Coverage for python/lsst/daf/butler/datastores/fileDatastore.py: 8%
975 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-07 02:10 -0700
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-07 02:10 -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/>.
21from __future__ import annotations
23"""Generic file-based datastore code."""
25__all__ = ("FileDatastore",)
27import hashlib
28import logging
29from collections import defaultdict
30from collections.abc import Callable
31from dataclasses import dataclass
32from typing import (
33 TYPE_CHECKING,
34 Any,
35 ClassVar,
36 Dict,
37 Iterable,
38 List,
39 Mapping,
40 Optional,
41 Sequence,
42 Set,
43 Tuple,
44 Type,
45 Union,
46)
48from lsst.daf.butler import (
49 CompositesMap,
50 Config,
51 DatasetId,
52 DatasetRef,
53 DatasetRefURIs,
54 DatasetType,
55 DatasetTypeNotSupportedError,
56 Datastore,
57 DatastoreCacheManager,
58 DatastoreConfig,
59 DatastoreDisabledCacheManager,
60 DatastoreRecordData,
61 DatastoreValidationError,
62 FileDataset,
63 FileDescriptor,
64 FileTemplates,
65 FileTemplateValidationError,
66 Formatter,
67 FormatterFactory,
68 Location,
69 LocationFactory,
70 Progress,
71 StorageClass,
72 StoredDatastoreItemInfo,
73 StoredFileInfo,
74 ddl,
75)
76from lsst.daf.butler.core.repoRelocation import replaceRoot
77from lsst.daf.butler.core.utils import transactional
78from lsst.daf.butler.registry.interfaces import DatastoreRegistryBridge, ReadOnlyDatabaseError
79from lsst.resources import ResourcePath, ResourcePathExpression
80from lsst.utils.introspection import get_class_of, get_instance_of
81from lsst.utils.iteration import chunk_iterable
83# For VERBOSE logging usage.
84from lsst.utils.logging import VERBOSE, getLogger
85from lsst.utils.timer import time_this
86from sqlalchemy import BigInteger, String
88from ..registry.interfaces import FakeDatasetRef
89from .genericDatastore import GenericBaseDatastore
91if TYPE_CHECKING:
92 from lsst.daf.butler import AbstractDatastoreCacheManager, LookupKey
93 from lsst.daf.butler.registry.interfaces import DatasetIdRef, DatastoreRegistryBridgeManager
95log = getLogger(__name__)
98class _IngestPrepData(Datastore.IngestPrepData):
99 """Helper class for FileDatastore ingest implementation.
101 Parameters
102 ----------
103 datasets : `list` of `FileDataset`
104 Files to be ingested by this datastore.
105 """
107 def __init__(self, datasets: List[FileDataset]):
108 super().__init__(ref for dataset in datasets for ref in dataset.refs)
109 self.datasets = datasets
112@dataclass(frozen=True)
113class DatastoreFileGetInformation:
114 """Collection of useful parameters needed to retrieve a file from
115 a Datastore.
116 """
118 location: Location
119 """The location from which to read the dataset."""
121 formatter: Formatter
122 """The `Formatter` to use to deserialize the dataset."""
124 info: StoredFileInfo
125 """Stored information about this file and its formatter."""
127 assemblerParams: Mapping[str, Any]
128 """Parameters to use for post-processing the retrieved dataset."""
130 formatterParams: Mapping[str, Any]
131 """Parameters that were understood by the associated formatter."""
133 component: Optional[str]
134 """The component to be retrieved (can be `None`)."""
136 readStorageClass: StorageClass
137 """The `StorageClass` of the dataset being read."""
140class FileDatastore(GenericBaseDatastore):
141 """Generic Datastore for file-based implementations.
143 Should always be sub-classed since key abstract methods are missing.
145 Parameters
146 ----------
147 config : `DatastoreConfig` or `str`
148 Configuration as either a `Config` object or URI to file.
149 bridgeManager : `DatastoreRegistryBridgeManager`
150 Object that manages the interface between `Registry` and datastores.
151 butlerRoot : `str`, optional
152 New datastore root to use to override the configuration value.
154 Raises
155 ------
156 ValueError
157 If root location does not exist and ``create`` is `False` in the
158 configuration.
159 """
161 defaultConfigFile: ClassVar[Optional[str]] = None
162 """Path to configuration defaults. Accessed within the ``config`` resource
163 or relative to a search path. Can be None if no defaults specified.
164 """
166 root: ResourcePath
167 """Root directory URI of this `Datastore`."""
169 locationFactory: LocationFactory
170 """Factory for creating locations relative to the datastore root."""
172 formatterFactory: FormatterFactory
173 """Factory for creating instances of formatters."""
175 templates: FileTemplates
176 """File templates that can be used by this `Datastore`."""
178 composites: CompositesMap
179 """Determines whether a dataset should be disassembled on put."""
181 defaultConfigFile = "datastores/fileDatastore.yaml"
182 """Path to configuration defaults. Accessed within the ``config`` resource
183 or relative to a search path. Can be None if no defaults specified.
184 """
186 _retrieve_dataset_method: Callable[[str], DatasetType | None] | None = None
187 """Callable that is used in trusted mode to retrieve registry definition
188 of a named dataset type.
189 """
191 @classmethod
192 def setConfigRoot(cls, root: str, config: Config, full: Config, overwrite: bool = True) -> None:
193 """Set any filesystem-dependent config options for this Datastore to
194 be appropriate for a new empty repository with the given root.
196 Parameters
197 ----------
198 root : `str`
199 URI to the root of the data repository.
200 config : `Config`
201 A `Config` to update. Only the subset understood by
202 this component will be updated. Will not expand
203 defaults.
204 full : `Config`
205 A complete config with all defaults expanded that can be
206 converted to a `DatastoreConfig`. Read-only and will not be
207 modified by this method.
208 Repository-specific options that should not be obtained
209 from defaults when Butler instances are constructed
210 should be copied from ``full`` to ``config``.
211 overwrite : `bool`, optional
212 If `False`, do not modify a value in ``config`` if the value
213 already exists. Default is always to overwrite with the provided
214 ``root``.
216 Notes
217 -----
218 If a keyword is explicitly defined in the supplied ``config`` it
219 will not be overridden by this method if ``overwrite`` is `False`.
220 This allows explicit values set in external configs to be retained.
221 """
222 Config.updateParameters(
223 DatastoreConfig,
224 config,
225 full,
226 toUpdate={"root": root},
227 toCopy=("cls", ("records", "table")),
228 overwrite=overwrite,
229 )
231 @classmethod
232 def makeTableSpec(cls, datasetIdColumnType: type) -> ddl.TableSpec:
233 return ddl.TableSpec(
234 fields=[
235 ddl.FieldSpec(name="dataset_id", dtype=datasetIdColumnType, primaryKey=True),
236 ddl.FieldSpec(name="path", dtype=String, length=256, nullable=False),
237 ddl.FieldSpec(name="formatter", dtype=String, length=128, nullable=False),
238 ddl.FieldSpec(name="storage_class", dtype=String, length=64, nullable=False),
239 # Use empty string to indicate no component
240 ddl.FieldSpec(name="component", dtype=String, length=32, primaryKey=True),
241 # TODO: should checksum be Base64Bytes instead?
242 ddl.FieldSpec(name="checksum", dtype=String, length=128, nullable=True),
243 ddl.FieldSpec(name="file_size", dtype=BigInteger, nullable=True),
244 ],
245 unique=frozenset(),
246 indexes=[ddl.IndexSpec("path")],
247 )
249 def __init__(
250 self,
251 config: Union[DatastoreConfig, str],
252 bridgeManager: DatastoreRegistryBridgeManager,
253 butlerRoot: str | None = None,
254 ):
255 super().__init__(config, bridgeManager)
256 if "root" not in self.config:
257 raise ValueError("No root directory specified in configuration")
259 self._bridgeManager = bridgeManager
261 # Name ourselves either using an explicit name or a name
262 # derived from the (unexpanded) root
263 if "name" in self.config:
264 self.name = self.config["name"]
265 else:
266 # We use the unexpanded root in the name to indicate that this
267 # datastore can be moved without having to update registry.
