instrument.Instrument`, `str`, or None
490 A derived class object of a `lsst.daf.butler.instrument`, a string
491 corresponding to a fully qualified `lsst.daf.butler.instrument`
492 name, or None if the pipeline does not have an instrument.
493 """
494 return self._pipelineIR.instrument
496 def addTask(self, task: Union[PipelineTask, str], label: str) -> None:
497 """Add a new task to the pipeline, or replace a task that is already
498 associated with the supplied label.
500 Parameters
501 ----------
502 task: `PipelineTask` or `str`
503 Either a derived class object of a `PipelineTask` or a string
504 corresponding to a fully qualified `PipelineTask` name.
505 label: `str`
506 A label that is used to identify the `PipelineTask` being added
507 """
508 if isinstance(task, str):
509 taskName = task
510 elif issubclass(task, PipelineTask):
511 taskName = f"{task.__module__}.{task.__qualname__}"
512 else:
513 raise ValueError(
514 "task must be either a child class of PipelineTask or a string containing"
515 " a fully qualified name to one"
516 )
517 if not label:
518 # in some cases (with command line-generated pipeline) tasks can
519 # be defined without label which is not acceptable, use task
520 # _DefaultName in that case
521 if isinstance(task, str):
522 task = doImport(task)
523 label = task._DefaultName
524 self._pipelineIR.tasks[label] = pipelineIR.TaskIR(label, taskName)
526 def removeTask(self, label: str) -> None:
527 """Remove a task from the pipeline.
529 Parameters
530 ----------
531 label : `str`
532 The label used to identify the task that is to be removed
534 Raises
535 ------
536 KeyError
537 If no task with that label exists in the pipeline
539 """
540 self._pipelineIR.tasks.pop(label)
542 def addConfigOverride(self, label: str, key: str, value: object) -> None:
543 """Apply single config override.
545 Parameters
546 ----------
547 label : `str`
548 Label of the task.
549 key: `str`
550 Fully-qualified field name.
551 value : object
552 Value to be given to a field.
553 """
554 self._addConfigImpl(label, pipelineIR.ConfigIR(rest={key: value}))
556 def addConfigFile(self, label: str, filename: str) -> None:
557 """Add overrides from a specified file.
559 Parameters
560 ----------
561 label : `str`
562 The label used to identify the task associated with config to
563 modify
564 filename : `str`
565 Path to the override file.
566 """
567 self._addConfigImpl(label, pipelineIR.ConfigIR(file=[filename]))
569 def addConfigPython(self, label: str, pythonString: str) -> None:
570 """Add Overrides by running a snippet of python code against a config.
572 Parameters
573 ----------
574 label : `str`
575 The label used to identity the task associated with config to
576 modify.
577 pythonString: `str`
578 A string which is valid python code to be executed. This is done
579 with config as the only local accessible value.
580 """
581 self._addConfigImpl(label, pipelineIR.ConfigIR(python=pythonString))
583 def _addConfigImpl(self, label: str, newConfig: pipelineIR.ConfigIR) -> None:
584 if label == "parameters":
585 if newConfig.rest.keys() - self._pipelineIR.parameters.mapping.keys():
586 raise ValueError("Cannot override parameters that are not defined in pipeline")
587 self._pipelineIR.parameters.mapping.update(newConfig.rest)
588 if newConfig.file:
589 raise ValueError("Setting parameters section with config file is not supported")
590 if newConfig.python:
591 raise ValueError("Setting parameters section using python block in unsupported")
592 return
593 if label not in self._pipelineIR.tasks:
594 raise LookupError(f"There are no tasks labeled '{label}' in the pipeline")
595 self._pipelineIR.tasks[label].add_or_update_config(newConfig)
597 def toFile(self, filename: str) -> None:
598 self._pipelineIR.to_file(filename)
600 def write_to_uri(self, uri: ResourcePathExpression) -> None:
601 """Write the pipeline to a file or directory.
603 Parameters
604 ----------
605 uri : convertible to `ResourcePath`
606 URI to write to; may have any scheme with `ResourcePath` write
607 support or no scheme for a local file/directory. Should have a
608 ``.yaml``.
609 """
610 labels = [td.label for td in self._toExpandedPipelineImpl(checkContracts=False)]
611 self._pipelineIR.reorder_tasks(labels)
612 self._pipelineIR.write_to_uri(uri)
614 def toExpandedPipeline(self) -> Generator[TaskDef, None, None]:
615 """Returns a generator of TaskDefs which can be used to create quantum
616 graphs.
