Coverage for python/lsst/daf/butler/_butler.py: 8%
718 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-07-12 10:56 -0700
« prev ^ index » next coverage.py v7.2.7, created at 2023-07-12 10:56 -0700
1# This file is part of daf_butler.
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/>.
22"""Butler top level classes.
23"""
24from __future__ import annotations
26__all__ = (
27 "Butler",
28 "ButlerValidationError",
29)
31import collections.abc
32import contextlib
33import io
34import logging
35import numbers
36import os
37import warnings
38from collections import Counter, defaultdict
39from collections.abc import Iterable, Iterator, MutableMapping, Sequence
40from typing import TYPE_CHECKING, Any, ClassVar, TextIO
42from deprecated.sphinx import deprecated
43from lsst.resources import ResourcePath, ResourcePathExpression
44from lsst.utils import doImportType
45from lsst.utils.introspection import get_class_of
46from lsst.utils.logging import VERBOSE, getLogger
47from sqlalchemy.exc import IntegrityError
49from ._butlerConfig import ButlerConfig
50from ._butlerRepoIndex import ButlerRepoIndex
51from ._dataset_existence import DatasetExistence
52from ._deferredDatasetHandle import DeferredDatasetHandle
53from ._limited_butler import LimitedButler
54from .core import (
55 Config,
56 ConfigSubset,
57 DataCoordinate,
58 DataId,
59 DataIdValue,
60 DatasetIdGenEnum,
61 DatasetRef,
62 DatasetRefURIs,
63 DatasetType,
64 Datastore,
65 Dimension,
66 DimensionConfig,
67 DimensionElement,
68 DimensionRecord,
69 DimensionUniverse,
70 FileDataset,
71 Progress,
72 StorageClass,
73 StorageClassFactory,
74 Timespan,
75 ValidationError,
76)
77from .core.repoRelocation import BUTLER_ROOT_TAG
78from .core.utils import transactional
79from .registry import (
80 CollectionType,
81 ConflictingDefinitionError,
82 DataIdError,
83 MissingDatasetTypeError,
84 NoDefaultCollectionError,
85 Registry,
86 RegistryConfig,
87 RegistryDefaults,
88)
89from .transfers import RepoExportContext
91if TYPE_CHECKING:
92 from lsst.resources import ResourceHandleProtocol
94 from .transfers import RepoImportBackend
96log = getLogger(__name__)
99class ButlerValidationError(ValidationError):
100 """There is a problem with the Butler configuration."""
102 pass
105class Butler(LimitedButler):
106 """Main entry point for the data access system.
108 Parameters
109 ----------
110 config : `ButlerConfig`, `Config` or `str`, optional.
111 Configuration. Anything acceptable to the
112 `ButlerConfig` constructor. If a directory path
113 is given the configuration will be read from a ``butler.yaml`` file in
114 that location. If `None` is given default values will be used.
115 butler : `Butler`, optional.
116 If provided, construct a new Butler that uses the same registry and
117 datastore as the given one, but with the given collection and run.
118 Incompatible with the ``config``, ``searchPaths``, and ``writeable``
119 arguments.
120 collections : `str` or `~collections.abc.Iterable` [ `str` ], optional
121 An expression specifying the collections to be searched (in order) when
122 reading datasets.
123 This may be a `str` collection name or an iterable thereof.
124 See :ref:`daf_butler_collection_expressions` for more information.
125 These collections are not registered automatically and must be
126 manually registered before they are used by any method, but they may be
127 manually registered after the `Butler` is initialized.
128 run : `str`, optional
129 Name of the `~CollectionType.RUN` collection new datasets should be
130 inserted into. If ``collections`` is `None` and ``run`` is not `None`,
131 ``collections`` will be set to ``[run]``. If not `None`, this
132 collection will automatically be registered. If this is not set (and
133 ``writeable`` is not set either), a read-only butler will be created.
134 searchPaths : `list` of `str`, optional
135 Directory paths to search when calculating the full Butler
136 configuration. Not used if the supplied config is already a
137 `ButlerConfig`.
138 writeable : `bool`, optional
139 Explicitly sets whether the butler supports write operations. If not
140 provided, a read-write butler is created if any of ``run``, ``tags``,
141 or ``chains`` is non-empty.
142 inferDefaults : `bool`, optional
143 If `True` (default) infer default data ID values from the values
144 present in the datasets in ``collections``: if all collections have the
145 same value (or no value) for a governor dimension, that value will be
146 the default for that dimension. Nonexistent collections are ignored.
147 If a default value is provided explicitly for a governor dimension via
148 ``**kwargs``, no default will be inferred for that dimension.
149 **kwargs : `str`
150 Default data ID key-value pairs. These may only identify "governor"
151 dimensions like ``instrument`` and ``skymap``.
153 Examples
154 --------
155 While there are many ways to control exactly how a `Butler` interacts with
156 the collections in its `Registry`, the most common cases are still simple.
158 For a read-only `Butler` that searches one collection, do::
160 butler = Butler("/path/to/repo", collections=["u/alice/DM-50000"])
162 For a read-write `Butler` that writes to and reads from a
163 `~CollectionType.RUN` collection::
165 butler = Butler("/path/to/repo", run="u/alice/DM-50000/a")
167 The `Butler` passed to a ``PipelineTask`` is often much more complex,
168 because we want to write to one `~CollectionType.RUN` collection but read
169 from several others (as well)::
171 butler = Butler("/path/to/repo", run="u/alice/DM-50000/a",
172 collections=["u/alice/DM-50000/a",
173 "u/bob/DM-49998",
174 "HSC/defaults"])
176 This butler will `put` new datasets to the run ``u/alice/DM-50000/a``.
177 Datasets will be read first from that run (since it appears first in the
178 chain), and then from ``u/bob/DM-49998`` and finally ``HSC/defaults``.
180 Finally, one can always create a `Butler` with no collections::
182 butler = Butler("/path/to/repo", writeable=True)
184 This can be extremely useful when you just want to use ``butler.registry``,
185 e.g. for inserting dimension data or managing collections, or when the
186 collections you want to use with the butler are not consistent.
187 Passing ``writeable`` explicitly here is only necessary if you want to be
188 able to make changes to the repo - usually the value for ``writeable`` can
189 be guessed from the collection arguments provided, but it defaults to
190 `False` when there are not collection arguments.
191 """
193 def __init__(
194 self,
195 config: Config | ResourcePathExpression | None = None,
196 *,
197 butler: Butler | None = None,
198 collections: Any = None,
199 run: str | None = None,
200 searchPaths: Sequence[ResourcePathExpression] | None = None,
201 writeable: bool | None = None,
202 inferDefaults: bool = True,
203 **kwargs: str,
204 ):
205 defaults = RegistryDefaults(collections=collections, run=run, infer=inferDefaults, **kwargs)
206 # Load registry, datastore, etc. from config or existing butler.
207 if butler is not None:
208 if config is not None or searchPaths is not None or writeable is not None:
209 raise TypeError(
210 "Cannot pass 'config', 'searchPaths', or 'writeable' arguments with 'butler' argument."
211 )
212 self.registry = butler.registry.copy(defaults)
213 self._datastore = butler._datastore
214 self.storageClasses = butler.storageClasses
215 self._config: ButlerConfig = butler._config
216 else:
217 self._config = ButlerConfig(config, searchPaths=searchPaths)
218 try:
219 if "root" in self._config:
220 butlerRoot = self._config["root"]
221 else:
222 butlerRoot = self._config.configDir
223 if writeable is None:
224 writeable = run is not None
225 self.registry = Registry.fromConfig(
226 self._config, butlerRoot=butlerRoot, writeable=writeable, defaults=defaults
227 )
228 self._datastore = Datastore.fromConfig(
229 self._config, self.registry.getDatastoreBridgeManager(), butlerRoot=butlerRoot
230 )
231 self.storageClasses = StorageClassFactory()
232 self.storageClasses.addFromConfig(self._config)
233 except Exception:
234 # Failures here usually mean that configuration is incomplete,
235 # just issue an error message which includes config file URI.
236 log.error(f"Failed to instantiate Butler from config {self._config.configFile}.")
237 raise
239 # For execution butler the datastore needs a special
240 # dependency-inversion trick. This is not used by regular butler,
241 # but we do not have a way to distinguish regular butler from execution
242 # butler.
243 self._datastore.set_retrieve_dataset_type_method(self._retrieve_dataset_type)
245 if "run" in self._config or "collection" in self._config:
246 raise ValueError("Passing a run or collection via configuration is no longer supported.")
248 GENERATION: ClassVar[int] = 3
249 """This is a Generation 3 Butler.
251 This attribute may be removed in the future, once the Generation 2 Butler
252 interface has been fully retired; it should only be used in transitional
253 code.
254 """
256 def _retrieve_dataset_type(self, name: str) -> DatasetType | None:
257 """Return DatasetType defined in registry given dataset type name."""
258 try:
259 return self.registry.getDatasetType(name)
260 except MissingDatasetTypeError:
261 return None
263 @classmethod
264 def get_repo_uri(cls, label: str, return_label: bool = False) -> ResourcePath:
265 """Look up the label in a butler repository index.
267 Parameters
268 ----------
269 label : `str`
270 Label of the Butler repository to look up.
271 return_label : `bool`, optional
272 If ``label`` cannot be found in the repository index (either
273 because index is not defined or ``label`` is not in the index) and
274 ``return_label`` is `True` then return ``ResourcePath(label)``.
275 If ``return_label`` is `False` (default) then an exception will be
276 raised instead.
278 Returns
279 -------
280 uri : `lsst.resources.ResourcePath`
281 URI to the Butler repository associated with the given label or
282 default value if it is provided.
284 Raises
285 ------
286 KeyError
287 Raised if the label is not found in the index, or if an index
288 is not defined, and ``return_label`` is `False`.
290 Notes
291 -----
292 See `~lsst.daf.butler.ButlerRepoIndex` for details on how the
293 information is discovered.
294 """
295 return ButlerRepoIndex.get_repo_uri(label, return_label)
297 @classmethod
298 def get_known_repos(cls) -> set[str]:
299 """Retrieve the list of known repository labels.
301 Returns
302 -------
303 repos : `set` of `str`
304 All the known labels. Can be empty if no index can be found.
306 Notes
307 -----
308 See `~lsst.daf.butler.ButlerRepoIndex` for details on how the
309 information is discovered.
310 """
311 return ButlerRepoIndex.get_known_repos()
313 @staticmethod
314 def makeRepo(
315 root: ResourcePathExpression,
316 config: Config | str | None = None,
317 dimensionConfig: Config | str | None = None,
318 standalone: bool = False,
319 searchPaths: list[str] | None = None,
320 forceConfigRoot: bool = True,
321 outfile: ResourcePathExpression | None = None,
322 overwrite: bool = False,
323 ) -> Config:
324 """Create an empty data repository by adding a butler.yaml config
325 to a repository root directory.
327 Parameters
328 ----------
329 root : `lsst.resources.ResourcePathExpression`
330 Path or URI to the root location of the new repository. Will be
331 created if it does not exist.
332 config : `Config` or `str`, optional
333 Configuration to write to the repository, after setting any
334 root-dependent Registry or Datastore config options. Can not
335 be a `ButlerConfig` or a `ConfigSubset`. If `None`, default
336 configuration will be used. Root-dependent config options
337 specified in this config are overwritten if ``forceConfigRoot``
338 is `True`.
339 dimensionConfig : `Config` or `str`, optional
340 Configuration for dimensions, will be used to initialize registry
341 database.
342 standalone : `bool`
343 If True, write all expanded defaults, not just customized or
344 repository-specific settings.
345 This (mostly) decouples the repository from the default
346 configuration, insulating it from changes to the defaults (which
347 may be good or bad, depending on the nature of the changes).
348 Future *additions* to the defaults will still be picked up when
349 initializing `Butlers` to repos created with ``standalone=True``.
350 searchPaths : `list` of `str`, optional
351 Directory paths to search when calculating the full butler
352 configuration.
353 forceConfigRoot : `bool`, optional
354 If `False`, any values present in the supplied ``config`` that
355 would normally be reset are not overridden and will appear
356 directly in the output config. This allows non-standard overrides
357 of the root directory for a datastore or registry to be given.
358 If this parameter is `True` the values for ``root`` will be
359 forced into the resulting config if appropriate.