268 self.name = "{}@{}".format(type(self).__name__, self.config["root"])
270 # Support repository relocation in config
271 # Existence of self.root is checked in subclass
272 self.root = ResourcePath(
273 replaceRoot(self.config["root"], butlerRoot), forceDirectory=True, forceAbsolute=True
274 )
276 self.locationFactory = LocationFactory(self.root)
277 self.formatterFactory = FormatterFactory()
279 # Now associate formatters with storage classes
280 self.formatterFactory.registerFormatters(self.config["formatters"], universe=bridgeManager.universe)
282 # Read the file naming templates
283 self.templates = FileTemplates(self.config["templates"], universe=bridgeManager.universe)
285 # See if composites should be disassembled
286 self.composites = CompositesMap(self.config["composites"], universe=bridgeManager.universe)
288 tableName = self.config["records", "table"]
289 try:
290 # Storage of paths and formatters, keyed by dataset_id
291 self._table = bridgeManager.opaque.register(
292 tableName, self.makeTableSpec(bridgeManager.datasetIdColumnType)
293 )
294 # Interface to Registry.
295 self._bridge = bridgeManager.register(self.name)
296 except ReadOnlyDatabaseError:
297 # If the database is read only and we just tried and failed to
298 # create a table, it means someone is trying to create a read-only
299 # butler client for an empty repo. That should be okay, as long
300 # as they then try to get any datasets before some other client
301 # creates the table. Chances are they'rejust validating
302 # configuration.
303 pass
305 # Determine whether checksums should be used - default to False
306 self.useChecksum = self.config.get("checksum", False)
308 # Determine whether we can fall back to configuration if a
309 # requested dataset is not known to registry
310 self.trustGetRequest = self.config.get("trust_get_request", False)
312 # Create a cache manager
313 self.cacheManager: AbstractDatastoreCacheManager
314 if "cached" in self.config:
315 self.cacheManager = DatastoreCacheManager(self.config["cached"], universe=bridgeManager.universe)
316 else:
317 self.cacheManager = DatastoreDisabledCacheManager("", universe=bridgeManager.universe)
319 # Check existence and create directory structure if necessary
320 if not self.root.exists():
321 if "create" not in self.config or not self.config["create"]:
322 raise ValueError(f"No valid root and not allowed to create one at: {self.root}")
323 try:
324 self.root.mkdir()
325 except Exception as e:
326 raise ValueError(
327 f"Can not create datastore root '{self.root}', check permissions. Got error: {e}"
328 ) from e
330 def __str__(self) -> str:
331 return str(self.root)
333 @property
334 def bridge(self) -> DatastoreRegistryBridge:
335 return self._bridge
337 def _artifact_exists(self, location: Location) -> bool:
338 """Check that an artifact exists in this datastore at the specified
339 location.
341 Parameters
342 ----------
343 location : `Location`
344 Expected location of the artifact associated with this datastore.
346 Returns
347 -------
348 exists : `bool`
349 True if the location can be found, false otherwise.
350 """
351 log.debug("Checking if resource exists: %s", location.uri)
352 return location.uri.exists()
354 def _delete_artifact(self, location: Location) -> None:
355 """Delete the artifact from the datastore.
357 Parameters
358 ----------
359 location : `Location`
360 Location of the artifact associated with this datastore.
361 """
362 if location.pathInStore.isabs():
363 raise RuntimeError(f"Cannot delete artifact with absolute uri {location.uri}.")
365 try:
366 location.uri.remove()
367 except FileNotFoundError:
368 log.debug("File %s did not exist and so could not be deleted.", location.uri)
369 raise
370 except Exception as e:
371 log.critical("Failed to delete file: %s (%s)", location.uri, e)
372 raise
373 log.debug("Successfully deleted file: %s", location.uri)
375 def addStoredItemInfo(self, refs: Iterable[DatasetRef], infos: Iterable[StoredFileInfo]) -> None:
376 # Docstring inherited from GenericBaseDatastore
377 records = [info.rebase(ref).to_record() for ref, info in zip(refs, infos)]
378 self._table.insert(*records, transaction=self._transaction)
380 def getStoredItemsInfo(self, ref: DatasetIdRef) -> List[StoredFileInfo]:
381 # Docstring inherited from GenericBaseDatastore
383 # Look for the dataset_id -- there might be multiple matches
384 # if we have disassembled the dataset.
385 records = self._table.fetch(dataset_id=ref.id)
386 return [StoredFileInfo.from_record(record) for record in records]
388 def _get_stored_records_associated_with_refs(
389 self, refs: Iterable[DatasetIdRef]
390 ) -> Dict[DatasetId, List[StoredFileInfo]]:
391 """Retrieve all records associated with the provided refs.
393 Parameters
394 ----------
395 refs : iterable of `DatasetIdRef`
396 The refs for which records are to be retrieved.
398 Returns
399 -------
400 records : `dict` of [`DatasetId`, `list` of `StoredFileInfo`]
401 The matching records indexed by the ref ID. The number of entries
402 in the dict can be smaller than the number of requested refs.
403 """
404 records = self._table.fetch(dataset_id=[ref.id for ref in refs])
406 # Uniqueness is dataset_id + component so can have multiple records
407 # per ref.
408 records_by_ref = defaultdict(list)
409 for record in records:
410 records_by_ref[record["dataset_id"]].append(StoredFileInfo.from_record(record))
411 return records_by_ref
413 def _refs_associated_with_artifacts(
414 self, paths: List[Union[str, ResourcePath]]
415 ) -> Dict[str, Set[DatasetId]]:
416 """Return paths and associated dataset refs.
418 Parameters
419 ----------
420 paths : `list` of `str` or `lsst.resources.ResourcePath`
421 All the paths to include in search.
423 Returns
424 -------
425 mapping : `dict` of [`str`, `set` [`DatasetId`]]
426 Mapping of each path to a set of associated database IDs.
427 """
428 records = self._table.fetch(path=[str(path) for path in paths])
429 result = defaultdict(set)
430 for row in records:
431 result[row["path"]].add(row["dataset_id"])
432 return result
434 def _registered_refs_per_artifact(self, pathInStore: ResourcePath) -> Set[DatasetId]:
435 """Return all dataset refs associated with the supplied path.
437 Parameters
438 ----------
439 pathInStore : `lsst.resources.ResourcePath`
440 Path of interest in the data store.
442 Returns
443 -------
444 ids : `set` of `int`
445 All `DatasetRef` IDs associated with this path.
446 """
447 records = list(self._table.fetch(path=str(pathInStore)))
448 ids = {r["dataset_id"] for r in records}
449 return ids
451 def removeStoredItemInfo(self, ref: DatasetIdRef) -> None:
452 # Docstring inherited from GenericBaseDatastore
453 self._table.delete(["dataset_id"], {"dataset_id": ref.id})
455 def _get_dataset_locations_info(self, ref: DatasetIdRef) -> List[Tuple[Location, StoredFileInfo]]:
456 r"""Find all the `Location`\ s of the requested dataset in the
457 `Datastore` and the associated stored file information.
459 Parameters
460 ----------
461 ref : `DatasetRef`
462 Reference to the required `Dataset`.
464 Returns
465 -------
466 results : `list` [`tuple` [`Location`, `StoredFileInfo` ]]
467 Location of the dataset within the datastore and
468 stored information about each file and its formatter.