618 Returns
619 -------
620 generator : generator of `TaskDef`
621 The generator returned will be the sorted iterator of tasks which
622 are to be used in constructing a quantum graph.
624 Raises
625 ------
626 NotImplementedError
627 If a dataId is supplied in a config block. This is in place for
628 future use
629 """
630 yield from self._toExpandedPipelineImpl()
632 def _toExpandedPipelineImpl(self, checkContracts=True) -> Iterable[TaskDef]:
633 taskDefs = []
634 for label in self._pipelineIR.tasks:
635 taskDefs.append(self._buildTaskDef(label))
637 # lets evaluate the contracts
638 if self._pipelineIR.contracts is not None:
639 label_to_config = {x.label: x.config for x in taskDefs}
640 for contract in self._pipelineIR.contracts:
641 # execute this in its own line so it can raise a good error
642 # message if there was problems with the eval
643 success = eval(contract.contract, None, label_to_config)
644 if not success:
645 extra_info = f": {contract.msg}" if contract.msg is not None else ""
646 raise pipelineIR.ContractError(
647 f"Contract(s) '{contract.contract}' were not satisfied{extra_info}"
648 )
650 taskDefs = sorted(taskDefs, key=lambda x: x.label)
651 yield from pipeTools.orderPipeline(taskDefs)
653 def _buildTaskDef(self, label: str) -> TaskDef:
654 if (taskIR := self._pipelineIR.tasks.get(label)) is None:
655 raise NameError(f"Label {label} does not appear in this pipeline")
656 taskClass = doImport(taskIR.klass)
657 taskName = taskClass.__qualname__
658 config = taskClass.ConfigClass()
659 overrides = ConfigOverrides()
660 if self._pipelineIR.instrument is not None:
661 overrides.addInstrumentOverride(self._pipelineIR.instrument, taskClass._DefaultName)
662 if taskIR.config is not None:
663 for configIR in (configIr.formatted(self._pipelineIR.parameters) for configIr in taskIR.config):
664 if configIR.dataId is not None:
665 raise NotImplementedError(
666 "Specializing a config on a partial data id is not yet "
667 "supported in Pipeline definition"
668 )
669 # only apply override if it applies to everything
670 if configIR.dataId is None:
671 if configIR.file:
672 for configFile in configIR.file:
673 overrides.addFileOverride(os.path.expandvars(configFile))
674 if configIR.python is not None:
675 overrides.addPythonOverride(configIR.python)
676 for key, value in configIR.rest.items():
677 overrides.addValueOverride(key, value)
678 overrides.applyTo(config)
679 return TaskDef(taskName=taskName, config=config, taskClass=taskClass, label=label)
681 def __iter__(self) -> Generator[TaskDef, None, None]:
682 return self.toExpandedPipeline()
684 def __getitem__(self, item: str) -> TaskDef:
685 return self._buildTaskDef(item)
687 def __len__(self):
688 return len(self._pipelineIR.tasks)
690 def __eq__(self, other: object):
691 if not isinstance(other, Pipeline):
692 return False
693 return self._pipelineIR == other._pipelineIR
696@dataclass(frozen=True)
697class TaskDatasetTypes:
698 """An immutable struct that extracts and classifies the dataset types used
699 by a `PipelineTask`
700 """
702 initInputs: NamedValueSet[DatasetType]
703 """Dataset types that are needed as inputs in order to construct this Task.
705 Task-level `initInputs` may be classified as either
706 `~PipelineDatasetTypes.initInputs` or
707 `~PipelineDatasetTypes.initIntermediates` at the Pipeline level.
708 """
710 initOutputs: NamedValueSet[DatasetType]
711 """Dataset types that may be written after constructing this Task.
713 Task-level `initOutputs` may be classified as either
714 `~PipelineDatasetTypes.initOutputs` or
715 `~PipelineDatasetTypes.initIntermediates` at the Pipeline level.
716 """
718 inputs: NamedValueSet[DatasetType]
719 """Dataset types that are regular inputs to this Task.
721 If an input dataset needed for a Quantum cannot be found in the input
722 collection(s) or produced by another Task in the Pipeline, that Quantum
723 (and all dependent Quanta) will not be produced.