360 outfile : `lss.resources.ResourcePathExpression`, optional
361 If not-`None`, the output configuration will be written to this
362 location rather than into the repository itself. Can be a URI
363 string. Can refer to a directory that will be used to write
364 ``butler.yaml``.
365 overwrite : `bool`, optional
366 Create a new configuration file even if one already exists
367 in the specified output location. Default is to raise
368 an exception.
370 Returns
371 -------
372 config : `Config`
373 The updated `Config` instance written to the repo.
375 Raises
376 ------
377 ValueError
378 Raised if a ButlerConfig or ConfigSubset is passed instead of a
379 regular Config (as these subclasses would make it impossible to
380 support ``standalone=False``).
381 FileExistsError
382 Raised if the output config file already exists.
383 os.error
384 Raised if the directory does not exist, exists but is not a
385 directory, or cannot be created.
387 Notes
388 -----
389 Note that when ``standalone=False`` (the default), the configuration
390 search path (see `ConfigSubset.defaultSearchPaths`) that was used to
391 construct the repository should also be used to construct any Butlers
392 to avoid configuration inconsistencies.
393 """
394 if isinstance(config, (ButlerConfig, ConfigSubset)):
395 raise ValueError("makeRepo must be passed a regular Config without defaults applied.")
397 # Ensure that the root of the repository exists or can be made
398 root_uri = ResourcePath(root, forceDirectory=True)
399 root_uri.mkdir()
401 config = Config(config)
403 # If we are creating a new repo from scratch with relative roots,
404 # do not propagate an explicit root from the config file
405 if "root" in config:
406 del config["root"]
408 full = ButlerConfig(config, searchPaths=searchPaths) # this applies defaults
409 imported_class = doImportType(full["datastore", "cls"])
410 if not issubclass(imported_class, Datastore):
411 raise TypeError(f"Imported datastore class {full['datastore', 'cls']} is not a Datastore")
412 datastoreClass: type[Datastore] = imported_class
413 datastoreClass.setConfigRoot(BUTLER_ROOT_TAG, config, full, overwrite=forceConfigRoot)
415 # if key exists in given config, parse it, otherwise parse the defaults
416 # in the expanded config
417 if config.get(("registry", "db")):
418 registryConfig = RegistryConfig(config)
419 else:
420 registryConfig = RegistryConfig(full)
421 defaultDatabaseUri = registryConfig.makeDefaultDatabaseUri(BUTLER_ROOT_TAG)
422 if defaultDatabaseUri is not None:
423 Config.updateParameters(
424 RegistryConfig, config, full, toUpdate={"db": defaultDatabaseUri}, overwrite=forceConfigRoot
425 )
426 else:
427 Config.updateParameters(RegistryConfig, config, full, toCopy=("db",), overwrite=forceConfigRoot)
429 if standalone:
430 config.merge(full)
431 else:
432 # Always expand the registry.managers section into the per-repo
433 # config, because after the database schema is created, it's not
434 # allowed to change anymore. Note that in the standalone=True
435 # branch, _everything_ in the config is expanded, so there's no
436 # need to special case this.
437 Config.updateParameters(RegistryConfig, config, full, toMerge=("managers",), overwrite=False)
438 configURI: ResourcePathExpression
439 if outfile is not None:
440 # When writing to a separate location we must include
441 # the root of the butler repo in the config else it won't know
442 # where to look.
443 config["root"] = root_uri.geturl()
444 configURI = outfile
445 else:
446 configURI = root_uri
447 # Strip obscore configuration, if it is present, before writing config
448 # to a file, obscore config will be stored in registry.
449 if (obscore_config_key := ("registry", "managers", "obscore", "config")) in config:
450 config_to_write = config.copy()
451 del config_to_write[obscore_config_key]
452 config_to_write.dumpToUri(configURI, overwrite=overwrite)
453 # configFile attribute is updated, need to copy it to original.
454 config.configFile = config_to_write.configFile
455 else:
456 config.dumpToUri(configURI, overwrite=overwrite)
458 # Create Registry and populate tables
459 registryConfig = RegistryConfig(config.get("registry"))
460 dimensionConfig = DimensionConfig(dimensionConfig)
461 Registry.createFromConfig(registryConfig, dimensionConfig=dimensionConfig, butlerRoot=root_uri)
463 log.verbose("Wrote new Butler configuration file to %s", configURI)
465 return config
467 @classmethod
468 def _unpickle(
469 cls,
470 config: ButlerConfig,
471 collections: tuple[str, ...] | None,
472 run: str | None,
473 defaultDataId: dict[str, str],
474 writeable: bool,
475 ) -> Butler:
476 """Callable used to unpickle a Butler.
478 We prefer not to use ``Butler.__init__`` directly so we can force some
479 of its many arguments to be keyword-only (note that ``__reduce__``
480 can only invoke callables with positional arguments).
482 Parameters
483 ----------
484 config : `ButlerConfig`
485 Butler configuration, already coerced into a true `ButlerConfig`
486 instance (and hence after any search paths for overrides have been
487 utilized).
488 collections : `tuple` [ `str` ]
489 Names of the default collections to read from.
490 run : `str`, optional
491 Name of the default `~CollectionType.RUN` collection to write to.
492 defaultDataId : `dict` [ `str`, `str` ]
493 Default data ID values.
494 writeable : `bool`
495 Whether the Butler should support write operations.
497 Returns
498 -------
499 butler : `Butler`
500 A new `Butler` instance.
501 """
502 # MyPy doesn't recognize that the kwargs below are totally valid; it
503 # seems to think '**defaultDataId* is a _positional_ argument!
504 return cls(
505 config=config,
506 collections=collections,
507 run=run,
508 writeable=writeable,
509 **defaultDataId, # type: ignore
510 )
512 def __reduce__(self) -> tuple:
513 """Support pickling."""
514 return (
515 Butler._unpickle,
516 (
517 self._config,
518 self.collections,
519 self.run,
520 self.registry.defaults.dataId.byName(),
521 self.registry.isWriteable(),
522 ),
523 )
525 def __str__(self) -> str:
526 return "Butler(collections={}, run={}, datastore='{}', registry='{}')".format(
527 self.collections, self.run, self._datastore, self.registry
528 )
530 def isWriteable(self) -> bool:
531 """Return `True` if this `Butler` supports write operations."""
532 return self.registry.isWriteable()
534 @contextlib.contextmanager
535 def transaction(self) -> Iterator[None]:
536 """Context manager supporting `Butler` transactions.
538 Transactions can be nested.
539 """
540 with self.registry.transaction():
541 with self._datastore.transaction():
542 yield
544 def _standardizeArgs(
545 self,
546 datasetRefOrType: DatasetRef | DatasetType | str,
547 dataId: DataId | None = None,
548 for_put: bool = True,
549 **kwargs: Any,
550 ) -> tuple[DatasetType, DataId | None]:
551 """Standardize the arguments passed to several Butler APIs.
553 Parameters
554 ----------
555 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
556 When `DatasetRef` the `dataId` should be `None`.
557 Otherwise the `DatasetType` or name thereof.
558 dataId : `dict` or `DataCoordinate`
559 A `dict` of `Dimension` link name, value pairs that label the
560 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
561 should be provided as the second argument.
562 for_put : `bool`, optional
563 If `True` this call is invoked as part of a `Butler.put()`.
564 Otherwise it is assumed to be part of a `Butler.get()`. This
565 parameter is only relevant if there is dataset type
566 inconsistency.
567 **kwargs
568 Additional keyword arguments used to augment or construct a
569 `DataCoordinate`. See `DataCoordinate.standardize`
570 parameters.
572 Returns
573 -------
574 datasetType : `DatasetType`
575 A `DatasetType` instance extracted from ``datasetRefOrType``.
576 dataId : `dict` or `DataId`, optional
577 Argument that can be used (along with ``kwargs``) to construct a
578 `DataId`.
580 Notes
581 -----
582 Butler APIs that conceptually need a DatasetRef also allow passing a
583 `DatasetType` (or the name of one) and a `DataId` (or a dict and
584 keyword arguments that can be used to construct one) separately. This
585 method accepts those arguments and always returns a true `DatasetType`
586 and a `DataId` or `dict`.
588 Standardization of `dict` vs `DataId` is best handled by passing the
589 returned ``dataId`` (and ``kwargs``) to `Registry` APIs, which are
590 generally similarly flexible.
591 """
592 externalDatasetType: DatasetType | None = None
593 internalDatasetType: DatasetType | None = None
594 if isinstance(datasetRefOrType, DatasetRef):
595 if dataId is not None or kwargs:
596 raise ValueError("DatasetRef given, cannot use dataId as well")
597 externalDatasetType = datasetRefOrType.datasetType
598 dataId = datasetRefOrType.dataId
599 else:
600 # Don't check whether DataId is provided, because Registry APIs
601 # can usually construct a better error message when it wasn't.
602 if isinstance(datasetRefOrType, DatasetType):
603 externalDatasetType = datasetRefOrType
604 else:
605 internalDatasetType = self.registry.getDatasetType(datasetRefOrType)
607 # Check that they are self-consistent
608 if externalDatasetType is not None:
609 internalDatasetType = self.registry.getDatasetType(externalDatasetType.name)
610 if externalDatasetType != internalDatasetType:
611 # We can allow differences if they are compatible, depending
612 # on whether this is a get or a put. A get requires that
613 # the python type associated with the datastore can be
614 # converted to the user type. A put requires that the user
615 # supplied python type can be converted to the internal
616 # type expected by registry.
617 relevantDatasetType = internalDatasetType
618 if for_put:
619 is_compatible = internalDatasetType.is_compatible_with(externalDatasetType)
620 else:
621 is_compatible = externalDatasetType.is_compatible_with(internalDatasetType)
622 relevantDatasetType = externalDatasetType
623 if not is_compatible:
624 raise ValueError(
625 f"Supplied dataset type ({externalDatasetType}) inconsistent with "
626 f"registry definition ({internalDatasetType})"
627 )
628 # Override the internal definition.
629 internalDatasetType = relevantDatasetType
631 assert internalDatasetType is not None
632 return internalDatasetType, dataId
634 def _rewrite_data_id(
635 self, dataId: DataId | None, datasetType: DatasetType, **kwargs: Any
636 ) -> tuple[DataId | None, dict[str, Any]]:
637 """Rewrite a data ID taking into account dimension records.
639 Take a Data ID and keyword args and rewrite it if necessary to
640 allow the user to specify dimension records rather than dimension
641 primary values.
643 This allows a user to include a dataId dict with keys of
644 ``exposure.day_obs`` and ``exposure.seq_num`` instead of giving
645 the integer exposure ID. It also allows a string to be given
646 for a dimension value rather than the integer ID if that is more
647 convenient. For example, rather than having to specifying the
648 detector with ``detector.full_name``, a string given for ``detector``
649 will be interpreted as the full name and converted to the integer
650 value.
652 Keyword arguments can also use strings for dimensions like detector
653 and exposure but python does not allow them to include ``.`` and
654 so the ``exposure.day_obs`` syntax can not be used in a keyword
655 argument.
657 Parameters
658 ----------
659 dataId : `dict` or `DataCoordinate`
660 A `dict` of `Dimension` link name, value pairs that will label the
661 `DatasetRef` within a Collection.
662 datasetType : `DatasetType`
663 The dataset type associated with this dataId. Required to
664 determine the relevant dimensions.
665 **kwargs
666 Additional keyword arguments used to augment or construct a
667 `DataId`. See `DataId` parameters.
669 Returns
670 -------
671 dataId : `dict` or `DataCoordinate`
672 The, possibly rewritten, dataId. If given a `DataCoordinate` and
673 no keyword arguments, the original dataId will be returned
674 unchanged.
675 **kwargs : `dict`
676 Any unused keyword arguments (would normally be empty dict).
677 """
678 # Do nothing if we have a standalone DataCoordinate.
679 if isinstance(dataId, DataCoordinate) and not kwargs:
680 return dataId, kwargs
682 # Process dimension records that are using record information
683 # rather than ids
684 newDataId: dict[str, DataIdValue] = {}
685 byRecord: dict[str, dict[str, Any]] = defaultdict(dict)
687 # if all the dataId comes from keyword parameters we do not need
688 # to do anything here because they can't be of the form
689 # exposure.obs_id because a "." is not allowed in a keyword parameter.
690 if dataId:
691 for k, v in dataId.items():
692 # If we have a Dimension we do not need to do anything
693 # because it cannot be a compound key.