469 """
470 # Get the file information (this will fail if no file)
471 records = self.getStoredItemsInfo(ref)
473 # Use the path to determine the location -- we need to take
474 # into account absolute URIs in the datastore record
475 return [(r.file_location(self.locationFactory), r) for r in records]
477 def _can_remove_dataset_artifact(self, ref: DatasetIdRef, location: Location) -> bool:
478 """Check that there is only one dataset associated with the
479 specified artifact.
481 Parameters
482 ----------
483 ref : `DatasetRef` or `FakeDatasetRef`
484 Dataset to be removed.
485 location : `Location`
486 The location of the artifact to be removed.
488 Returns
489 -------
490 can_remove : `Bool`
491 True if the artifact can be safely removed.
492 """
493 # Can't ever delete absolute URIs.
494 if location.pathInStore.isabs():
495 return False
497 # Get all entries associated with this path
498 allRefs = self._registered_refs_per_artifact(location.pathInStore)
499 if not allRefs:
500 raise RuntimeError(f"Datastore inconsistency error. {location.pathInStore} not in registry")
502 # Remove these refs from all the refs and if there is nothing left
503 # then we can delete
504 remainingRefs = allRefs - {ref.id}
506 if remainingRefs:
507 return False
508 return True
510 def _get_expected_dataset_locations_info(self, ref: DatasetRef) -> List[Tuple[Location, StoredFileInfo]]:
511 """Predict the location and related file information of the requested
512 dataset in this datastore.
514 Parameters
515 ----------
516 ref : `DatasetRef`
517 Reference to the required `Dataset`.
519 Returns
520 -------
521 results : `list` [`tuple` [`Location`, `StoredFileInfo` ]]
522 Expected Location of the dataset within the datastore and
523 placeholder information about each file and its formatter.
525 Notes
526 -----
527 Uses the current configuration to determine how we would expect the
528 datastore files to have been written if we couldn't ask registry.
529 This is safe so long as there has been no change to datastore
530 configuration between writing the dataset and wanting to read it.
531 Will not work for files that have been ingested without using the
532 standard file template or default formatter.
533 """
535 # If we have a component ref we always need to ask the questions
536 # of the composite. If the composite is disassembled this routine
537 # should return all components. If the composite was not
538 # disassembled the composite is what is stored regardless of
539 # component request. Note that if the caller has disassembled
540 # a composite there is no way for this guess to know that
541 # without trying both the composite and component ref and seeing
542 # if there is something at the component Location even without
543 # disassembly being enabled.
544 if ref.datasetType.isComponent():
545 ref = ref.makeCompositeRef()
547 # See if the ref is a composite that should be disassembled
548 doDisassembly = self.composites.shouldBeDisassembled(ref)
550 all_info: List[Tuple[Location, Formatter, StorageClass, Optional[str]]] = []
552 if doDisassembly:
553 for component, componentStorage in ref.datasetType.storageClass.components.items():
554 compRef = ref.makeComponentRef(component)
555 location, formatter = self._determine_put_formatter_location(compRef)
556 all_info.append((location, formatter, componentStorage, component))
558 else:
559 # Always use the composite ref if no disassembly
560 location, formatter = self._determine_put_formatter_location(ref)
561 all_info.append((location, formatter, ref.datasetType.storageClass, None))
563 # Convert the list of tuples to have StoredFileInfo as second element
564 return [
565 (
566 location,
567 StoredFileInfo(
568 formatter=formatter,
569 path=location.pathInStore.path,
570 storageClass=storageClass,
571 component=component,
572 checksum=None,
573 file_size=-1,
574 dataset_id=ref.id,
575 ),
576 )
577 for location, formatter, storageClass, component in all_info
578 ]
580 def _prepare_for_get(
581 self, ref: DatasetRef, parameters: Optional[Mapping[str, Any]] = None
582 ) -> List[DatastoreFileGetInformation]:
583 """Check parameters for ``get`` and obtain formatter and
584 location.
586 Parameters
587 ----------
588 ref : `DatasetRef`
589 Reference to the required Dataset.
590 parameters : `dict`
591 `StorageClass`-specific parameters that specify, for example,
592 a slice of the dataset to be loaded.
594 Returns
595 -------
596 getInfo : `list` [`DatastoreFileGetInformation`]
597 Parameters needed to retrieve each file.
598 """
599 log.debug("Retrieve %s from %s with parameters %s", ref, self.name, parameters)
601 # For trusted mode need to reset storage class.
602 ref = self._cast_storage_class(ref)
604 # Get file metadata and internal metadata
605 fileLocations = self._get_dataset_locations_info(ref)
606 if not fileLocations:
607 if not self.trustGetRequest:
608 raise FileNotFoundError(f"Could not retrieve dataset {ref}.")
609 # Assume the dataset is where we think it should be
610 fileLocations = self._get_expected_dataset_locations_info(ref)
612 # The storage class we want to use eventually
613 refStorageClass = ref.datasetType.storageClass
615 if len(fileLocations) > 1:
616 disassembled = True
618 # If trust is involved it is possible that there will be
619 # components listed here that do not exist in the datastore.
620 # Explicitly check for file artifact existence and filter out any
621 # that are missing.
622 if self.trustGetRequest:
623 fileLocations = [loc for loc in fileLocations if loc[0].uri.exists()]
625 # For now complain only if we have no components at all. One
626 # component is probably a problem but we can punt that to the
627 # assembler.
628 if not fileLocations:
629 raise FileNotFoundError(f"None of the component files for dataset {ref} exist.")
631 else:
632 disassembled = False
634 # Is this a component request?
635 refComponent = ref.datasetType.component()
637 fileGetInfo = []
638 for location, storedFileInfo in fileLocations:
639 # The storage class used to write the file
640 writeStorageClass = storedFileInfo.storageClass
642 # If this has been disassembled we need read to match the write
643 if disassembled:
644 readStorageClass = writeStorageClass
645 else:
646 readStorageClass = refStorageClass
648 formatter = get_instance_of(
649 storedFileInfo.formatter,
650 FileDescriptor(
651 location,
652 readStorageClass=readStorageClass,
653 storageClass=writeStorageClass,
654 parameters=parameters,
655 ),
656 ref.dataId,
657 )
659 formatterParams, notFormatterParams = formatter.segregateParameters()
661 # Of the remaining parameters, extract the ones supported by
662 # this StorageClass (for components not all will be handled)
663 assemblerParams = readStorageClass.filterParameters(notFormatterParams)
665 # The ref itself could be a component if the dataset was
666 # disassembled by butler, or we disassembled in datastore and
667 # components came from the datastore records
668 component = storedFileInfo.component if storedFileInfo.component else refComponent
670 fileGetInfo.append(
671 DatastoreFileGetInformation(
672 location,
673 formatter,
674 storedFileInfo,
675 assemblerParams,
676 formatterParams,
677 component,
678 readStorageClass,
679 )
680 )
682 return fileGetInfo
684 def _prepare_for_put(self, inMemoryDataset: Any, ref: DatasetRef) -> Tuple[Location, Formatter]:
685 """Check the arguments for ``put`` and obtain formatter and
686 location.
688 Parameters
689 ----------
690 inMemoryDataset : `object`
691 The dataset to store.
692 ref : `DatasetRef`
693 Reference to the associated Dataset.
695 Returns
696 -------
697 location : `Location`
698 The location to write the dataset.
699 formatter : `Formatter`
700 The `Formatter` to use to write the dataset.
702 Raises
703 ------
704 TypeError
705 Supplied object and storage class are inconsistent.
706 DatasetTypeNotSupportedError
707 The associated `DatasetType` is not handled by this datastore.
708 """
709 self._validate_put_parameters(inMemoryDataset, ref)
710 return self._determine_put_formatter_location(ref)
712 def _determine_put_formatter_location(self, ref: DatasetRef) -> Tuple[Location, Formatter]:
713 """Calculate the formatter and output location to use for put.