725 Task-level `inputs` may be classified as either
726 `~PipelineDatasetTypes.inputs` or `~PipelineDatasetTypes.intermediates`
727 at the Pipeline level.
728 """
730 prerequisites: NamedValueSet[DatasetType]
731 """Dataset types that are prerequisite inputs to this Task.
733 Prerequisite inputs must exist in the input collection(s) before the
734 pipeline is run, but do not constrain the graph - if a prerequisite is
735 missing for a Quantum, `PrerequisiteMissingError` is raised.
737 Prerequisite inputs are not resolved until the second stage of
738 QuantumGraph generation.
739 """
741 outputs: NamedValueSet[DatasetType]
742 """Dataset types that are produced by this Task.
744 Task-level `outputs` may be classified as either
745 `~PipelineDatasetTypes.outputs` or `~PipelineDatasetTypes.intermediates`
746 at the Pipeline level.
747 """
749 @classmethod
750 def fromTaskDef(
751 cls,
752 taskDef: TaskDef,
753 *,
754 registry: Registry,
755 include_configs: bool = True,
756 storage_class_mapping: Optional[Mapping[str, str]] = None,
757 ) -> TaskDatasetTypes:
758 """Extract and classify the dataset types from a single `PipelineTask`.
760 Parameters
761 ----------
762 taskDef: `TaskDef`
763 An instance of a `TaskDef` class for a particular `PipelineTask`.
764 registry: `Registry`
765 Registry used to construct normalized `DatasetType` objects and
766 retrieve those that are incomplete.
767 include_configs : `bool`, optional
768 If `True` (default) include config dataset types as
769 ``initOutputs``.
770 storage_class_mapping : `Mapping` of `str` to `StorageClass`, optional
771 If a taskdef contains a component dataset type that is unknown
772 to the registry, its parent StorageClass will be looked up in this
773 mapping if it is supplied. If the mapping does not contain the
774 composite dataset type, or the mapping is not supplied an exception
775 will be raised.
777 Returns
778 -------
779 types: `TaskDatasetTypes`
780 The dataset types used by this task.
782 Raises
783 ------
784 ValueError
785 Raised if dataset type connection definition differs from
786 registry definition.
787 LookupError
788 Raised if component parent StorageClass could not be determined
789 and storage_class_mapping does not contain the composite type, or
790 is set to None.
791 """
793 def makeDatasetTypesSet(connectionType: str, freeze: bool = True) -> NamedValueSet[DatasetType]:
794 """Constructs a set of true `DatasetType` objects
796 Parameters
797 ----------
798 connectionType : `str`
799 Name of the connection type to produce a set for, corresponds
800 to an attribute of type `list` on the connection class instance
801 freeze : `bool`, optional
802 If `True`, call `NamedValueSet.freeze` on the object returned.
804 Returns
805 -------
806 datasetTypes : `NamedValueSet`
807 A set of all datasetTypes which correspond to the input
808 connection type specified in the connection class of this
809 `PipelineTask`
811 Raises
812 ------
813 ValueError
814 Raised if dataset type connection definition differs from
815 registry definition.
816 LookupError
817 Raised if component parent StorageClass could not be determined
818 and storage_class_mapping does not contain the composite type,
819 or is set to None.
821 Notes
822 -----
823 This function is a closure over the variables ``registry`` and
824 ``taskDef``, and ``storage_class_mapping``.
825 """
826 datasetTypes = NamedValueSet()
827 for c in iterConnections(taskDef.connections, connectionType):
828 dimensions = set(getattr(c, "dimensions", set()))
829 if "skypix" in dimensions:
830 try:
831 datasetType = registry.getDatasetType(c.name)
832 except LookupError as err:
833 raise LookupError(
834 f"DatasetType '{c.name}' referenced by "
835 f"{type(taskDef.connections).__name__} uses 'skypix' as a dimension "
836 f"placeholder, but does not already exist in the registry. "
837 f"Note that reference catalog names are now used as the dataset "
838 f"type name instead of 'ref_cat'."
839 ) from err
840 rest1 = set(registry.dimensions.extract(dimensions - set(["skypix"])).names)
841 rest2 = set(
842 dim.name for dim in datasetType.dimensions if not isinstance(dim, SkyPixDimension)
843 )
844 if rest1 != rest2:
845 raise ValueError(
846 f"Non-skypix dimensions for dataset type {c.name} declared in "
847 f"connections ({rest1}) are inconsistent with those in "
848 f"registry's version of this dataset ({rest2})."