694 if isinstance(k, str) and "." in k:
695 # Someone is using a more human-readable dataId
696 dimensionName, record = k.split(".", 1)
697 byRecord[dimensionName][record] = v
698 elif isinstance(k, Dimension):
699 newDataId[k.name] = v
700 else:
701 newDataId[k] = v
703 # Go through the updated dataId and check the type in case someone is
704 # using an alternate key. We have already filtered out the compound
705 # keys dimensions.record format.
706 not_dimensions = {}
708 # Will need to look in the dataId and the keyword arguments
709 # and will remove them if they need to be fixed or are unrecognized.
710 for dataIdDict in (newDataId, kwargs):
711 # Use a list so we can adjust the dict safely in the loop
712 for dimensionName in list(dataIdDict):
713 value = dataIdDict[dimensionName]
714 try:
715 dimension = self.dimensions.getStaticDimensions()[dimensionName]
716 except KeyError:
717 # This is not a real dimension
718 not_dimensions[dimensionName] = value
719 del dataIdDict[dimensionName]
720 continue
722 # Convert an integral type to an explicit int to simplify
723 # comparisons here
724 if isinstance(value, numbers.Integral):
725 value = int(value)
727 if not isinstance(value, dimension.primaryKey.getPythonType()):
728 for alternate in dimension.alternateKeys:
729 if isinstance(value, alternate.getPythonType()):
730 byRecord[dimensionName][alternate.name] = value
731 del dataIdDict[dimensionName]
732 log.debug(
733 "Converting dimension %s to %s.%s=%s",
734 dimensionName,
735 dimensionName,
736 alternate.name,
737 value,
738 )
739 break
740 else:
741 log.warning(
742 "Type mismatch found for value '%r' provided for dimension %s. "
743 "Could not find matching alternative (primary key has type %s) "
744 "so attempting to use as-is.",
745 value,
746 dimensionName,
747 dimension.primaryKey.getPythonType(),
748 )
750 # By this point kwargs and newDataId should only include valid
751 # dimensions. Merge kwargs in to the new dataId and log if there
752 # are dimensions in both (rather than calling update).
753 for k, v in kwargs.items():
754 if k in newDataId and newDataId[k] != v:
755 log.debug(
756 "Keyword arg %s overriding explicit value in dataId of %s with %s", k, newDataId[k], v
757 )
758 newDataId[k] = v
759 # No need to retain any values in kwargs now.
760 kwargs = {}
762 # If we have some unrecognized dimensions we have to try to connect
763 # them to records in other dimensions. This is made more complicated
764 # by some dimensions having records with clashing names. A mitigation
765 # is that we can tell by this point which dimensions are missing
766 # for the DatasetType but this does not work for calibrations
767 # where additional dimensions can be used to constrain the temporal
768 # axis.
769 if not_dimensions:
770 # Search for all dimensions even if we have been given a value
771 # explicitly. In some cases records are given as well as the
772 # actually dimension and this should not be an error if they
773 # match.
774 mandatoryDimensions = datasetType.dimensions.names # - provided
776 candidateDimensions: set[str] = set()
777 candidateDimensions.update(mandatoryDimensions)
779 # For calibrations we may well be needing temporal dimensions
780 # so rather than always including all dimensions in the scan
781 # restrict things a little. It is still possible for there
782 # to be confusion over day_obs in visit vs exposure for example.
783 # If we are not searching calibration collections things may
784 # fail but they are going to fail anyway because of the
785 # ambiguousness of the dataId...
786 if datasetType.isCalibration():
787 for dim in self.dimensions.getStaticDimensions():
788 if dim.temporal:
789 candidateDimensions.add(str(dim))
791 # Look up table for the first association with a dimension
792 guessedAssociation: dict[str, dict[str, Any]] = defaultdict(dict)
794 # Keep track of whether an item is associated with multiple
795 # dimensions.
796 counter: Counter[str] = Counter()
797 assigned: dict[str, set[str]] = defaultdict(set)
799 # Go through the missing dimensions and associate the
800 # given names with records within those dimensions
801 matched_dims = set()
802 for dimensionName in candidateDimensions:
803 dimension = self.dimensions.getStaticDimensions()[dimensionName]
804 fields = dimension.metadata.names | dimension.uniqueKeys.names
805 for field in not_dimensions:
806 if field in fields:
807 guessedAssociation[dimensionName][field] = not_dimensions[field]
808 counter[dimensionName] += 1
809 assigned[field].add(dimensionName)
810 matched_dims.add(field)
812 # Calculate the fields that matched nothing.
813 never_found = set(not_dimensions) - matched_dims
815 if never_found:
816 raise ValueError(f"Unrecognized keyword args given: {never_found}")
818 # There is a chance we have allocated a single dataId item
819 # to multiple dimensions. Need to decide which should be retained.
820 # For now assume that the most popular alternative wins.
821 # This means that day_obs with seq_num will result in
822 # exposure.day_obs and not visit.day_obs
823 # Also prefer an explicitly missing dimension over an inferred
824 # temporal dimension.
825 for fieldName, assignedDimensions in assigned.items():
826 if len(assignedDimensions) > 1:
827 # Pick the most popular (preferring mandatory dimensions)
828 requiredButMissing = assignedDimensions.intersection(mandatoryDimensions)
829 if requiredButMissing:
830 candidateDimensions = requiredButMissing
831 else:
832 candidateDimensions = assignedDimensions
834 # If this is a choice between visit and exposure and
835 # neither was a required part of the dataset type,
836 # (hence in this branch) always prefer exposure over
837 # visit since exposures are always defined and visits
838 # are defined from exposures.
839 if candidateDimensions == {"exposure", "visit"}:
840 candidateDimensions = {"exposure"}
842 # Select the relevant items and get a new restricted
843 # counter.
844 theseCounts = {k: v for k, v in counter.items() if k in candidateDimensions}
845 duplicatesCounter: Counter[str] = Counter()
846 duplicatesCounter.update(theseCounts)
848 # Choose the most common. If they are equally common
849 # we will pick the one that was found first.
850 # Returns a list of tuples
851 selected = duplicatesCounter.most_common(1)[0][0]
853 log.debug(
854 "Ambiguous dataId entry '%s' associated with multiple dimensions: %s."
855 " Removed ambiguity by choosing dimension %s.",
856 fieldName,
857 ", ".join(assignedDimensions),
858 selected,
859 )
861 for candidateDimension in assignedDimensions:
862 if candidateDimension != selected:
863 del guessedAssociation[candidateDimension][fieldName]
865 # Update the record look up dict with the new associations
866 for dimensionName, values in guessedAssociation.items():
867 if values: # A dict might now be empty
868 log.debug("Assigned non-dimension dataId keys to dimension %s: %s", dimensionName, values)
869 byRecord[dimensionName].update(values)
871 if byRecord:
872 # Some record specifiers were found so we need to convert
873 # them to the Id form
874 for dimensionName, values in byRecord.items():
875 if dimensionName in newDataId:
876 log.debug(
877 "DataId specified explicit %s dimension value of %s in addition to"
878 " general record specifiers for it of %s. Ignoring record information.",
879 dimensionName,
880 newDataId[dimensionName],
881 str(values),
882 )
883 # Get the actual record and compare with these values.
884 try:
885 recs = list(self.registry.queryDimensionRecords(dimensionName, dataId=newDataId))
886 except DataIdError:
887 raise ValueError(
888 f"Could not find dimension '{dimensionName}'"
889 f" with dataId {newDataId} as part of comparing with"
890 f" record values {byRecord[dimensionName]}"
891 ) from None
892 if len(recs) == 1:
893 errmsg: list[str] = []
894 for k, v in values.items():
895 if (recval := getattr(recs[0], k)) != v:
896 errmsg.append(f"{k}({recval} != {v})")
897 if errmsg:
898 raise ValueError(
899 f"Dimension {dimensionName} in dataId has explicit value"
900 " inconsistent with records: " + ", ".join(errmsg)
901 )
902 else:
903 # Multiple matches for an explicit dimension
904 # should never happen but let downstream complain.
905 pass
906 continue
908 # Build up a WHERE expression
909 bind = {k: v for k, v in values.items()}
910 where = " AND ".join(f"{dimensionName}.{k} = {k}" for k in bind)
912 # Hopefully we get a single record that matches
913 records = set(
914 self.registry.queryDimensionRecords(
915 dimensionName, dataId=newDataId, where=where, bind=bind, **kwargs
916 )
917 )
919 if len(records) != 1:
920 if len(records) > 1:
921 # visit can have an ambiguous answer without involving
922 # visit_system. The default visit_system is defined
923 # by the instrument.
924 if (
925 dimensionName == "visit"
926 and "visit_system_membership" in self.dimensions
927 and "visit_system" in self.dimensions["instrument"].metadata
928 ):
929 instrument_records = list(
930 self.registry.queryDimensionRecords(
931 "instrument",
932 dataId=newDataId,
933 **kwargs,
934 )
935 )
936 if len(instrument_records) == 1:
937 visit_system = instrument_records[0].visit_system
938 if visit_system is None:
939 # Set to a value that will never match.
940 visit_system = -1
942 # Look up each visit in the
943 # visit_system_membership records.
944 for rec in records:
945 membership = list(
946 self.registry.queryDimensionRecords(
947 # Use bind to allow zero results.
948 # This is a fully-specified query.
949 "visit_system_membership",
950 where="instrument = inst AND visit_system = system AND visit = v",
951 bind=dict(
952 inst=instrument_records[0].name, system=visit_system, v=rec.id
953 ),
954 )
955 )
956 if membership:
957 # This record is the right answer.
958 records = {rec}
959 break
961 # The ambiguity may have been resolved so check again.
962 if len(records) > 1:
963 log.debug("Received %d records from constraints of %s", len(records), str(values))
964 for r in records:
965 log.debug("- %s", str(r))
966 raise ValueError(
967 f"DataId specification for dimension {dimensionName} is not"
968 f" uniquely constrained to a single dataset by {values}."
969 f" Got {len(records)} results."
970 )
971 else:
972 raise ValueError(
973 f"DataId specification for dimension {dimensionName} matched no"
974 f" records when constrained by {values}"
975 )
977 # Get the primary key from the real dimension object
978 dimension = self.dimensions.getStaticDimensions()[dimensionName]
979 if not isinstance(dimension, Dimension):
980 raise RuntimeError(
981 f"{dimension.name} is not a true dimension, and cannot be used in data IDs."
982 )
983 newDataId[dimensionName] = getattr(records.pop(), dimension.primaryKey.name)
985 return newDataId, kwargs
987 def _findDatasetRef(
988 self,
989 datasetRefOrType: DatasetRef | DatasetType | str,
990 dataId: DataId | None = None,
991 *,
992 collections: Any = None,
993 predict: bool = False,
994 run: str | None = None,
995 **kwargs: Any,
996 ) -> DatasetRef:
997 """Shared logic for methods that start with a search for a dataset in
998 the registry.
1000 Parameters
1001 ----------
1002 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1003 When `DatasetRef` the `dataId` should be `None`.
1004 Otherwise the `DatasetType` or name thereof.
1005 dataId : `dict` or `DataCoordinate`, optional
1006 A `dict` of `Dimension` link name, value pairs that label the
1007 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1008 should be provided as the first argument.
1009 collections : Any, optional
1010 Collections to be searched, overriding ``self.collections``.
1011 Can be any of the types supported by the ``collections`` argument
1012 to butler construction.
1013 predict : `bool`, optional
1014 If `True`, return a newly created `DatasetRef` with a unique
1015 dataset ID if finding a reference in the `Registry` fails.
1016 Defaults to `False`.
1017 run : `str`, optional
1018 Run collection name to use for creating `DatasetRef` for predicted
1019 datasets. Only used if ``predict`` is `True`.
1020 **kwargs
1021 Additional keyword arguments used to augment or construct a
1022 `DataId`. See `DataId` parameters.
1024 Returns
1025 -------
1026 ref : `DatasetRef`
1027 A reference to the dataset identified by the given arguments.
1028 This can be the same dataset reference as given if it was
1029 resolved.
1031 Raises
1032 ------
1033 LookupError
1034 Raised if no matching dataset exists in the `Registry` (and
1035 ``predict`` is `False`).
1036 ValueError
1037 Raised if a resolved `DatasetRef` was passed as an input, but it
1038 differs from the one found in the registry.