715 Parameters
716 ----------
717 ref : `DatasetRef`
718 Reference to the associated Dataset.
720 Returns
721 -------
722 location : `Location`
723 The location to write the dataset.
724 formatter : `Formatter`
725 The `Formatter` to use to write the dataset.
726 """
727 # Work out output file name
728 try:
729 template = self.templates.getTemplate(ref)
730 except KeyError as e:
731 raise DatasetTypeNotSupportedError(f"Unable to find template for {ref}") from e
733 # Validate the template to protect against filenames from different
734 # dataIds returning the same and causing overwrite confusion.
735 template.validateTemplate(ref)
737 location = self.locationFactory.fromPath(template.format(ref))
739 # Get the formatter based on the storage class
740 storageClass = ref.datasetType.storageClass
741 try:
742 formatter = self.formatterFactory.getFormatter(
743 ref, FileDescriptor(location, storageClass=storageClass), ref.dataId
744 )
745 except KeyError as e:
746 raise DatasetTypeNotSupportedError(
747 f"Unable to find formatter for {ref} in datastore {self.name}"
748 ) from e
750 # Now that we know the formatter, update the location
751 location = formatter.makeUpdatedLocation(location)
753 return location, formatter
755 def _overrideTransferMode(self, *datasets: FileDataset, transfer: Optional[str] = None) -> Optional[str]:
756 # Docstring inherited from base class
757 if transfer != "auto":
758 return transfer
760 # See if the paths are within the datastore or not
761 inside = [self._pathInStore(d.path) is not None for d in datasets]
763 if all(inside):
764 transfer = None
765 elif not any(inside):
766 # Allow ResourcePath to use its own knowledge
767 transfer = "auto"
768 else:
769 # This can happen when importing from a datastore that
770 # has had some datasets ingested using "direct" mode.
771 # Also allow ResourcePath to sort it out but warn about it.
772 # This can happen if you are importing from a datastore
773 # that had some direct transfer datasets.
774 log.warning(
775 "Some datasets are inside the datastore and some are outside. Using 'split' "
776 "transfer mode. This assumes that the files outside the datastore are "
777 "still accessible to the new butler since they will not be copied into "
778 "the target datastore."
779 )
780 transfer = "split"
782 return transfer
784 def _pathInStore(self, path: ResourcePathExpression) -> Optional[str]:
785 """Return path relative to datastore root
787 Parameters
788 ----------
789 path : `lsst.resources.ResourcePathExpression`
790 Path to dataset. Can be absolute URI. If relative assumed to
791 be relative to the datastore. Returns path in datastore
792 or raises an exception if the path it outside.
794 Returns
795 -------
796 inStore : `str`
797 Path relative to datastore root. Returns `None` if the file is
798 outside the root.
799 """
800 # Relative path will always be relative to datastore
801 pathUri = ResourcePath(path, forceAbsolute=False)
802 return pathUri.relative_to(self.root)
804 def _standardizeIngestPath(
805 self, path: Union[str, ResourcePath], *, transfer: Optional[str] = None
806 ) -> Union[str, ResourcePath]:
807 """Standardize the path of a to-be-ingested file.
809 Parameters
810 ----------
811 path : `str` or `lsst.resources.ResourcePath`
812 Path of a file to be ingested. This parameter is not expected
813 to be all the types that can be used to construct a
814 `~lsst.resources.ResourcePath`.
815 transfer : `str`, optional
816 How (and whether) the dataset should be added to the datastore.
817 See `ingest` for details of transfer modes.
818 This implementation is provided only so
819 `NotImplementedError` can be raised if the mode is not supported;
820 actual transfers are deferred to `_extractIngestInfo`.
822 Returns
823 -------
824 path : `str` or `lsst.resources.ResourcePath`
825 New path in what the datastore considers standard form. If an
826 absolute URI was given that will be returned unchanged.
828 Notes
829 -----
830 Subclasses of `FileDatastore` can implement this method instead
831 of `_prepIngest`. It should not modify the data repository or given
832 file in any way.
834 Raises
835 ------
836 NotImplementedError
837 Raised if the datastore does not support the given transfer mode
838 (including the case where ingest is not supported at all).
839 FileNotFoundError
840 Raised if one of the given files does not exist.
841 """
842 if transfer not in (None, "direct", "split") + self.root.transferModes:
843 raise NotImplementedError(f"Transfer mode {transfer} not supported.")
845 # A relative URI indicates relative to datastore root
846 srcUri = ResourcePath(path, forceAbsolute=False)
847 if not srcUri.isabs():
848 srcUri = self.root.join(path)
850 if not srcUri.exists():
851 raise FileNotFoundError(
852 f"Resource at {srcUri} does not exist; note that paths to ingest "
853 f"are assumed to be relative to {self.root} unless they are absolute."
854 )
856 if transfer is None:
857 relpath = srcUri.relative_to(self.root)
858 if not relpath:
859 raise RuntimeError(
860 f"Transfer is none but source file ({srcUri}) is not within datastore ({self.root})"
861 )
863 # Return the relative path within the datastore for internal
864 # transfer
865 path = relpath
867 return path
869 def _extractIngestInfo(
870 self,
871 path: ResourcePathExpression,
872 ref: DatasetRef,
873 *,
874 formatter: Union[Formatter, Type[Formatter]],
875 transfer: Optional[str] = None,
876 record_validation_info: bool = True,
877 ) -> StoredFileInfo:
878 """Relocate (if necessary) and extract `StoredFileInfo` from a
879 to-be-ingested file.
881 Parameters
882 ----------
883 path : `lsst.resources.ResourcePathExpression`
884 URI or path of a file to be ingested.
885 ref : `DatasetRef`
886 Reference for the dataset being ingested. Guaranteed to have
887 ``dataset_id not None`.
888 formatter : `type` or `Formatter`
889 `Formatter` subclass to use for this dataset or an instance.
890 transfer : `str`, optional
891 How (and whether) the dataset should be added to the datastore.
892 See `ingest` for details of transfer modes.
893 record_validation_info : `bool`, optional
894 If `True`, the default, the datastore can record validation
895 information associated with the file. If `False` the datastore
896 will not attempt to track any information such as checksums
897 or file sizes. This can be useful if such information is tracked
898 in an external system or if the file is to be compressed in place.
899 It is up to the datastore whether this parameter is relevant.
901 Returns
902 -------
903 info : `StoredFileInfo`
904 Internal datastore record for this file. This will be inserted by
905 the caller; the `_extractIngestInfo` is only responsible for
906 creating and populating the struct.
908 Raises
909 ------
910 FileNotFoundError
911 Raised if one of the given files does not exist.
912 FileExistsError
913 Raised if transfer is not `None` but the (internal) location the
914 file would be moved to is already occupied.
915 """
916 if self._transaction is None:
917 raise RuntimeError("Ingest called without transaction enabled")
919 # Create URI of the source path, do not need to force a relative
920 # path to absolute.
921 srcUri = ResourcePath(path, forceAbsolute=False)
923 # Track whether we have read the size of the source yet
924 have_sized = False
926 tgtLocation: Optional[Location]
927 if transfer is None or transfer == "split":
928 # A relative path is assumed to be relative to the datastore
929 # in this context
930 if not srcUri.isabs():
931 tgtLocation = self.locationFactory.fromPath(srcUri.ospath)
932 else:
933 # Work out the path in the datastore from an absolute URI
934 # This is required to be within the datastore.