849 )
850 else:
851 # Component dataset types are not explicitly in the
852 # registry. This complicates consistency checks with
853 # registry and requires we work out the composite storage
854 # class.
855 registryDatasetType = None
856 try:
857 registryDatasetType = registry.getDatasetType(c.name)
858 except KeyError:
859 compositeName, componentName = DatasetType.splitDatasetTypeName(c.name)
860 if componentName:
861 if storage_class_mapping is None or compositeName not in storage_class_mapping:
862 raise LookupError(
863 "Component parent class cannot be determined, and "
864 "composite name was not in storage class mapping, or no "
865 "storage_class_mapping was supplied"
866 )
867 else:
868 parentStorageClass = storage_class_mapping[compositeName]
869 else:
870 parentStorageClass = None
871 datasetType = c.makeDatasetType(
872 registry.dimensions, parentStorageClass=parentStorageClass
873 )
874 registryDatasetType = datasetType
875 else:
876 datasetType = c.makeDatasetType(
877 registry.dimensions, parentStorageClass=registryDatasetType.parentStorageClass
878 )
880 if registryDatasetType and datasetType != registryDatasetType:
881 try:
882 # Explicitly check for storage class just to make
883 # more specific message.
884 _ = datasetType.storageClass
885 except KeyError:
886 raise ValueError(
887 "Storage class does not exist for supplied dataset type "
888 f"{datasetType} for {taskDef.label}."
889 ) from None
890 raise ValueError(
891 f"Supplied dataset type ({datasetType}) inconsistent with "
892 f"registry definition ({registryDatasetType}) "
893 f"for {taskDef.label}."
894 )
895 datasetTypes.add(datasetType)
896 if freeze:
897 datasetTypes.freeze()
898 return datasetTypes
900 # optionally add initOutput dataset for config
901 initOutputs = makeDatasetTypesSet("initOutputs", freeze=False)
902 if include_configs:
903 initOutputs.add(
904 DatasetType(
905 taskDef.configDatasetName,
906 registry.dimensions.empty,
907 storageClass="Config",
908 )
909 )
910 initOutputs.freeze()
912 # optionally add output dataset for metadata
913 outputs = makeDatasetTypesSet("outputs", freeze=False)
914 if taskDef.metadataDatasetName is not None:
915 # Metadata is supposed to be of the TaskMetadata type, its
916 # dimensions correspond to a task quantum.
917 dimensions = registry.dimensions.extract(taskDef.connections.dimensions)
919 # Allow the storage class definition to be read from the existing
920 # dataset type definition if present.
921 try:
922 current = registry.getDatasetType(taskDef.metadataDatasetName)
923 except KeyError:
924 # No previous definition so use the default.
925 storageClass = "TaskMetadata" if _TASK_METADATA_TYPE is TaskMetadata else "PropertySet"
926 else:
927 storageClass = current.storageClass.name
929 outputs |= {DatasetType(taskDef.metadataDatasetName, dimensions, storageClass)}
930 if taskDef.logOutputDatasetName is not None:
931 # Log output dimensions correspond to a task quantum.
932 dimensions = registry.dimensions.extract(taskDef.connections.dimensions)
933 outputs |= {DatasetType(taskDef.logOutputDatasetName, dimensions, "ButlerLogRecords")}
935 outputs.freeze()
937 return cls(
938 initInputs=makeDatasetTypesSet("initInputs"),
939 initOutputs=initOutputs,
940 inputs=makeDatasetTypesSet("inputs"),
941 prerequisites=makeDatasetTypesSet("prerequisiteInputs"),
942 outputs=outputs,
943 )
946@dataclass(frozen=True)
947class PipelineDatasetTypes:
948 """An immutable struct that classifies the dataset types used in a
949 `Pipeline`.
950 """
952 packagesDatasetName: ClassVar[str] = "packages"
953 """Name of a dataset type used to save package versions.
954 """
956 initInputs: NamedValueSet[DatasetType]
957 """Dataset types that are needed as inputs in order to construct the Tasks
958 in this Pipeline.
960 This does not include dataset types that are produced when constructing
961 other Tasks in the Pipeline (these are classified as `initIntermediates`).