1039 TypeError
1040 Raised if no collections were provided.
1041 """
1042 datasetType, dataId = self._standardizeArgs(datasetRefOrType, dataId, for_put=False, **kwargs)
1043 if isinstance(datasetRefOrType, DatasetRef):
1044 if collections is not None:
1045 warnings.warn("Collections should not be specified with DatasetRef", stacklevel=3)
1046 return datasetRefOrType
1047 timespan: Timespan | None = None
1049 dataId, kwargs = self._rewrite_data_id(dataId, datasetType, **kwargs)
1051 if datasetType.isCalibration():
1052 # Because this is a calibration dataset, first try to make a
1053 # standardize the data ID without restricting the dimensions to
1054 # those of the dataset type requested, because there may be extra
1055 # dimensions that provide temporal information for a validity-range
1056 # lookup.
1057 dataId = DataCoordinate.standardize(
1058 dataId, universe=self.dimensions, defaults=self.registry.defaults.dataId, **kwargs
1059 )
1060 if dataId.graph.temporal:
1061 dataId = self.registry.expandDataId(dataId)
1062 timespan = dataId.timespan
1063 else:
1064 # Standardize the data ID to just the dimensions of the dataset
1065 # type instead of letting registry.findDataset do it, so we get the
1066 # result even if no dataset is found.
1067 dataId = DataCoordinate.standardize(
1068 dataId, graph=datasetType.dimensions, defaults=self.registry.defaults.dataId, **kwargs
1069 )
1070 # Always lookup the DatasetRef, even if one is given, to ensure it is
1071 # present in the current collection.
1072 ref = self.registry.findDataset(datasetType, dataId, collections=collections, timespan=timespan)
1073 if ref is None:
1074 if predict:
1075 if run is None:
1076 run = self.run
1077 if run is None:
1078 raise TypeError("Cannot predict dataset ID/location with run=None.")
1079 return DatasetRef(datasetType, dataId, run=run)
1080 else:
1081 if collections is None:
1082 collections = self.registry.defaults.collections
1083 raise LookupError(
1084 f"Dataset {datasetType.name} with data ID {dataId} "
1085 f"could not be found in collections {collections}."
1086 )
1087 if datasetType != ref.datasetType:
1088 # If they differ it is because the user explicitly specified
1089 # a compatible dataset type to this call rather than using the
1090 # registry definition. The DatasetRef must therefore be recreated
1091 # using the user definition such that the expected type is
1092 # returned.
1093 ref = DatasetRef(datasetType, ref.dataId, run=ref.run, id=ref.id)
1095 return ref
1097 @transactional
1098 @deprecated(
1099 reason="Butler.put() now behaves like Butler.putDirect() when given a DatasetRef."
1100 " Please use Butler.put(). Be aware that you may need to adjust your usage if you"
1101 " were relying on the run parameter to determine the run."
1102 " Will be removed after v27.0.",
1103 version="v26.0",
1104 category=FutureWarning,
1105 )
1106 def putDirect(self, obj: Any, ref: DatasetRef, /) -> DatasetRef:
1107 # Docstring inherited.
1108 return self.put(obj, ref)
1110 @transactional
1111 def put(
1112 self,
1113 obj: Any,
1114 datasetRefOrType: DatasetRef | DatasetType | str,
1115 /,
1116 dataId: DataId | None = None,
1117 *,
1118 run: str | None = None,
1119 **kwargs: Any,
1120 ) -> DatasetRef:
1121 """Store and register a dataset.
1123 Parameters
1124 ----------
1125 obj : `object`
1126 The dataset.
1127 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1128 When `DatasetRef` is provided, ``dataId`` should be `None`.
1129 Otherwise the `DatasetType` or name thereof. If a fully resolved
1130 `DatasetRef` is given the run and ID are used directly.
1131 dataId : `dict` or `DataCoordinate`
1132 A `dict` of `Dimension` link name, value pairs that label the
1133 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1134 should be provided as the second argument.
1135 run : `str`, optional
1136 The name of the run the dataset should be added to, overriding
1137 ``self.run``. Not used if a resolved `DatasetRef` is provided.
1138 **kwargs
1139 Additional keyword arguments used to augment or construct a
1140 `DataCoordinate`. See `DataCoordinate.standardize`
1141 parameters. Not used if a resolve `DatasetRef` is provided.
1143 Returns
1144 -------
1145 ref : `DatasetRef`
1146 A reference to the stored dataset, updated with the correct id if
1147 given.
1149 Raises
1150 ------
1151 TypeError
1152 Raised if the butler is read-only or if no run has been provided.
1153 """
1154 if isinstance(datasetRefOrType, DatasetRef):
1155 # This is a direct put of predefined DatasetRef.
1156 log.debug("Butler put direct: %s", datasetRefOrType)
1157 if run is not None:
1158 warnings.warn("Run collection is not used for DatasetRef", stacklevel=3)
1159 # If registry already has a dataset with the same dataset ID,
1160 # dataset type and DataId, then _importDatasets will do nothing and
1161 # just return an original ref. We have to raise in this case, there
1162 # is a datastore check below for that.
1163 self.registry._importDatasets([datasetRefOrType], expand=True)
1164 # Before trying to write to the datastore check that it does not
1165 # know this dataset. This is prone to races, of course.
1166 if self._datastore.knows(datasetRefOrType):
1167 raise ConflictingDefinitionError(f"Datastore already contains dataset: {datasetRefOrType}")
1168 # Try to write dataset to the datastore, if it fails due to a race
1169 # with another write, the content of stored data may be
1170 # unpredictable.
1171 try:
1172 self._datastore.put(obj, datasetRefOrType)
1173 except IntegrityError as e:
1174 raise ConflictingDefinitionError(f"Datastore already contains dataset: {e}")
1175 return datasetRefOrType
1177 log.debug("Butler put: %s, dataId=%s, run=%s", datasetRefOrType, dataId, run)
1178 if not self.isWriteable():
1179 raise TypeError("Butler is read-only.")
1180 datasetType, dataId = self._standardizeArgs(datasetRefOrType, dataId, **kwargs)
1182 # Handle dimension records in dataId
1183 dataId, kwargs = self._rewrite_data_id(dataId, datasetType, **kwargs)
1185 # Add Registry Dataset entry.
1186 dataId = self.registry.expandDataId(dataId, graph=datasetType.dimensions, **kwargs)
1187 (ref,) = self.registry.insertDatasets(datasetType, run=run, dataIds=[dataId])
1188 self._datastore.put(obj, ref)
1190 return ref
1192 @deprecated(
1193 reason="Butler.get() now behaves like Butler.getDirect() when given a DatasetRef."
1194 " Please use Butler.get(). Will be removed after v27.0.",
1195 version="v26.0",
1196 category=FutureWarning,
1197 )
1198 def getDirect(
1199 self,
1200 ref: DatasetRef,
1201 *,
1202 parameters: dict[str, Any] | None = None,
1203 storageClass: StorageClass | str | None = None,
1204 ) -> Any:
1205 """Retrieve a stored dataset.
1207 Parameters
1208 ----------
1209 ref : `DatasetRef`
1210 Resolved reference to an already stored dataset.
1211 parameters : `dict`
1212 Additional StorageClass-defined options to control reading,
1213 typically used to efficiently read only a subset of the dataset.
1214 storageClass : `StorageClass` or `str`, optional
1215 The storage class to be used to override the Python type
1216 returned by this method. By default the returned type matches
1217 the dataset type definition for this dataset. Specifying a
1218 read `StorageClass` can force a different type to be returned.
1219 This type must be compatible with the original type.
1221 Returns
1222 -------
1223 obj : `object`
1224 The dataset.
1225 """
1226 return self._datastore.get(ref, parameters=parameters, storageClass=storageClass)
1228 @deprecated(
1229 reason="Butler.getDeferred() now behaves like getDirectDeferred() when given a DatasetRef. "
1230 "Please use Butler.getDeferred(). Will be removed after v27.0.",
1231 version="v26.0",
1232 category=FutureWarning,
1233 )
1234 def getDirectDeferred(
1235 self,
1236 ref: DatasetRef,
1237 *,
1238 parameters: dict | None = None,
1239 storageClass: str | StorageClass | None = None,
1240 ) -> DeferredDatasetHandle:
1241 """Create a `DeferredDatasetHandle` which can later retrieve a dataset,
1242 from a resolved `DatasetRef`.
1244 Parameters
1245 ----------
1246 ref : `DatasetRef`
1247 Resolved reference to an already stored dataset.
1248 parameters : `dict`
1249 Additional StorageClass-defined options to control reading,
1250 typically used to efficiently read only a subset of the dataset.
1251 storageClass : `StorageClass` or `str`, optional
1252 The storage class to be used to override the Python type
1253 returned by this method. By default the returned type matches
1254 the dataset type definition for this dataset. Specifying a
1255 read `StorageClass` can force a different type to be returned.
1256 This type must be compatible with the original type.
1258 Returns
1259 -------
1260 obj : `DeferredDatasetHandle`
1261 A handle which can be used to retrieve a dataset at a later time.
1263 Raises
1264 ------
1265 LookupError
1266 Raised if no matching dataset exists in the `Registry`.
1267 """
1268 # Check that dataset actually exists.
1269 if not self._datastore.exists(ref):
1270 raise LookupError(f"Dataset reference {ref} does not exist.")
1271 return DeferredDatasetHandle(butler=self, ref=ref, parameters=parameters, storageClass=storageClass)
1273 def getDeferred(
1274 self,
1275 datasetRefOrType: DatasetRef | DatasetType | str,
1276 /,
1277 dataId: DataId | None = None,
1278 *,
1279 parameters: dict | None = None,
1280 collections: Any = None,
1281 storageClass: str | StorageClass | None = None,
1282 **kwargs: Any,
1283 ) -> DeferredDatasetHandle:
1284 """Create a `DeferredDatasetHandle` which can later retrieve a dataset,
1285 after an immediate registry lookup.
1287 Parameters
1288 ----------
1289 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1290 When `DatasetRef` the `dataId` should be `None`.
1291 Otherwise the `DatasetType` or name thereof.
1292 dataId : `dict` or `DataCoordinate`, optional
1293 A `dict` of `Dimension` link name, value pairs that label the
1294 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1295 should be provided as the first argument.
1296 parameters : `dict`
1297 Additional StorageClass-defined options to control reading,
1298 typically used to efficiently read only a subset of the dataset.
1299 collections : Any, optional
1300 Collections to be searched, overriding ``self.collections``.
1301 Can be any of the types supported by the ``collections`` argument
1302 to butler construction.
1303 storageClass : `StorageClass` or `str`, optional
1304 The storage class to be used to override the Python type
1305 returned by this method. By default the returned type matches
1306 the dataset type definition for this dataset. Specifying a
1307 read `StorageClass` can force a different type to be returned.
1308 This type must be compatible with the original type.
1309 **kwargs
1310 Additional keyword arguments used to augment or construct a
1311 `DataId`. See `DataId` parameters.
1313 Returns
1314 -------
1315 obj : `DeferredDatasetHandle`
1316 A handle which can be used to retrieve a dataset at a later time.
1318 Raises
1319 ------
1320 LookupError
1321 Raised if no matching dataset exists in the `Registry`.
1322 ValueError
1323 Raised if a resolved `DatasetRef` was passed as an input, but it
1324 differs from the one found in the registry.
1325 TypeError
1326 Raised if no collections were provided.
1327 """
1328 if isinstance(datasetRefOrType, DatasetRef) and not self._datastore.exists(datasetRefOrType):
1329 raise LookupError(f"Dataset reference {datasetRefOrType} does not exist.")
1330 ref = self._findDatasetRef(datasetRefOrType, dataId, collections=collections, **kwargs)
1331 return DeferredDatasetHandle(butler=self, ref=ref, parameters=parameters, storageClass=storageClass)
1333 def get(
1334 self,
1335 datasetRefOrType: DatasetRef | DatasetType | str,
1336 /,
1337 dataId: DataId | None = None,
1338 *,
1339 parameters: dict[str, Any] | None = None,
1340 collections: Any = None,
1341 storageClass: StorageClass | str | None = None,
1342 **kwargs: Any,
1343 ) -> Any:
1344 """Retrieve a stored dataset.
1346 Parameters
1347 ----------
1348 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1349 When `DatasetRef` the `dataId` should be `None`.
1350 Otherwise the `DatasetType` or name thereof.
1351 If a resolved `DatasetRef`, the associated dataset
1352 is returned directly without additional querying.