935 pathInStore = srcUri.relative_to(self.root)
936 if pathInStore is None and transfer is None:
937 raise RuntimeError(
938 f"Unexpectedly learned that {srcUri} is not within datastore {self.root}"
939 )
940 if pathInStore:
941 tgtLocation = self.locationFactory.fromPath(pathInStore)
942 elif transfer == "split":
943 # Outside the datastore but treat that as a direct ingest
944 # instead.
945 tgtLocation = None
946 else:
947 raise RuntimeError(f"Unexpected transfer mode encountered: {transfer} for URI {srcUri}")
948 elif transfer == "direct":
949 # Want to store the full URI to the resource directly in
950 # datastore. This is useful for referring to permanent archive
951 # storage for raw data.
952 # Trust that people know what they are doing.
953 tgtLocation = None
954 else:
955 # Work out the name we want this ingested file to have
956 # inside the datastore
957 tgtLocation = self._calculate_ingested_datastore_name(srcUri, ref, formatter)
958 if not tgtLocation.uri.dirname().exists():
959 log.debug("Folder %s does not exist yet.", tgtLocation.uri.dirname())
960 tgtLocation.uri.dirname().mkdir()
962 # if we are transferring from a local file to a remote location
963 # it may be more efficient to get the size and checksum of the
964 # local file rather than the transferred one
965 if record_validation_info and srcUri.isLocal:
966 size = srcUri.size()
967 checksum = self.computeChecksum(srcUri) if self.useChecksum else None
968 have_sized = True
970 # Transfer the resource to the destination.
971 # Allow overwrite of an existing file. This matches the behavior
972 # of datastore.put() in that it trusts that registry would not
973 # be asking to overwrite unless registry thought that the
974 # overwrite was allowed.
975 tgtLocation.uri.transfer_from(
976 srcUri, transfer=transfer, transaction=self._transaction, overwrite=True
977 )
979 if tgtLocation is None:
980 # This means we are using direct mode
981 targetUri = srcUri
982 targetPath = str(srcUri)
983 else:
984 targetUri = tgtLocation.uri
985 targetPath = tgtLocation.pathInStore.path
987 # the file should exist in the datastore now
988 if record_validation_info:
989 if not have_sized:
990 size = targetUri.size()
991 checksum = self.computeChecksum(targetUri) if self.useChecksum else None
992 else:
993 # Not recording any file information.
994 size = -1
995 checksum = None
997 return StoredFileInfo(
998 formatter=formatter,
999 path=targetPath,
1000 storageClass=ref.datasetType.storageClass,
1001 component=ref.datasetType.component(),
1002 file_size=size,
1003 checksum=checksum,
1004 dataset_id=ref.id,
1005 )
1007 def _prepIngest(self, *datasets: FileDataset, transfer: Optional[str] = None) -> _IngestPrepData:
1008 # Docstring inherited from Datastore._prepIngest.
1009 filtered = []
1010 for dataset in datasets:
1011 acceptable = [ref for ref in dataset.refs if self.constraints.isAcceptable(ref)]
1012 if not acceptable:
1013 continue
1014 else:
1015 dataset.refs = acceptable
1016 if dataset.formatter is None:
1017 dataset.formatter = self.formatterFactory.getFormatterClass(dataset.refs[0])
1018 else:
1019 assert isinstance(dataset.formatter, (type, str))
1020 formatter_class = get_class_of(dataset.formatter)
1021 if not issubclass(formatter_class, Formatter):
1022 raise TypeError(f"Requested formatter {dataset.formatter} is not a Formatter class.")
1023 dataset.formatter = formatter_class
1024 dataset.path = self._standardizeIngestPath(dataset.path, transfer=transfer)
1025 filtered.append(dataset)
1026 return _IngestPrepData(filtered)
1028 @transactional
1029 def _finishIngest(
1030 self,
1031 prepData: Datastore.IngestPrepData,
1032 *,
1033 transfer: Optional[str] = None,
1034 record_validation_info: bool = True,
1035 ) -> None:
1036 # Docstring inherited from Datastore._finishIngest.
1037 refsAndInfos = []
1038 progress = Progress("lsst.daf.butler.datastores.FileDatastore.ingest", level=logging.DEBUG)
1039 for dataset in progress.wrap(prepData.datasets, desc="Ingesting dataset files"):
1040 # Do ingest as if the first dataset ref is associated with the file
1041 info = self._extractIngestInfo(
1042 dataset.path,
1043 dataset.refs[0],
1044 formatter=dataset.formatter,
1045 transfer=transfer,
1046 record_validation_info=record_validation_info,
1047 )
1048 refsAndInfos.extend([(ref, info) for ref in dataset.refs])
1049 self._register_datasets(refsAndInfos)
1051 def _calculate_ingested_datastore_name(
1052 self,
1053 srcUri: ResourcePath,
1054 ref: DatasetRef,
1055 formatter: Formatter | Type[Formatter] | None = None,
1056 ) -> Location:
1057 """Given a source URI and a DatasetRef, determine the name the
1058 dataset will have inside datastore.
1060 Parameters
1061 ----------
1062 srcUri : `lsst.resources.ResourcePath`
1063 URI to the source dataset file.
1064 ref : `DatasetRef`
1065 Ref associated with the newly-ingested dataset artifact. This
1066 is used to determine the name within the datastore.
1067 formatter : `Formatter` or Formatter class.
1068 Formatter to use for validation. Can be a class or an instance.
1069 No validation of the file extension is performed if the
1070 ``formatter`` is `None`. This can be used if the caller knows
1071 that the source URI and target URI will use the same formatter.
1073 Returns
1074 -------
1075 location : `Location`
1076 Target location for the newly-ingested dataset.
1077 """
1078 # Ingesting a file from outside the datastore.
1079 # This involves a new name.
1080 template = self.templates.getTemplate(ref)
1081 location = self.locationFactory.fromPath(template.format(ref))
1083 # Get the extension
1084 ext = srcUri.getExtension()
1086 # Update the destination to include that extension
1087 location.updateExtension(ext)
1089 # Ask the formatter to validate this extension
1090 if formatter is not None:
1091 formatter.validateExtension(location)
1093 return location
1095 def _write_in_memory_to_artifact(self, inMemoryDataset: Any, ref: DatasetRef) -> StoredFileInfo:
1096 """Write out in memory dataset to datastore.
1098 Parameters
1099 ----------
1100 inMemoryDataset : `object`
1101 Dataset to write to datastore.
1102 ref : `DatasetRef`
1103 Registry information associated with this dataset.
1105 Returns
1106 -------
1107 info : `StoredFileInfo`
1108 Information describing the artifact written to the datastore.
1109 """
1110 # May need to coerce the in memory dataset to the correct
1111 # python type, but first we need to make sure the storage class
1112 # reflects the one defined in the data repository.
1113 ref = self._cast_storage_class(ref)
1114 inMemoryDataset = ref.datasetType.storageClass.coerce_type(inMemoryDataset)
1116 location, formatter = self._prepare_for_put(inMemoryDataset, ref)
1117 uri = location.uri
1119 if not uri.dirname().exists():
1120 log.debug("Folder %s does not exist yet so creating it.", uri.dirname())
1121 uri.dirname().mkdir()
1123 if self._transaction is None:
1124 raise RuntimeError("Attempting to write artifact without transaction enabled")
1126 def _removeFileExists(uri: ResourcePath) -> None:
1127 """Remove a file and do not complain if it is not there.
1129 This is important since a formatter might fail before the file
1130 is written and we should not confuse people by writing spurious
1131 error messages to the log.
1132 """
1133 try:
1134 uri.remove()
1135 except FileNotFoundError:
1136 pass
1138 # Register a callback to try to delete the uploaded data if
1139 # something fails below
1140 self._transaction.registerUndo("artifactWrite", _removeFileExists, uri)
1142 data_written = False
1143 if not uri.isLocal:
1144 # This is a remote URI. Some datasets can be serialized directly
1145 # to bytes and sent to the remote datastore without writing a
1146 # file. If the dataset is intended to be saved to the cache
1147 # a file is always written and direct write to the remote
1148 # datastore is bypassed.