962 """
964 initOutputs: NamedValueSet[DatasetType]
965 """Dataset types that may be written after constructing the Tasks in this
966 Pipeline.
968 This does not include dataset types that are also used as inputs when
969 constructing other Tasks in the Pipeline (these are classified as
970 `initIntermediates`).
971 """
973 initIntermediates: NamedValueSet[DatasetType]
974 """Dataset types that are both used when constructing one or more Tasks
975 in the Pipeline and produced as a side-effect of constructing another
976 Task in the Pipeline.
977 """
979 inputs: NamedValueSet[DatasetType]
980 """Dataset types that are regular inputs for the full pipeline.
982 If an input dataset needed for a Quantum cannot be found in the input
983 collection(s), that Quantum (and all dependent Quanta) will not be
984 produced.
985 """
987 prerequisites: NamedValueSet[DatasetType]
988 """Dataset types that are prerequisite inputs for the full Pipeline.
990 Prerequisite inputs must exist in the input collection(s) before the
991 pipeline is run, but do not constrain the graph - if a prerequisite is
992 missing for a Quantum, `PrerequisiteMissingError` is raised.
994 Prerequisite inputs are not resolved until the second stage of
995 QuantumGraph generation.
996 """
998 intermediates: NamedValueSet[DatasetType]
999 """Dataset types that are output by one Task in the Pipeline and consumed
1000 as inputs by one or more other Tasks in the Pipeline.
1001 """
1003 outputs: NamedValueSet[DatasetType]
1004 """Dataset types that are output by a Task in the Pipeline and not consumed
1005 by any other Task in the Pipeline.
1006 """
1008 byTask: Mapping[str, TaskDatasetTypes]
1009 """Per-Task dataset types, keyed by label in the `Pipeline`.
1011 This is guaranteed to be zip-iterable with the `Pipeline` itself (assuming
1012 neither has been modified since the dataset types were extracted, of
1013 course).
1014 """
1016 @classmethod
1017 def fromPipeline(
1018 cls,
1019 pipeline: Union[Pipeline, Iterable[TaskDef]],
1020 *,
1021 registry: Registry,
1022 include_configs: bool = True,
1023 include_packages: bool = True,
1024 ) -> PipelineDatasetTypes:
1025 """Extract and classify the dataset types from all tasks in a
1026 `Pipeline`.
1028 Parameters
1029 ----------
1030 pipeline: `Pipeline` or `Iterable` [ `TaskDef` ]
1031 A collection of tasks that can be run together.
1032 registry: `Registry`
1033 Registry used to construct normalized `DatasetType` objects and
1034 retrieve those that are incomplete.
1035 include_configs : `bool`, optional
1036 If `True` (default) include config dataset types as
1037 ``initOutputs``.
1038 include_packages : `bool`, optional
1039 If `True` (default) include the dataset type for software package
1040 versions in ``initOutputs``.
1042 Returns
1043 -------
1044 types: `PipelineDatasetTypes`
1045 The dataset types used by this `Pipeline`.
1047 Raises
1048 ------
1049 ValueError
1050 Raised if Tasks are inconsistent about which datasets are marked
1051 prerequisite. This indicates that the Tasks cannot be run as part
1052 of the same `Pipeline`.