1353 dataId : `dict` or `DataCoordinate`
1354 A `dict` of `Dimension` link name, value pairs that label the
1355 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1356 should be provided as the first argument.
1357 parameters : `dict`
1358 Additional StorageClass-defined options to control reading,
1359 typically used to efficiently read only a subset of the dataset.
1360 collections : Any, optional
1361 Collections to be searched, overriding ``self.collections``.
1362 Can be any of the types supported by the ``collections`` argument
1363 to butler construction.
1364 storageClass : `StorageClass` or `str`, optional
1365 The storage class to be used to override the Python type
1366 returned by this method. By default the returned type matches
1367 the dataset type definition for this dataset. Specifying a
1368 read `StorageClass` can force a different type to be returned.
1369 This type must be compatible with the original type.
1370 **kwargs
1371 Additional keyword arguments used to augment or construct a
1372 `DataCoordinate`. See `DataCoordinate.standardize`
1373 parameters.
1375 Returns
1376 -------
1377 obj : `object`
1378 The dataset.
1380 Raises
1381 ------
1382 LookupError
1383 Raised if no matching dataset exists in the `Registry`.
1384 TypeError
1385 Raised if no collections were provided.
1387 Notes
1388 -----
1389 When looking up datasets in a `~CollectionType.CALIBRATION` collection,
1390 this method requires that the given data ID include temporal dimensions
1391 beyond the dimensions of the dataset type itself, in order to find the
1392 dataset with the appropriate validity range. For example, a "bias"
1393 dataset with native dimensions ``{instrument, detector}`` could be
1394 fetched with a ``{instrument, detector, exposure}`` data ID, because
1395 ``exposure`` is a temporal dimension.
1396 """
1397 log.debug("Butler get: %s, dataId=%s, parameters=%s", datasetRefOrType, dataId, parameters)
1398 ref = self._findDatasetRef(datasetRefOrType, dataId, collections=collections, **kwargs)
1399 return self._datastore.get(ref, parameters=parameters, storageClass=storageClass)
1401 def getURIs(
1402 self,
1403 datasetRefOrType: DatasetRef | DatasetType | str,
1404 /,
1405 dataId: DataId | None = None,
1406 *,
1407 predict: bool = False,
1408 collections: Any = None,
1409 run: str | None = None,
1410 **kwargs: Any,
1411 ) -> DatasetRefURIs:
1412 """Return the URIs associated with the dataset.
1414 Parameters
1415 ----------
1416 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1417 When `DatasetRef` the `dataId` should be `None`.
1418 Otherwise the `DatasetType` or name thereof.
1419 dataId : `dict` or `DataCoordinate`
1420 A `dict` of `Dimension` link name, value pairs that label the
1421 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1422 should be provided as the first argument.
1423 predict : `bool`
1424 If `True`, allow URIs to be returned of datasets that have not
1425 been written.
1426 collections : Any, optional
1427 Collections to be searched, overriding ``self.collections``.
1428 Can be any of the types supported by the ``collections`` argument
1429 to butler construction.
1430 run : `str`, optional
1431 Run to use for predictions, overriding ``self.run``.
1432 **kwargs
1433 Additional keyword arguments used to augment or construct a
1434 `DataCoordinate`. See `DataCoordinate.standardize`
1435 parameters.
1437 Returns
1438 -------
1439 uris : `DatasetRefURIs`
1440 The URI to the primary artifact associated with this dataset (if
1441 the dataset was disassembled within the datastore this may be
1442 `None`), and the URIs to any components associated with the dataset
1443 artifact. (can be empty if there are no components).
1444 """
1445 ref = self._findDatasetRef(
1446 datasetRefOrType, dataId, predict=predict, run=run, collections=collections, **kwargs
1447 )
1448 return self._datastore.getURIs(ref, predict)
1450 def getURI(
1451 self,
1452 datasetRefOrType: DatasetRef | DatasetType | str,
1453 /,
1454 dataId: DataId | None = None,
1455 *,
1456 predict: bool = False,
1457 collections: Any = None,
1458 run: str | None = None,
1459 **kwargs: Any,
1460 ) -> ResourcePath:
1461 """Return the URI to the Dataset.
1463 Parameters
1464 ----------
1465 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1466 When `DatasetRef` the `dataId` should be `None`.
1467 Otherwise the `DatasetType` or name thereof.
1468 dataId : `dict` or `DataCoordinate`
1469 A `dict` of `Dimension` link name, value pairs that label the
1470 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1471 should be provided as the first argument.
1472 predict : `bool`
1473 If `True`, allow URIs to be returned of datasets that have not
1474 been written.
1475 collections : Any, optional
1476 Collections to be searched, overriding ``self.collections``.
1477 Can be any of the types supported by the ``collections`` argument
1478 to butler construction.
1479 run : `str`, optional
1480 Run to use for predictions, overriding ``self.run``.
1481 **kwargs
1482 Additional keyword arguments used to augment or construct a
1483 `DataCoordinate`. See `DataCoordinate.standardize`
1484 parameters.
1486 Returns
1487 -------
1488 uri : `lsst.resources.ResourcePath`
1489 URI pointing to the Dataset within the datastore. If the
1490 Dataset does not exist in the datastore, and if ``predict`` is
1491 `True`, the URI will be a prediction and will include a URI
1492 fragment "#predicted".
1493 If the datastore does not have entities that relate well
1494 to the concept of a URI the returned URI string will be
1495 descriptive. The returned URI is not guaranteed to be obtainable.
1497 Raises
1498 ------
1499 LookupError
1500 A URI has been requested for a dataset that does not exist and
1501 guessing is not allowed.
1502 ValueError
1503 Raised if a resolved `DatasetRef` was passed as an input, but it
1504 differs from the one found in the registry.
1505 TypeError
1506 Raised if no collections were provided.
1507 RuntimeError
1508 Raised if a URI is requested for a dataset that consists of
1509 multiple artifacts.
1510 """
1511 primary, components = self.getURIs(
1512 datasetRefOrType, dataId=dataId, predict=predict, collections=collections, run=run, **kwargs
1513 )
1515 if primary is None or components:
1516 raise RuntimeError(
1517 f"Dataset ({datasetRefOrType}) includes distinct URIs for components. "
1518 "Use Butler.getURIs() instead."
1519 )
1520 return primary
1522 def retrieveArtifacts(
1523 self,
1524 refs: Iterable[DatasetRef],
1525 destination: ResourcePathExpression,
1526 transfer: str = "auto",
1527 preserve_path: bool = True,
1528 overwrite: bool = False,
1529 ) -> list[ResourcePath]:
1530 """Retrieve the artifacts associated with the supplied refs.
1532 Parameters
1533 ----------
1534 refs : iterable of `DatasetRef`
1535 The datasets for which artifacts are to be retrieved.
1536 A single ref can result in multiple artifacts. The refs must
1537 be resolved.
1538 destination : `lsst.resources.ResourcePath` or `str`
1539 Location to write the artifacts.
1540 transfer : `str`, optional
1541 Method to use to transfer the artifacts. Must be one of the options
1542 supported by `~lsst.resources.ResourcePath.transfer_from()`.
1543 "move" is not allowed.
1544 preserve_path : `bool`, optional
1545 If `True` the full path of the artifact within the datastore
1546 is preserved. If `False` the final file component of the path
1547 is used.
1548 overwrite : `bool`, optional
1549 If `True` allow transfers to overwrite existing files at the
1550 destination.
1552 Returns
1553 -------
1554 targets : `list` of `lsst.resources.ResourcePath`
1555 URIs of file artifacts in destination location. Order is not
1556 preserved.
1558 Notes
1559 -----
1560 For non-file datastores the artifacts written to the destination
1561 may not match the representation inside the datastore. For example
1562 a hierarchical data structure in a NoSQL database may well be stored
1563 as a JSON file.
1564 """
1565 return self._datastore.retrieveArtifacts(
1566 refs,
1567 ResourcePath(destination),
1568 transfer=transfer,
1569 preserve_path=preserve_path,
1570 overwrite=overwrite,
1571 )
1573 def exists(
1574 self,
1575 dataset_ref_or_type: DatasetRef | DatasetType | str,
1576 /,
1577 data_id: DataId | None = None,
1578 *,
1579 full_check: bool = True,
1580 collections: Any = None,
1581 **kwargs: Any,
1582 ) -> DatasetExistence:
1583 """Indicate whether a dataset is known to Butler registry and
1584 datastore.
1586 Parameters
1587 ----------
1588 dataset_ref_or_type : `DatasetRef`, `DatasetType`, or `str`
1589 When `DatasetRef` the `dataId` should be `None`.
1590 Otherwise the `DatasetType` or name thereof.
1591 data_id : `dict` or `DataCoordinate`
1592 A `dict` of `Dimension` link name, value pairs that label the
1593 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1594 should be provided as the first argument.
1595 full_check : `bool`, optional
1596 If `True`, an additional check will be made for dataset artifact
1597 existence. This will involve additional overhead due to the need
1598 to query an external system. If `False` registry and datastore
1599 will solely be asked if they know about the dataset but no
1600 check for the artifact will be performed.
1601 collections : Any, optional
1602 Collections to be searched, overriding ``self.collections``.
1603 Can be any of the types supported by the ``collections`` argument
1604 to butler construction.
1605 **kwargs
1606 Additional keyword arguments used to augment or construct a
1607 `DataCoordinate`. See `DataCoordinate.standardize`
1608 parameters.
1610 Returns
1611 -------
1612 existence : `DatasetExistence`
1613 Object indicating whether the dataset is known to registry and
1614 datastore. Evaluates to `True` if the dataset is present and known
1615 to both.
1616 """
1617 existence = DatasetExistence.UNRECOGNIZED
1619 if isinstance(dataset_ref_or_type, DatasetRef):
1620 if collections is not None:
1621 warnings.warn("Collections should not be specified with DatasetRef", stacklevel=2)
1622 if data_id is not None:
1623 warnings.warn("A DataID should not be specified with DatasetRef", stacklevel=2)
1624 ref = dataset_ref_or_type
1625 registry_ref = self.registry.getDataset(dataset_ref_or_type.id)
1626 if registry_ref is not None:
1627 existence |= DatasetExistence.RECORDED
1629 if dataset_ref_or_type != registry_ref:
1630 # This could mean that storage classes differ, so we should
1631 # check for that but use the registry ref for the rest of
1632 # the method.
1633 if registry_ref.is_compatible_with(dataset_ref_or_type):
1634 # Use the registry version from now on.
1635 ref = registry_ref
1636 else:
1637 raise ValueError(
1638 f"The ref given to exists() ({ref}) has the same dataset ID as one "
1639 f"in registry but has different incompatible values ({registry_ref})."
1640 )
1641 else:
1642 try:
1643 ref = self._findDatasetRef(dataset_ref_or_type, data_id, collections=collections, **kwargs)
1644 except (LookupError, TypeError, NoDefaultCollectionError):
1645 return existence
1646 existence |= DatasetExistence.RECORDED
1648 if self._datastore.knows(ref):
1649 existence |= DatasetExistence.DATASTORE
1651 if full_check:
1652 if self._datastore.exists(ref):
1653 existence |= DatasetExistence._ARTIFACT
1654 elif existence.value != DatasetExistence.UNRECOGNIZED.value:
1655 # Do not add this flag if we have no other idea about a dataset.
1656 existence |= DatasetExistence(DatasetExistence._ASSUMED)
1658 return existence
1660 def _exists_many(
1661 self,
1662 refs: Iterable[DatasetRef],
1663 /,
1664 *,
1665 full_check: bool = True,
1666 ) -> dict[DatasetRef, DatasetExistence]:
1667 """Indicate whether multiple datasets are known to Butler registry and
1668 datastore.
1670 This is an experimental API that may change at any moment.
1672 Parameters
1673 ----------
1674 refs : iterable of `DatasetRef`
1675 The datasets to be checked.
1676 full_check : `bool`, optional
1677 If `True`, an additional check will be made for dataset artifact
1678 existence. This will involve additional overhead due to the need
1679 to query an external system. If `False` registry and datastore
1680 will solely be asked if they know about the dataset but no
1681 check for the artifact will be performed.
1683 Returns
1684 -------
1685 existence : dict of [`DatasetRef`, `DatasetExistence`]
1686 Mapping from the given dataset refs to an enum indicating the
1687 status of the dataset in registry and datastore.
1688 Each value evaluates to `True` if the dataset is present and known
1689 to both.