1149 if not self.cacheManager.should_be_cached(ref):
1150 try:
1151 serializedDataset = formatter.toBytes(inMemoryDataset)
1152 except NotImplementedError:
1153 # Fallback to the file writing option.
1154 pass
1155 except Exception as e:
1156 raise RuntimeError(
1157 f"Failed to serialize dataset {ref} of type {type(inMemoryDataset)} to bytes."
1158 ) from e
1159 else:
1160 log.debug("Writing bytes directly to %s", uri)
1161 uri.write(serializedDataset, overwrite=True)
1162 log.debug("Successfully wrote bytes directly to %s", uri)
1163 data_written = True
1165 if not data_written:
1166 # Did not write the bytes directly to object store so instead
1167 # write to temporary file. Always write to a temporary even if
1168 # using a local file system -- that gives us atomic writes.
1169 # If a process is killed as the file is being written we do not
1170 # want it to remain in the correct place but in corrupt state.
1171 # For local files write to the output directory not temporary dir.
1172 prefix = uri.dirname() if uri.isLocal else None
1173 with ResourcePath.temporary_uri(suffix=uri.getExtension(), prefix=prefix) as temporary_uri:
1174 # Need to configure the formatter to write to a different
1175 # location and that needs us to overwrite internals
1176 log.debug("Writing dataset to temporary location at %s", temporary_uri)
1177 with formatter._updateLocation(Location(None, temporary_uri)):
1178 try:
1179 formatter.write(inMemoryDataset)
1180 except Exception as e:
1181 raise RuntimeError(
1182 f"Failed to serialize dataset {ref} of type"
1183 f" {type(inMemoryDataset)} to "
1184 f"temporary location {temporary_uri}"
1185 ) from e
1187 # Use move for a local file since that becomes an efficient
1188 # os.rename. For remote resources we use copy to allow the
1189 # file to be cached afterwards.
1190 transfer = "move" if uri.isLocal else "copy"
1192 uri.transfer_from(temporary_uri, transfer=transfer, overwrite=True)
1194 if transfer == "copy":
1195 # Cache if required
1196 self.cacheManager.move_to_cache(temporary_uri, ref)
1198 log.debug("Successfully wrote dataset to %s via a temporary file.", uri)
1200 # URI is needed to resolve what ingest case are we dealing with
1201 return self._extractIngestInfo(uri, ref, formatter=formatter)
1203 def _read_artifact_into_memory(
1204 self,
1205 getInfo: DatastoreFileGetInformation,
1206 ref: DatasetRef,
1207 isComponent: bool = False,
1208 cache_ref: Optional[DatasetRef] = None,
1209 ) -> Any:
1210 """Read the artifact from datastore into in memory object.
1212 Parameters
1213 ----------
1214 getInfo : `DatastoreFileGetInformation`
1215 Information about the artifact within the datastore.
1216 ref : `DatasetRef`
1217 The registry information associated with this artifact.
1218 isComponent : `bool`
1219 Flag to indicate if a component is being read from this artifact.
1220 cache_ref : `DatasetRef`, optional
1221 The DatasetRef to use when looking up the file in the cache.
1222 This ref must have the same ID as the supplied ref but can
1223 be a parent ref or component ref to indicate to the cache whether
1224 a composite file is being requested from the cache or a component
1225 file. Without this the cache will default to the supplied ref but
1226 it can get confused with read-only derived components for
1227 disassembled composites.
1229 Returns
1230 -------
1231 inMemoryDataset : `object`
1232 The artifact as a python object.
1233 """
1234 location = getInfo.location
1235 uri = location.uri
1236 log.debug("Accessing data from %s", uri)
1238 if cache_ref is None:
1239 cache_ref = ref
1240 if cache_ref.id != ref.id:
1241 raise ValueError(
1242 "The supplied cache dataset ref refers to a different dataset than expected:"
1243 f" {ref.id} != {cache_ref.id}"
1244 )
1246 # Cannot recalculate checksum but can compare size as a quick check
1247 # Do not do this if the size is negative since that indicates
1248 # we do not know.
1249 recorded_size = getInfo.info.file_size
1250 resource_size = uri.size()
1251 if recorded_size >= 0 and resource_size != recorded_size:
1252 raise RuntimeError(
1253 "Integrity failure in Datastore. "
1254 f"Size of file {uri} ({resource_size}) "
1255 f"does not match size recorded in registry of {recorded_size}"
1256 )
1258 # For the general case we have choices for how to proceed.
1259 # 1. Always use a local file (downloading the remote resource to a
1260 # temporary file if needed).
1261 # 2. Use a threshold size and read into memory and use bytes.
1262 # Use both for now with an arbitrary hand off size.
1263 # This allows small datasets to be downloaded from remote object
1264 # stores without requiring a temporary file.
1266 formatter = getInfo.formatter
1267 nbytes_max = 10_000_000 # Arbitrary number that we can tune
1268 if resource_size <= nbytes_max and formatter.can_read_bytes():
1269 with self.cacheManager.find_in_cache(cache_ref, uri.getExtension()) as cached_file:
1270 if cached_file is not None:
1271 desired_uri = cached_file
1272 msg = f" (cached version of {uri})"
1273 else:
1274 desired_uri = uri
1275 msg = ""
1276 with time_this(log, msg="Reading bytes from %s%s", args=(desired_uri, msg)):
1277 serializedDataset = desired_uri.read()
1278 log.debug(
1279 "Deserializing %s from %d bytes from location %s with formatter %s",
1280 f"component {getInfo.component}" if isComponent else "",
1281 len(serializedDataset),
1282 uri,
1283 formatter.name(),
1284 )
1285 try:
1286 result = formatter.fromBytes(
1287 serializedDataset, component=getInfo.component if isComponent else None
1288 )
1289 except Exception as e:
1290 raise ValueError(
1291 f"Failure from formatter '{formatter.name()}' for dataset {ref.id}"
1292 f" ({ref.datasetType.name} from {uri}): {e}"
1293 ) from e
1294 else:
1295 # Read from file.
1297 # Have to update the Location associated with the formatter
1298 # because formatter.read does not allow an override.
1299 # This could be improved.
1300 location_updated = False
1301 msg = ""
1303 # First check in cache for local version.
1304 # The cache will only be relevant for remote resources but
1305 # no harm in always asking. Context manager ensures that cache
1306 # file is not deleted during cache expiration.
1307 with self.cacheManager.find_in_cache(cache_ref, uri.getExtension()) as cached_file:
1308 if cached_file is not None:
1309 msg = f"(via cache read of remote file {uri})"
1310 uri = cached_file
1311 location_updated = True
1313 with uri.as_local() as local_uri:
1314 can_be_cached = False
1315 if uri != local_uri:
1316 # URI was remote and file was downloaded
1317 cache_msg = ""
1318 location_updated = True
1320 if self.cacheManager.should_be_cached(cache_ref):
1321 # In this scenario we want to ask if the downloaded
1322 # file should be cached but we should not cache
1323 # it until after we've used it (to ensure it can't
1324 # be expired whilst we are using it).
1325 can_be_cached = True
1327 # Say that it is "likely" to be cached because
1328 # if the formatter read fails we will not be
1329 # caching this file.
1330 cache_msg = " and likely cached"
1332 msg = f"(via download to local file{cache_msg})"
1334 # Calculate the (possibly) new location for the formatter
1335 # to use.