1053 """
1054 allInputs = NamedValueSet()
1055 allOutputs = NamedValueSet()
1056 allInitInputs = NamedValueSet()
1057 allInitOutputs = NamedValueSet()
1058 prerequisites = NamedValueSet()
1059 byTask = dict()
1060 if include_packages:
1061 allInitOutputs.add(
1062 DatasetType(
1063 cls.packagesDatasetName,
1064 registry.dimensions.empty,
1065 storageClass="Packages",
1066 )
1067 )
1068 # create a list of TaskDefs in case the input is a generator
1069 pipeline = list(pipeline)
1071 # collect all the output dataset types
1072 typeStorageclassMap: Dict[str, str] = {}
1073 for taskDef in pipeline:
1074 for outConnection in iterConnections(taskDef.connections, "outputs"):
1075 typeStorageclassMap[outConnection.name] = outConnection.storageClass
1077 for taskDef in pipeline:
1078 thisTask = TaskDatasetTypes.fromTaskDef(
1079 taskDef,
1080 registry=registry,
1081 include_configs=include_configs,
1082 storage_class_mapping=typeStorageclassMap,
1083 )
1084 allInitInputs |= thisTask.initInputs
1085 allInitOutputs |= thisTask.initOutputs
1086 allInputs |= thisTask.inputs
1087 prerequisites |= thisTask.prerequisites
1088 allOutputs |= thisTask.outputs
1089 byTask[taskDef.label] = thisTask
1090 if not prerequisites.isdisjoint(allInputs):
1091 raise ValueError(
1092 "{} marked as both prerequisites and regular inputs".format(
1093 {dt.name for dt in allInputs & prerequisites}
1094 )
1095 )
1096 if not prerequisites.isdisjoint(allOutputs):
1097 raise ValueError(
1098 "{} marked as both prerequisites and outputs".format(
1099 {dt.name for dt in allOutputs & prerequisites}
1100 )
1101 )
1102 # Make sure that components which are marked as inputs get treated as
1103 # intermediates if there is an output which produces the composite
1104 # containing the component
1105 intermediateComponents = NamedValueSet()
1106 intermediateComposites = NamedValueSet()
1107 outputNameMapping = {dsType.name: dsType for dsType in allOutputs}
1108 for dsType in allInputs:
1109 # get the name of a possible component
1110 name, component = dsType.nameAndComponent()
1111 # if there is a component name, that means this is a component
1112 # DatasetType, if there is an output which produces the parent of
1113 # this component, treat this input as an intermediate
1114 if component is not None:
1115 # This needs to be in this if block, because someone might have
1116 # a composite that is a pure input from existing data
1117 if name in outputNameMapping:
1118 intermediateComponents.add(dsType)
1119 intermediateComposites.add(outputNameMapping[name])
1121 def checkConsistency(a: NamedValueSet, b: NamedValueSet):
1122 common = a.names & b.names
1123 for name in common:
1124 if a[name] != b[name]:
1125 raise ValueError(f"Conflicting definitions for dataset type: {a[name]} != {b[name]}.")
1127 checkConsistency(allInitInputs, allInitOutputs)
1128 checkConsistency(allInputs, allOutputs)
1129 checkConsistency(allInputs, intermediateComposites)
1130 checkConsistency(allOutputs, intermediateComposites)
1132 def frozen(s: NamedValueSet) -> NamedValueSet:
1133 s.freeze()
1134 return s
1136 return cls(
1137 initInputs=frozen(allInitInputs - allInitOutputs),
1138 initIntermediates=frozen(allInitInputs & allInitOutputs),
1139 initOutputs=frozen(allInitOutputs - allInitInputs),
1140 inputs=frozen(allInputs - allOutputs - intermediateComponents),
1141 intermediates=frozen(allInputs & allOutputs | intermediateComponents),
1142 outputs=frozen(allOutputs - allInputs - intermediateComposites),
1143 prerequisites=frozen(prerequisites),
1144 byTask=MappingProxyType(byTask), # MappingProxyType -> frozen view of dict for immutability
1145 )
1147 @classmethod
1148 def initOutputNames(
1149 cls,
1150 pipeline: Union[Pipeline, Iterable[TaskDef]],
1151 *,
1152 include_configs: bool = True,
1153 include_packages: bool = True,
1154 ) -> Iterator[str]:
1155 """Return the names of dataset types ot task initOutputs, Configs,
1156 and package versions for a pipeline.
1158 Parameters
1159 ----------
1160 pipeline: `Pipeline` or `Iterable` [ `TaskDef` ]
1161 A `Pipeline` instance or collection of `TaskDef` instances.
1162 include_configs : `bool`, optional
1163 If `True` (default) include config dataset types.
1164 include_packages : `bool`, optional
1165 If `True` (default) include the dataset type for package versions.
1167 Yields
1168 ------
1169 datasetTypeName : `str`
1170 Name of the dataset type.
1171 """
1172 if include_packages:
1173 # Package versions dataset type
1174 yield cls.packagesDatasetName
1176 if isinstance(pipeline, Pipeline):
1177 pipeline = pipeline.toExpandedPipeline()
1179 for taskDef in pipeline:
1181 # all task InitOutputs
1182 for name in taskDef.connections.initOutputs:
1183 attribute = getattr(taskDef.connections, name)
1184 yield attribute.name
1186 # config dataset name
1187 if include_configs:
1188 yield taskDef.configDatasetName