1690 """
1691 existence = {ref: DatasetExistence.UNRECOGNIZED for ref in refs}
1693 # Registry does not have a bulk API to check for a ref.
1694 for ref in refs:
1695 registry_ref = self.registry.getDataset(ref.id)
1696 if registry_ref is not None:
1697 # It is possible, albeit unlikely, that the given ref does
1698 # not match the one in registry even though the UUID matches.
1699 # When checking a single ref we raise, but it's impolite to
1700 # do that when potentially hundreds of refs are being checked.
1701 # We could change the API to only accept UUIDs and that would
1702 # remove the ability to even check and remove the worry
1703 # about differing storage classes. Given the ongoing discussion
1704 # on refs vs UUIDs and whether to raise or have a new
1705 # private flag, treat this as a private API for now.
1706 existence[ref] |= DatasetExistence.RECORDED
1708 # Ask datastore if it knows about these refs.
1709 knows = self._datastore.knows_these(refs)
1710 for ref, known in knows.items():
1711 if known:
1712 existence[ref] |= DatasetExistence.DATASTORE
1714 if full_check:
1715 mexists = self._datastore.mexists(refs)
1716 for ref, exists in mexists.items():
1717 if exists:
1718 existence[ref] |= DatasetExistence._ARTIFACT
1719 else:
1720 # Do not set this flag if nothing is known about the dataset.
1721 for ref in existence.keys():
1722 if existence[ref] != DatasetExistence.UNRECOGNIZED:
1723 existence[ref] |= DatasetExistence._ASSUMED
1725 return existence
1727 @deprecated(
1728 reason="Butler.datasetExists() has been replaced by Butler.exists(). Will be removed after v27.0.",
1729 version="v26.0",
1730 category=FutureWarning,
1731 )
1732 def datasetExists(
1733 self,
1734 datasetRefOrType: DatasetRef | DatasetType | str,
1735 dataId: DataId | None = None,
1736 *,
1737 collections: Any = None,
1738 **kwargs: Any,
1739 ) -> bool:
1740 """Return True if the Dataset is actually present in the Datastore.
1742 Parameters
1743 ----------
1744 datasetRefOrType : `DatasetRef`, `DatasetType`, or `str`
1745 When `DatasetRef` the `dataId` should be `None`.
1746 Otherwise the `DatasetType` or name thereof.
1747 dataId : `dict` or `DataCoordinate`
1748 A `dict` of `Dimension` link name, value pairs that label the
1749 `DatasetRef` within a Collection. When `None`, a `DatasetRef`
1750 should be provided as the first argument.
1751 collections : Any, optional
1752 Collections to be searched, overriding ``self.collections``.
1753 Can be any of the types supported by the ``collections`` argument
1754 to butler construction.
1755 **kwargs
1756 Additional keyword arguments used to augment or construct a
1757 `DataCoordinate`. See `DataCoordinate.standardize`
1758 parameters.
1760 Raises
1761 ------
1762 LookupError
1763 Raised if the dataset is not even present in the Registry.
1764 ValueError
1765 Raised if a resolved `DatasetRef` was passed as an input, but it
1766 differs from the one found in the registry.
1767 NoDefaultCollectionError
1768 Raised if no collections were provided.
1769 """
1770 # A resolved ref may be given that is not known to this butler.
1771 if isinstance(datasetRefOrType, DatasetRef):
1772 ref = self.registry.getDataset(datasetRefOrType.id)
1773 if ref is None:
1774 raise LookupError(
1775 f"Resolved DatasetRef with id {datasetRefOrType.id} is not known to registry."
1776 )
1777 else:
1778 ref = self._findDatasetRef(datasetRefOrType, dataId, collections=collections, **kwargs)
1779 return self._datastore.exists(ref)
1781 def removeRuns(self, names: Iterable[str], unstore: bool = True) -> None:
1782 """Remove one or more `~CollectionType.RUN` collections and the
1783 datasets within them.
1785 Parameters
1786 ----------
1787 names : `~collections.abc.Iterable` [ `str` ]
1788 The names of the collections to remove.
1789 unstore : `bool`, optional
1790 If `True` (default), delete datasets from all datastores in which
1791 they are present, and attempt to rollback the registry deletions if
1792 datastore deletions fail (which may not always be possible). If
1793 `False`, datastore records for these datasets are still removed,
1794 but any artifacts (e.g. files) will not be.
1796 Raises
1797 ------
1798 TypeError
1799 Raised if one or more collections are not of type
1800 `~CollectionType.RUN`.
1801 """
1802 if not self.isWriteable():
1803 raise TypeError("Butler is read-only.")
1804 names = list(names)
1805 refs: list[DatasetRef] = []
1806 for name in names:
1807 collectionType = self.registry.getCollectionType(name)
1808 if collectionType is not CollectionType.RUN:
1809 raise TypeError(f"The collection type of '{name}' is {collectionType.name}, not RUN.")
1810 refs.extend(self.registry.queryDatasets(..., collections=name, findFirst=True))
1811 with self._datastore.transaction():
1812 with self.registry.transaction():
1813 if unstore:
1814 self._datastore.trash(refs)
1815 else:
1816 self._datastore.forget(refs)
1817 for name in names:
1818 self.registry.removeCollection(name)
1819 if unstore:
1820 # Point of no return for removing artifacts
1821 self._datastore.emptyTrash()
1823 def pruneDatasets(
1824 self,
1825 refs: Iterable[DatasetRef],
1826 *,
1827 disassociate: bool = True,
1828 unstore: bool = False,
1829 tags: Iterable[str] = (),
1830 purge: bool = False,
1831 ) -> None:
1832 # docstring inherited from LimitedButler
1834 if not self.isWriteable():
1835 raise TypeError("Butler is read-only.")
1836 if purge:
1837 if not disassociate:
1838 raise TypeError("Cannot pass purge=True without disassociate=True.")
1839 if not unstore:
1840 raise TypeError("Cannot pass purge=True without unstore=True.")
1841 elif disassociate:
1842 tags = tuple(tags)
1843 if not tags:
1844 raise TypeError("No tags provided but disassociate=True.")
1845 for tag in tags:
1846 collectionType = self.registry.getCollectionType(tag)
1847 if collectionType is not CollectionType.TAGGED:
1848 raise TypeError(
1849 f"Cannot disassociate from collection '{tag}' "
1850 f"of non-TAGGED type {collectionType.name}."
1851 )
1852 # Transform possibly-single-pass iterable into something we can iterate
1853 # over multiple times.
1854 refs = list(refs)
1855 # Pruning a component of a DatasetRef makes no sense since registry
1856 # doesn't know about components and datastore might not store
1857 # components in a separate file
1858 for ref in refs:
1859 if ref.datasetType.component():
1860 raise ValueError(f"Can not prune a component of a dataset (ref={ref})")
1861 # We don't need an unreliable Datastore transaction for this, because
1862 # we've been extra careful to ensure that Datastore.trash only involves
1863 # mutating the Registry (it can _look_ at Datastore-specific things,
1864 # but shouldn't change them), and hence all operations here are
1865 # Registry operations.
1866 with self._datastore.transaction():
1867 with self.registry.transaction():
1868 if unstore:
1869 self._datastore.trash(refs)
1870 if purge:
1871 self.registry.removeDatasets(refs)
1872 elif disassociate:
1873 assert tags, "Guaranteed by earlier logic in this function."
1874 for tag in tags:
1875 self.registry.disassociate(tag, refs)
1876 # We've exited the Registry transaction, and apparently committed.
1877 # (if there was an exception, everything rolled back, and it's as if
1878 # nothing happened - and we never get here).
1879 # Datastore artifacts are not yet gone, but they're clearly marked
1880 # as trash, so if we fail to delete now because of (e.g.) filesystem
1881 # problems we can try again later, and if manual administrative
1882 # intervention is required, it's pretty clear what that should entail:
1883 # deleting everything on disk and in private Datastore tables that is
1884 # in the dataset_location_trash table.
1885 if unstore:
1886 # Point of no return for removing artifacts
1887 self._datastore.emptyTrash()
1889 @transactional
1890 def ingest(
1891 self,
1892 *datasets: FileDataset,
1893 transfer: str | None = "auto",
1894 run: str | None = None,
1895 idGenerationMode: DatasetIdGenEnum = DatasetIdGenEnum.UNIQUE,
1896 record_validation_info: bool = True,
1897 ) -> None:
1898 """Store and register one or more datasets that already exist on disk.
1900 Parameters
1901 ----------
1902 datasets : `FileDataset`
1903 Each positional argument is a struct containing information about
1904 a file to be ingested, including its URI (either absolute or
1905 relative to the datastore root, if applicable), a resolved
1906 `DatasetRef`, and optionally a formatter class or its
1907 fully-qualified string name. If a formatter is not provided, the
1908 formatter that would be used for `put` is assumed. On successful
1909 ingest all `FileDataset.formatter` attributes will be set to the
1910 formatter class used. `FileDataset.path` attributes may be modified
1911 to put paths in whatever the datastore considers a standardized
1912 form.
1913 transfer : `str`, optional
1914 If not `None`, must be one of 'auto', 'move', 'copy', 'direct',
1915 'split', 'hardlink', 'relsymlink' or 'symlink', indicating how to
1916 transfer the file.
1917 run : `str`, optional
1918 The name of the run ingested datasets should be added to,
1919 overriding ``self.run``. This parameter is now deprecated since
1920 the run is encoded in the ``FileDataset``.
1921 idGenerationMode : `DatasetIdGenEnum`, optional
1922 Specifies option for generating dataset IDs. By default unique IDs
1923 are generated for each inserted dataset.
1924 record_validation_info : `bool`, optional
1925 If `True`, the default, the datastore can record validation
1926 information associated with the file. If `False` the datastore
1927 will not attempt to track any information such as checksums
1928 or file sizes. This can be useful if such information is tracked
1929 in an external system or if the file is to be compressed in place.
1930 It is up to the datastore whether this parameter is relevant.
1932 Raises
1933 ------
1934 TypeError
1935 Raised if the butler is read-only or if no run was provided.
1936 NotImplementedError
1937 Raised if the `Datastore` does not support the given transfer mode.
1938 DatasetTypeNotSupportedError
1939 Raised if one or more files to be ingested have a dataset type that
1940 is not supported by the `Datastore`..
1941 FileNotFoundError
1942 Raised if one of the given files does not exist.
1943 FileExistsError
1944 Raised if transfer is not `None` but the (internal) location the
1945 file would be moved to is already occupied.
1947 Notes
1948 -----
1949 This operation is not fully exception safe: if a database operation
1950 fails, the given `FileDataset` instances may be only partially updated.
1952 It is atomic in terms of database operations (they will either all
1953 succeed or all fail) providing the database engine implements
1954 transactions correctly. It will attempt to be atomic in terms of
1955 filesystem operations as well, but this cannot be implemented
1956 rigorously for most datastores.
1957 """
1958 if not self.isWriteable():
1959 raise TypeError("Butler is read-only.")
1961 log.verbose("Ingesting %d file dataset%s.", len(datasets), "" if len(datasets) == 1 else "s")
1962 if not datasets:
1963 return
1965 progress = Progress("lsst.daf.butler.Butler.ingest", level=logging.DEBUG)
1967 # We need to reorganize all the inputs so that they are grouped
1968 # by dataset type and run. Multiple refs in a single FileDataset
1969 # are required to share the run and dataset type.
1970 GroupedData = MutableMapping[tuple[DatasetType, str], list[FileDataset]]
1971 groupedData: GroupedData = defaultdict(list)
1973 # Track DataIDs that are being ingested so we can spot issues early
1974 # with duplication. Retain previous FileDataset so we can report it.
1975 groupedDataIds: MutableMapping[
1976 tuple[DatasetType, str], dict[DataCoordinate, FileDataset]
1977 ] = defaultdict(dict)
1979 used_run = False
1981 # And the nested loop that populates it:
1982 for dataset in progress.wrap(datasets, desc="Grouping by dataset type"):
1983 # Somewhere to store pre-existing refs if we have an
1984 # execution butler.