1336 newLocation = Location(*local_uri.split()) if location_updated else None
1338 log.debug(
1339 "Reading%s from location %s %s with formatter %s",
1340 f" component {getInfo.component}" if isComponent else "",
1341 uri,
1342 msg,
1343 formatter.name(),
1344 )
1345 try:
1346 with formatter._updateLocation(newLocation):
1347 with time_this(
1348 log,
1349 msg="Reading%s from location %s %s with formatter %s",
1350 args=(
1351 f" component {getInfo.component}" if isComponent else "",
1352 uri,
1353 msg,
1354 formatter.name(),
1355 ),
1356 ):
1357 result = formatter.read(component=getInfo.component if isComponent else None)
1358 except Exception as e:
1359 raise ValueError(
1360 f"Failure from formatter '{formatter.name()}' for dataset {ref.id}"
1361 f" ({ref.datasetType.name} from {uri}): {e}"
1362 ) from e
1364 # File was read successfully so can move to cache
1365 if can_be_cached:
1366 self.cacheManager.move_to_cache(local_uri, cache_ref)
1368 return self._post_process_get(
1369 result, ref.datasetType.storageClass, getInfo.assemblerParams, isComponent=isComponent
1370 )
1372 def knows(self, ref: DatasetRef) -> bool:
1373 """Check if the dataset is known to the datastore.
1375 Does not check for existence of any artifact.
1377 Parameters
1378 ----------
1379 ref : `DatasetRef`
1380 Reference to the required dataset.
1382 Returns
1383 -------
1384 exists : `bool`
1385 `True` if the dataset is known to the datastore.
1386 """
1387 fileLocations = self._get_dataset_locations_info(ref)
1388 if fileLocations:
1389 return True
1390 return False
1392 def knows_these(self, refs: Iterable[DatasetRef]) -> dict[DatasetRef, bool]:
1393 # Docstring inherited from the base class.
1395 # The records themselves. Could be missing some entries.
1396 records = self._get_stored_records_associated_with_refs(refs)
1398 return {ref: ref.id in records for ref in refs}
1400 def _process_mexists_records(
1401 self,
1402 id_to_ref: Dict[DatasetId, DatasetRef],
1403 records: Dict[DatasetId, List[StoredFileInfo]],
1404 all_required: bool,
1405 artifact_existence: Optional[Dict[ResourcePath, bool]] = None,
1406 ) -> Dict[DatasetRef, bool]:
1407 """Helper function for mexists that checks the given records.
1409 Parameters
1410 ----------
1411 id_to_ref : `dict` of [`DatasetId`, `DatasetRef`]
1412 Mapping of the dataset ID to the dataset ref itself.
1413 records : `dict` of [`DatasetId`, `list` of `StoredFileInfo`]
1414 Records as generally returned by
1415 ``_get_stored_records_associated_with_refs``.
1416 all_required : `bool`
1417 Flag to indicate whether existence requires all artifacts
1418 associated with a dataset ID to exist or not for existence.
1419 artifact_existence : `dict` [`lsst.resources.ResourcePath`, `bool`]
1420 Optional mapping of datastore artifact to existence. Updated by
1421 this method with details of all artifacts tested. Can be `None`
1422 if the caller is not interested.
1424 Returns
1425 -------
1426 existence : `dict` of [`DatasetRef`, `bool`]
1427 Mapping from dataset to boolean indicating existence.
1428 """
1429 # The URIs to be checked and a mapping of those URIs to
1430 # the dataset ID.
1431 uris_to_check: List[ResourcePath] = []
1432 location_map: Dict[ResourcePath, DatasetId] = {}
1434 location_factory = self.locationFactory
1436 uri_existence: Dict[ResourcePath, bool] = {}
1437 for ref_id, infos in records.items():
1438 # Key is the dataset Id, value is list of StoredItemInfo
1439 uris = [info.file_location(location_factory).uri for info in infos]
1440 location_map.update({uri: ref_id for uri in uris})
1442 # Check the local cache directly for a dataset corresponding
1443 # to the remote URI.
1444 if self.cacheManager.file_count > 0:
1445 ref = id_to_ref[ref_id]
1446 for uri, storedFileInfo in zip(uris, infos):
1447 check_ref = ref
1448 if not ref.datasetType.isComponent() and (component := storedFileInfo.component):
1449 check_ref = ref.makeComponentRef(component)
1450 if self.cacheManager.known_to_cache(check_ref, uri.getExtension()):
1451 # Proxy for URI existence.
1452 uri_existence[uri] = True
1453 else:
1454 uris_to_check.append(uri)
1455 else:
1456 # Check all of them.
1457 uris_to_check.extend(uris)
1459 if artifact_existence is not None:
1460 # If a URI has already been checked remove it from the list
1461 # and immediately add the status to the output dict.
1462 filtered_uris_to_check = []
1463 for uri in uris_to_check:
1464 if uri in artifact_existence:
1465 uri_existence[uri] = artifact_existence[uri]
1466 else:
1467 filtered_uris_to_check.append(uri)
1468 uris_to_check = filtered_uris_to_check
1470 # Results.
1471 dataset_existence: Dict[DatasetRef, bool] = {}
1473 uri_existence.update(ResourcePath.mexists(uris_to_check))
1474 for uri, exists in uri_existence.items():
1475 dataset_id = location_map[uri]
1476 ref = id_to_ref[dataset_id]
1478 # Disassembled composite needs to check all locations.
1479 # all_required indicates whether all need to exist or not.
1480 if ref in dataset_existence:
1481 if all_required:
1482 exists = dataset_existence[ref] and exists
1483 else:
1484 exists = dataset_existence[ref] or exists
1485 dataset_existence[ref] = exists
1487 if artifact_existence is not None:
1488 artifact_existence.update(uri_existence)
1490 return dataset_existence
1492 def mexists(
1493 self, refs: Iterable[DatasetRef], artifact_existence: Optional[Dict[ResourcePath, bool]] = None
1494 ) -> Dict[DatasetRef, bool]:
1495 """Check the existence of multiple datasets at once.
1497 Parameters
1498 ----------
1499 refs : iterable of `DatasetRef`
1500 The datasets to be checked.
1501 artifact_existence : `dict` [`lsst.resources.ResourcePath`, `bool`]
1502 Optional mapping of datastore artifact to existence. Updated by
1503 this method with details of all artifacts tested. Can be `None`
1504 if the caller is not interested.
1506 Returns
1507 -------
1508 existence : `dict` of [`DatasetRef`, `bool`]
1509 Mapping from dataset to boolean indicating existence.
1511 Notes
1512 -----
1513 To minimize potentially costly remote existence checks, the local
1514 cache is checked as a proxy for existence. If a file for this
1515 `DatasetRef` does exist no check is done for the actual URI. This
1516 could result in possibly unexpected behavior if the dataset itself
1517 has been removed from the datastore by another process whilst it is
1518 still in the cache.
1519 """
1520 chunk_size = 10_000
1521 dataset_existence: Dict[DatasetRef, bool] = {}
1522 log.debug("Checking for the existence of multiple artifacts in datastore in chunks of %d", chunk_size)
1523 n_found_total = 0
1524 n_checked = 0
1525 n_chunks = 0
1526 for chunk in chunk_iterable(refs, chunk_size=chunk_size):
1527 chunk_result = self._mexists(chunk, artifact_existence)
1529 # The log message level and content depend on how many
1530 # datasets we are processing.
1531 n_results = len(chunk_result)
1533 # Use verbose logging to ensure that messages can be seen
1534 # easily if many refs are being checked.
1535 log_threshold = VERBOSE
1536 n_checked += n_results
1538 # This sum can take some time so only do it if we know the
1539 # result is going to be used.
1540 n_found = 0
1541 if log.isEnabledFor(log_threshold):
1542 # Can treat the booleans as 0, 1 integers and sum them.