1985 existingRefs: list[DatasetRef] = []
1987 for ref in dataset.refs:
1988 assert ref.run is not None # For mypy
1989 group_key = (ref.datasetType, ref.run)
1991 if ref.dataId in groupedDataIds[group_key]:
1992 raise ConflictingDefinitionError(
1993 f"Ingest conflict. Dataset {dataset.path} has same"
1994 " DataId as other ingest dataset"
1995 f" {groupedDataIds[group_key][ref.dataId].path} "
1996 f" ({ref.dataId})"
1997 )
1999 groupedDataIds[group_key][ref.dataId] = dataset
2001 if existingRefs:
2002 if len(dataset.refs) != len(existingRefs):
2003 # Keeping track of partially pre-existing datasets is hard
2004 # and should generally never happen. For now don't allow
2005 # it.
2006 raise ConflictingDefinitionError(
2007 f"For dataset {dataset.path} some dataIds already exist"
2008 " in registry but others do not. This is not supported."
2009 )
2011 # Store expanded form in the original FileDataset.
2012 dataset.refs = existingRefs
2013 else:
2014 groupedData[group_key].append(dataset)
2016 if not used_run and run is not None:
2017 warnings.warn(
2018 "All DatasetRefs to be ingested had resolved dataset IDs. The value given to the "
2019 f"'run' parameter ({run!r}) was not used and the parameter will be removed in the future.",
2020 category=FutureWarning,
2021 stacklevel=3, # Take into account the @transactional decorator.
2022 )
2024 # Now we can bulk-insert into Registry for each DatasetType.
2025 for (datasetType, this_run), grouped_datasets in progress.iter_item_chunks(
2026 groupedData.items(), desc="Bulk-inserting datasets by type"
2027 ):
2028 refs_to_import = []
2029 for dataset in grouped_datasets:
2030 refs_to_import.extend(dataset.refs)
2032 n_refs = len(refs_to_import)
2033 log.verbose(
2034 "Importing %d ref%s of dataset type %r into run %r",
2035 n_refs,
2036 "" if n_refs == 1 else "s",
2037 datasetType.name,
2038 this_run,
2039 )
2041 # Import the refs and expand the DataCoordinates since we can't
2042 # guarantee that they are expanded and Datastore will need
2043 # the records.
2044 imported_refs = self.registry._importDatasets(refs_to_import, expand=True)
2045 assert set(imported_refs) == set(refs_to_import)
2047 # Replace all the refs in the FileDataset with expanded versions.
2048 # Pull them off in the order we put them on the list.
2049 for dataset in grouped_datasets:
2050 n_dataset_refs = len(dataset.refs)
2051 dataset.refs = imported_refs[:n_dataset_refs]
2052 del imported_refs[:n_dataset_refs]
2054 # Bulk-insert everything into Datastore.
2055 # We do not know if any of the registry entries already existed
2056 # (_importDatasets only complains if they exist but differ) so
2057 # we have to catch IntegrityError explicitly.
2058 try:
2059 self._datastore.ingest(
2060 *datasets, transfer=transfer, record_validation_info=record_validation_info
2061 )
2062 except IntegrityError as e:
2063 raise ConflictingDefinitionError(f"Datastore already contains one or more datasets: {e}")
2065 @contextlib.contextmanager
2066 def export(
2067 self,
2068 *,
2069 directory: str | None = None,
2070 filename: str | None = None,
2071 format: str | None = None,
2072 transfer: str | None = None,
2073 ) -> Iterator[RepoExportContext]:
2074 """Export datasets from the repository represented by this `Butler`.
2076 This method is a context manager that returns a helper object
2077 (`RepoExportContext`) that is used to indicate what information from
2078 the repository should be exported.
2080 Parameters
2081 ----------
2082 directory : `str`, optional
2083 Directory dataset files should be written to if ``transfer`` is not
2084 `None`.
2085 filename : `str`, optional
2086 Name for the file that will include database information associated
2087 with the exported datasets. If this is not an absolute path and
2088 ``directory`` is not `None`, it will be written to ``directory``
2089 instead of the current working directory. Defaults to
2090 "export.{format}".
2091 format : `str`, optional
2092 File format for the database information file. If `None`, the
2093 extension of ``filename`` will be used.
2094 transfer : `str`, optional
2095 Transfer mode passed to `Datastore.export`.
2097 Raises
2098 ------
2099 TypeError
2100 Raised if the set of arguments passed is inconsistent.
2102 Examples
2103 --------
2104 Typically the `Registry.queryDataIds` and `Registry.queryDatasets`
2105 methods are used to provide the iterables over data IDs and/or datasets
2106 to be exported::
2108 with butler.export("exports.yaml") as export:
2109 # Export all flats, but none of the dimension element rows
2110 # (i.e. data ID information) associated with them.
2111 export.saveDatasets(butler.registry.queryDatasets("flat"),
2112 elements=())
2113 # Export all datasets that start with "deepCoadd_" and all of
2114 # their associated data ID information.
2115 export.saveDatasets(butler.registry.queryDatasets("deepCoadd_*"))
2116 """
2117 if directory is None and transfer is not None:
2118 raise TypeError("Cannot transfer without providing a directory.")
2119 if transfer == "move":
2120 raise TypeError("Transfer may not be 'move': export is read-only")
2121 if format is None:
2122 if filename is None:
2123 raise TypeError("At least one of 'filename' or 'format' must be provided.")
2124 else:
2125 _, format = os.path.splitext(filename)
2126 if not format:
2127 raise ValueError("Please specify a file extension to determine export format.")
2128 format = format[1:] # Strip leading ".""
2129 elif filename is None:
2130 filename = f"export.{format}"
2131 if directory is not None:
2132 filename = os.path.join(directory, filename)
2133 formats = self._config["repo_transfer_formats"]
2134 if format not in formats:
2135 raise ValueError(f"Unknown export format {format!r}, allowed: {','.join(formats.keys())}")
2136 BackendClass = get_class_of(formats[format, "export"])
2137 with open(filename, "w") as stream:
2138 backend = BackendClass(stream, universe=self.dimensions)
2139 try:
2140 helper = RepoExportContext(
2141 self.registry, self._datastore, backend=backend, directory=directory, transfer=transfer
2142 )
2143 yield helper
2144 except BaseException:
2145 raise
2146 else:
2147 helper._finish()
2149 def import_(
2150 self,
2151 *,
2152 directory: ResourcePathExpression | None = None,
2153 filename: ResourcePathExpression | TextIO | None = None,
2154 format: str | None = None,
2155 transfer: str | None = None,
2156 skip_dimensions: set | None = None,
2157 ) -> None:
2158 """Import datasets into this repository that were exported from a
2159 different butler repository via `~lsst.daf.butler.Butler.export`.
2161 Parameters
2162 ----------
2163 directory : `~lsst.resources.ResourcePathExpression`, optional
2164 Directory containing dataset files to import from. If `None`,
2165 ``filename`` and all dataset file paths specified therein must
2166 be absolute.
2167 filename : `~lsst.resources.ResourcePathExpression` or `TextIO`
2168 A stream or name of file that contains database information
2169 associated with the exported datasets, typically generated by
2170 `~lsst.daf.butler.Butler.export`. If this a string (name) or
2171 `~lsst.resources.ResourcePath` and is not an absolute path,
2172 it will first be looked for relative to ``directory`` and if not
2173 found there it will be looked for in the current working
2174 directory. Defaults to "export.{format}".
2175 format : `str`, optional
2176 File format for ``filename``. If `None`, the extension of
2177 ``filename`` will be used.
2178 transfer : `str`, optional
2179 Transfer mode passed to `~lsst.daf.butler.Datastore.ingest`.
2180 skip_dimensions : `set`, optional
2181 Names of dimensions that should be skipped and not imported.
2183 Raises
2184 ------
2185 TypeError
2186 Raised if the set of arguments passed is inconsistent, or if the
2187 butler is read-only.
2188 """
2189 if not self.isWriteable():
2190 raise TypeError("Butler is read-only.")
2191 if format is None:
2192 if filename is None:
2193 raise TypeError("At least one of 'filename' or 'format' must be provided.")
2194 else:
2195 _, format = os.path.splitext(filename) # type: ignore
2196 elif filename is None:
2197 filename = ResourcePath(f"export.{format}", forceAbsolute=False)
2198 if directory is not None:
2199 directory = ResourcePath(directory, forceDirectory=True)
2200 # mypy doesn't think this will work but it does in python >= 3.10.
2201 if isinstance(filename, ResourcePathExpression): # type: ignore
2202 filename = ResourcePath(filename, forceAbsolute=False) # type: ignore
2203 if not filename.isabs() and directory is not None:
2204 potential = directory.join(filename)
2205 exists_in_cwd = filename.exists()
2206 exists_in_dir = potential.exists()
2207 if exists_in_cwd and exists_in_dir:
2208 log.warning(
2209 "A relative path for filename was specified (%s) which exists relative to cwd. "
2210 "Additionally, the file exists relative to the given search directory (%s). "
2211 "Using the export file in the given directory.",
2212 filename,
2213 potential,
2214 )
2215 # Given they specified an explicit directory and that
2216 # directory has the export file in it, assume that that
2217 # is what was meant despite the file in cwd.
2218 filename = potential
2219 elif exists_in_dir:
2220 filename = potential
2221 elif not exists_in_cwd and not exists_in_dir:
2222 # Raise early.
2223 raise FileNotFoundError(
2224 f"Export file could not be found in {filename.abspath()} or {potential.abspath()}."
2225 )
2226 BackendClass: type[RepoImportBackend] = get_class_of(
2227 self._config["repo_transfer_formats"][format]["import"]
2228 )
2230 def doImport(importStream: TextIO | ResourceHandleProtocol) -> None:
2231 backend = BackendClass(importStream, self.registry) # type: ignore[call-arg]
2232 backend.register()
2233 with self.transaction():
2234 backend.load(
2235 self._datastore,
2236 directory=directory,
2237 transfer=transfer,
2238 skip_dimensions=skip_dimensions,
2239 )
2241 if isinstance(filename, ResourcePath):
2242 # We can not use open() here at the moment because of
2243 # DM-38589 since yaml does stream.read(8192) in a loop.
2244 stream = io.StringIO(filename.read().decode())
2245 doImport(stream)
2246 else:
2247 doImport(filename) # type: ignore
2249 def transfer_from(
2250 self,
2251 source_butler: LimitedButler,
2252 source_refs: Iterable[DatasetRef],
2253 transfer: str = "auto",
2254 skip_missing: bool = True,
2255 register_dataset_types: bool = False,
2256 transfer_dimensions: bool = False,
2257 ) -> collections.abc.Collection[DatasetRef]:
2258 """Transfer datasets to this Butler from a run in another Butler.
2260 Parameters
2261 ----------
2262 source_butler : `LimitedButler`
2263 Butler from which the datasets are to be transferred. If data IDs
2264 in ``source_refs`` are not expanded then this has to be a full
2265 `Butler` whose registry will be used to expand data IDs.
2266 source_refs : iterable of `DatasetRef`
2267 Datasets defined in the source butler that should be transferred to
2268 this butler.
2269 transfer : `str`, optional
2270 Transfer mode passed to `~lsst.daf.butler.Datastore.transfer_from`.
2271 skip_missing : `bool`
2272 If `True`, datasets with no datastore artifact associated with
2273 them are not transferred. If `False` a registry entry will be
2274 created even if no datastore record is created (and so will
2275 look equivalent to the dataset being unstored).
2276 register_dataset_types : `bool`
2277 If `True` any missing dataset types are registered. Otherwise
2278 an exception is raised.
2279 transfer_dimensions : `bool`, optional
2280 If `True`, dimension record data associated with the new datasets
2281 will be transferred.
2283 Returns
2284 -------
2285 refs : `list` of `DatasetRef`
2286 The refs added to this Butler.
2288 Notes
2289 -----
2290 The datastore artifact has to exist for a transfer
2291 to be made but non-existence is not an error.
2293 Datasets that already exist in this run will be skipped.
2295 The datasets are imported as part of a transaction, although
2296 dataset types are registered before the transaction is started.
2297 This means that it is possible for a dataset type to be registered
2298 even though transfer has failed.
2299 """
2300 if not self.isWriteable():
2301 raise TypeError("Butler is read-only.")
2302 progress = Progress("lsst.daf.butler.Butler.transfer_from", level=VERBOSE)
2304 # Will iterate through the refs multiple times so need to convert
2305 # to a list if this isn't a collection.
2306 if not isinstance(source_refs, collections.abc.Collection):
2307 source_refs = list(source_refs)
2309 original_count = len(source_refs)
2310 log.info("Transferring %d datasets into %s", original_count, str(self))
2312 # In some situations the datastore artifact may be missing
2313 # and we do not want that registry entry to be imported.
2314 # Asking datastore is not sufficient, the records may have been
2315 # purged, we have to ask for the (predicted) URI and check
2316 # existence explicitly. Execution butler is set up exactly like
2317 # this with no datastore records.
2318 artifact_existence: dict[ResourcePath, bool] = {}
2319 if skip_missing:
2320 dataset_existence = source_butler._datastore.mexists(
2321 source_refs, artifact_existence=artifact_existence
2322 )
2323 source_refs = [ref for ref, exists in dataset_existence.items() if exists]
2324 filtered_count = len(source_refs)
2325 n_missing = original_count - filtered_count
2326 log.verbose(
2327 "%d dataset%s removed because the artifact does not exist. Now have %d.",
2328 n_missing,
2329 "" if n_missing == 1 else "s",
2330 filtered_count,
2331 )
2333 # Importing requires that we group the refs by dataset type and run
2334 # before doing the import.
2335 source_dataset_types = set()
2336 grouped_refs = defaultdict(list)
2337 for ref in source_refs:
2338 grouped_refs[ref.datasetType, ref.run].append(ref)
2339 source_dataset_types.add(ref.datasetType)
2341 # Check to see if the dataset type in the source butler has
2342 # the same definition in the target butler and register missing
2343 # ones if requested. Registration must happen outside a transaction.
2344 newly_registered_dataset_types = set()
2345 for datasetType in source_dataset_types:
2346 if register_dataset_types:
2347 # Let this raise immediately if inconsistent. Continuing
2348 # on to find additional inconsistent dataset types
2349 # might result in additional unwanted dataset types being
2350 # registered.
2351 if self.registry.registerDatasetType(datasetType):
2352 newly_registered_dataset_types.add(datasetType)
2353 else:
2354 # If the dataset type is missing, let it fail immediately.
2355 target_dataset_type = self.registry.getDatasetType(datasetType.name)
2356 if target_dataset_type != datasetType:
2357 raise ConflictingDefinitionError(
2358 "Source butler dataset type differs from definition"
2359 f" in target butler: {datasetType} !="
2360 f" {target_dataset_type}"
2361 )
2362 if newly_registered_dataset_types:
2363 # We may have registered some even if there were inconsistencies
2364 # but should let people know (or else remove them again).
2365 log.log(
2366 VERBOSE,
2367 "Registered the following dataset types in the target Butler: %s",
2368 ", ".join(d.name for d in newly_registered_dataset_types),
2369 )
2370 else:
2371 log.log(VERBOSE, "All required dataset types are known to the target Butler")
2373 dimension_records: dict[DimensionElement, dict[DataCoordinate, DimensionRecord]] = defaultdict(dict)
2374 if transfer_dimensions:
2375 # Collect all the dimension records for these refs.
2376 # All dimensions are to be copied but the list of valid dimensions
2377 # come from this butler's universe.
2378 elements = frozenset(
2379 element
2380 for element in self.dimensions.getStaticElements()
2381 if element.hasTable() and element.viewOf is None
2382 )
2383 dataIds = {ref.dataId for ref in source_refs}
2384 # This logic comes from saveDataIds.
2385 for dataId in dataIds:
2386 # Need an expanded record, if not expanded that we need a full
2387 # butler with registry (allow mocks with registry too).
2388 if not dataId.hasRecords():
2389 if registry := getattr(source_butler, "registry", None):
2390 dataId = registry.expandDataId(dataId)
2391 else:
2392 raise TypeError("Input butler needs to be a full butler to expand DataId.")
2393 # If this butler doesn't know about a dimension in the source
2394 # butler things will break later.
2395 for record in dataId.records.values():
2396 if record is not None and record.definition in elements:
2397 dimension_records[record.definition].setdefault(record.dataId, record)
2399 handled_collections: set[str] = set()
2401 # Do all the importing in a single transaction.
2402 with self.transaction():
2403 if dimension_records:
2404 log.verbose("Ensuring that dimension records exist for transferred datasets.")
2405 for element, r in dimension_records.items():
2406 records = [r[dataId] for dataId in r]
2407 # Assume that if the record is already present that we can
2408 # use it without having to check that the record metadata
2409 # is consistent.
2410 self.registry.insertDimensionData(element, *records, skip_existing=True)
2412 n_imported = 0
2413 for (datasetType, run), refs_to_import in progress.iter_item_chunks(
2414 grouped_refs.items(), desc="Importing to registry by run and dataset type"
2415 ):
2416 if run not in handled_collections:
2417 # May need to create output collection. If source butler
2418 # has a registry, ask for documentation string.
2419 run_doc = None
2420 if registry := getattr(source_butler, "registry", None):
2421 run_doc = registry.getCollectionDocumentation(run)
2422 registered = self.registry.registerRun(run, doc=run_doc)
2423 handled_collections.add(run)
2424 if registered:
2425 log.log(VERBOSE, "Creating output run %s", run)
2427 n_refs = len(refs_to_import)
2428 log.verbose(
2429 "Importing %d ref%s of dataset type %s into run %s",
2430 n_refs,
2431 "" if n_refs == 1 else "s",
2432 datasetType.name,
2433 run,
2434 )
2436 # Assume we are using UUIDs and the source refs will match
2437 # those imported.
2438 imported_refs = self.registry._importDatasets(refs_to_import, expand=False)
2439 assert set(imported_refs) == set(refs_to_import)
2440 n_imported += len(imported_refs)
2442 assert len(source_refs) == n_imported
2443 log.verbose("Imported %d datasets into destination butler", n_imported)
2445 # Ask the datastore to transfer. The datastore has to check that
2446 # the source datastore is compatible with the target datastore.
2447 accepted, rejected = self._datastore.transfer_from(
2448 source_butler._datastore,
2449 source_refs,
2450 transfer=transfer,
2451 artifact_existence=artifact_existence,
2452 )
2453 if rejected:
2454 # For now, accept the registry entries but not the files.
2455 log.warning(
2456 "%d datasets were rejected and %d accepted for dataset type %s in run %r.",
2457 len(rejected),
2458 len(accepted),
2459 datasetType,
2460 run,
2461 )
2463 return source_refs
2465 def validateConfiguration(
2466 self,
2467 logFailures: bool = False,
2468 datasetTypeNames: Iterable[str] | None = None,
2469 ignore: Iterable[str] | None = None,
2470 ) -> None:
2471 """Validate butler configuration.
2473 Checks that each `DatasetType` can be stored in the `Datastore`.
2475 Parameters
2476 ----------
2477 logFailures : `bool`, optional
2478 If `True`, output a log message for every validation error
2479 detected.
2480 datasetTypeNames : iterable of `str`, optional
2481 The `DatasetType` names that should be checked. This allows
2482 only a subset to be selected.
2483 ignore : iterable of `str`, optional
2484 Names of DatasetTypes to skip over. This can be used to skip
2485 known problems. If a named `DatasetType` corresponds to a
2486 composite, all components of that `DatasetType` will also be
2487 ignored.
2489 Raises
2490 ------
2491 ButlerValidationError
2492 Raised if there is some inconsistency with how this Butler
2493 is configured.
2494 """
2495 if datasetTypeNames:
2496 datasetTypes = [self.registry.getDatasetType(name) for name in datasetTypeNames]
2497 else:
2498 datasetTypes = list(self.registry.queryDatasetTypes())
2500 # filter out anything from the ignore list
2501 if ignore:
2502 ignore = set(ignore)
2503 datasetTypes = [
2504 e for e in datasetTypes if e.name not in ignore and e.nameAndComponent()[0] not in ignore
2505 ]
2506 else:
2507 ignore = set()
2509 # For each datasetType that has an instrument dimension, create
2510 # a DatasetRef for each defined instrument
2511 datasetRefs = []
2513 # Find all the registered instruments (if "instrument" is in the
2514 # universe).
2515 if "instrument" in self.dimensions:
2516 instruments = {record.name for record in self.registry.queryDimensionRecords("instrument")}
2518 for datasetType in datasetTypes:
2519 if "instrument" in datasetType.dimensions:
2520 # In order to create a conforming dataset ref, create
2521 # fake DataCoordinate values for the non-instrument
2522 # dimensions. The type of the value does not matter here.
2523 dataId = {dim.name: 1 for dim in datasetType.dimensions if dim.name != "instrument"}
2525 for instrument in instruments:
2526 datasetRef = DatasetRef(
2527 datasetType,
2528 DataCoordinate.standardize(
2529 dataId, instrument=instrument, graph=datasetType.dimensions
2530 ),
2531 run="validate",
2532 )
2533 datasetRefs.append(datasetRef)
2535 entities: list[DatasetType | DatasetRef] = []
2536 entities.extend(datasetTypes)
2537 entities.extend(datasetRefs)
2539 datastoreErrorStr = None
2540 try:
2541 self._datastore.validateConfiguration(entities, logFailures=logFailures)
2542 except ValidationError as e:
2543 datastoreErrorStr = str(e)
2545 # Also check that the LookupKeys used by the datastores match
2546 # registry and storage class definitions
2547 keys = self._datastore.getLookupKeys()
2549 failedNames = set()
2550 failedDataId = set()
2551 for key in keys:
2552 if key.name is not None:
2553 if key.name in ignore:
2554 continue
2556 # skip if specific datasetType names were requested and this
2557 # name does not match
2558 if datasetTypeNames and key.name not in datasetTypeNames:
2559 continue
2561 # See if it is a StorageClass or a DatasetType
2562 if key.name in self.storageClasses:
2563 pass
2564 else:
2565 try:
2566 self.registry.getDatasetType(key.name)
2567 except KeyError:
2568 if logFailures:
2569 log.critical("Key '%s' does not correspond to a DatasetType or StorageClass", key)
2570 failedNames.add(key)
2571 else:
2572 # Dimensions are checked for consistency when the Butler
2573 # is created and rendezvoused with a universe.
2574 pass
2576 # Check that the instrument is a valid instrument
2577 # Currently only support instrument so check for that
2578 if key.dataId:
2579 dataIdKeys = set(key.dataId)
2580 if {"instrument"} != dataIdKeys:
2581 if logFailures:
2582 log.critical("Key '%s' has unsupported DataId override", key)
2583 failedDataId.add(key)
2584 elif key.dataId["instrument"] not in instruments:
2585 if logFailures:
2586 log.critical("Key '%s' has unknown instrument", key)
2587 failedDataId.add(key)
2589 messages = []
2591 if datastoreErrorStr:
2592 messages.append(datastoreErrorStr)
2594 for failed, msg in (
2595 (failedNames, "Keys without corresponding DatasetType or StorageClass entry: "),
2596 (failedDataId, "Keys with bad DataId entries: "),
2597 ):
2598 if failed:
2599 msg += ", ".join(str(k) for k in failed)
2600 messages.append(msg)
2602 if messages:
2603 raise ValidationError(";\n".join(messages))
2605 @property
2606 def collections(self) -> Sequence[str]:
2607 """The collections to search by default, in order
2608 (`~collections.abc.Sequence` [ `str` ]).
2610 This is an alias for ``self.registry.defaults.collections``. It cannot
2611 be set directly in isolation, but all defaults may be changed together
2612 by assigning a new `RegistryDefaults` instance to
2613 ``self.registry.defaults``.
2614 """
2615 return self.registry.defaults.collections
2617 @property
2618 def run(self) -> str | None:
2619 """Name of the run this butler writes outputs to by default (`str` or
2620 `None`).
2622 This is an alias for ``self.registry.defaults.run``. It cannot be set
2623 directly in isolation, but all defaults may be changed together by
2624 assigning a new `RegistryDefaults` instance to
2625 ``self.registry.defaults``.
2626 """
2627 return self.registry.defaults.run
2629 @property
2630 def dimensions(self) -> DimensionUniverse:
2631 # Docstring inherited.
2632 return self.registry.dimensions
2634 registry: Registry
2635 """The object that manages dataset metadata and relationships (`Registry`).
2637 Most operations that don't involve reading or writing butler datasets are
2638 accessible only via `Registry` methods.
2639 """
2641 datastore: Datastore
2642 """The object that manages actual dataset storage (`Datastore`).
2644 Direct user access to the datastore should rarely be necessary; the primary
2645 exception is the case where a `Datastore` implementation provides extra
2646 functionality beyond what the base class defines.
2647 """
2649 storageClasses: StorageClassFactory
2650 """An object that maps known storage class names to objects that fully
2651 describe them (`StorageClassFactory`).
2652 """