1543 n_found = sum(chunk_result.values())
1544 n_found_total += n_found
1546 # We are deliberately not trying to count the number of refs
1547 # provided in case it's in the millions. This means there is a
1548 # situation where the number of refs exactly matches the chunk
1549 # size and we will switch to the multi-chunk path even though
1550 # we only have a single chunk.
1551 if n_results < chunk_size and n_chunks == 0:
1552 # Single chunk will be processed so we can provide more detail.
1553 if n_results == 1:
1554 ref = list(chunk_result)[0]
1555 # Use debug logging to be consistent with `exists()`.
1556 log.debug(
1557 "Calling mexists() with single ref that does%s exist (%s).",
1558 "" if chunk_result[ref] else " not",
1559 ref,
1560 )
1561 else:
1562 # Single chunk but multiple files. Summarize.
1563 log.log(
1564 log_threshold,
1565 "Number of datasets found in datastore: %d out of %d datasets checked.",
1566 n_found,
1567 n_checked,
1568 )
1570 else:
1571 # Use incremental verbose logging when we have multiple chunks.
1572 log.log(
1573 log_threshold,
1574 "Number of datasets found in datastore for chunk %d: %d out of %d checked "
1575 "(running total from all chunks so far: %d found out of %d checked)",
1576 n_chunks,
1577 n_found,
1578 n_results,
1579 n_found_total,
1580 n_checked,
1581 )
1582 dataset_existence.update(chunk_result)
1583 n_chunks += 1
1585 return dataset_existence
1587 def _mexists(
1588 self, refs: Sequence[DatasetRef], artifact_existence: Optional[Dict[ResourcePath, bool]] = None
1589 ) -> Dict[DatasetRef, bool]:
1590 """Check the existence of multiple datasets at once.
1592 Parameters
1593 ----------
1594 refs : iterable of `DatasetRef`
1595 The datasets to be checked.
1596 artifact_existence : `dict` [`lsst.resources.ResourcePath`, `bool`]
1597 Optional mapping of datastore artifact to existence. Updated by
1598 this method with details of all artifacts tested. Can be `None`
1599 if the caller is not interested.
1601 Returns
1602 -------
1603 existence : `dict` of [`DatasetRef`, `bool`]
1604 Mapping from dataset to boolean indicating existence.
1605 """
1606 # Make a mapping from refs with the internal storage class to the given
1607 # refs that may have a different one. We'll use the internal refs
1608 # throughout this method and convert back at the very end.
1609 internal_ref_to_input_ref = {self._cast_storage_class(ref): ref for ref in refs}
1611 # Need a mapping of dataset_id to (internal) dataset ref since some
1612 # internal APIs work with dataset_id.
1613 id_to_ref = {ref.id: ref for ref in internal_ref_to_input_ref}
1615 # Set of all IDs we are checking for.
1616 requested_ids = set(id_to_ref.keys())
1618 # The records themselves. Could be missing some entries.
1619 records = self._get_stored_records_associated_with_refs(id_to_ref.values())
1621 dataset_existence = self._process_mexists_records(
1622 id_to_ref, records, True, artifact_existence=artifact_existence
1623 )
1625 # Set of IDs that have been handled.
1626 handled_ids = {ref.id for ref in dataset_existence.keys()}
1628 missing_ids = requested_ids - handled_ids
1629 if missing_ids:
1630 dataset_existence.update(
1631 self._mexists_check_expected(
1632 [id_to_ref[missing] for missing in missing_ids], artifact_existence
1633 )
1634 )
1636 return {
1637 internal_ref_to_input_ref[internal_ref]: existence
1638 for internal_ref, existence in dataset_existence.items()
1639 }
1641 def _mexists_check_expected(
1642 self, refs: Sequence[DatasetRef], artifact_existence: Optional[Dict[ResourcePath, bool]] = None
1643 ) -> Dict[DatasetRef, bool]:
1644 """Check existence of refs that are not known to datastore.
1646 Parameters
1647 ----------
1648 refs : iterable of `DatasetRef`
1649 The datasets to be checked. These are assumed not to be known
1650 to datastore.
1651 artifact_existence : `dict` [`lsst.resources.ResourcePath`, `bool`]
1652 Optional mapping of datastore artifact to existence. Updated by
1653 this method with details of all artifacts tested. Can be `None`
1654 if the caller is not interested.
1656 Returns
1657 -------
1658 existence : `dict` of [`DatasetRef`, `bool`]
1659 Mapping from dataset to boolean indicating existence.
1660 """
1661 dataset_existence: Dict[DatasetRef, bool] = {}
1662 if not self.trustGetRequest:
1663 # Must assume these do not exist
1664 for ref in refs:
1665 dataset_existence[ref] = False
1666 else:
1667 log.debug(
1668 "%d datasets were not known to datastore during initial existence check.",
1669 len(refs),
1670 )
1672 # Construct data structure identical to that returned
1673 # by _get_stored_records_associated_with_refs() but using
1674 # guessed names.
1675 records = {}
1676 id_to_ref = {}
1677 for missing_ref in refs:
1678 expected = self._get_expected_dataset_locations_info(missing_ref)
1679 dataset_id = missing_ref.id
1680 records[dataset_id] = [info for _, info in expected]
1681 id_to_ref[dataset_id] = missing_ref
1683 dataset_existence.update(
1684 self._process_mexists_records(
1685 id_to_ref,
1686 records,
1687 False,
1688 artifact_existence=artifact_existence,
1689 )
1690 )
1692 return dataset_existence
1694 def exists(self, ref: DatasetRef) -> bool:
1695 """Check if the dataset exists in the datastore.
1697 Parameters
1698 ----------
1699 ref : `DatasetRef`
1700 Reference to the required dataset.
1702 Returns
1703 -------
1704 exists : `bool`
1705 `True` if the entity exists in the `Datastore`.
1707 Notes
1708 -----
1709 The local cache is checked as a proxy for existence in the remote
1710 object store. It is possible that another process on a different
1711 compute node could remove the file from the object store even
1712 though it is present in the local cache.
1713 """
1714 ref = self._cast_storage_class(ref)
1715 fileLocations = self._get_dataset_locations_info(ref)
1717 # if we are being asked to trust that registry might not be correct
1718 # we ask for the expected locations and check them explicitly
1719 if not fileLocations:
1720 if not self.trustGetRequest:
1721 return False
1723 # First check the cache. If it is not found we must check
1724 # the datastore itself. Assume that any component in the cache
1725 # means that the dataset does exist somewhere.
1726 if self.cacheManager.known_to_cache(ref):
1727 return True
1729 # When we are guessing a dataset location we can not check
1730 # for the existence of every component since we can not
1731 # know if every component was written. Instead we check
1732 # for the existence of any of the expected locations.
1733 for location, _ in self._get_expected_dataset_locations_info(ref):
1734 if self._artifact_exists(location):
1735 return True
1736 return False
1738 # All listed artifacts must exist.
1739 for location, storedFileInfo in fileLocations:
1740 # Checking in cache needs the component ref.
1741 check_ref = ref
1742 if not ref.datasetType.isComponent() and (component := storedFileInfo.component):
1743 check_ref = ref.makeComponentRef(component)
1744 if self.cacheManager.known_to_cache(check_ref, location.getExtension()):
1745 continue
1747 if not self._artifact_exists(location):
1748 return False
1750 return True
1752 def getURIs(self, ref: DatasetRef, predict: bool = False) -> DatasetRefURIs:
1753 """Return URIs associated with dataset.
1755 Parameters
1756 ----------
1757 ref : `DatasetRef`
1758 Reference to the required dataset.
1759 predict : `bool`, optional
1760 If the datastore does not know about the dataset, should it
1761 return a predicted URI or not?
1763 Returns
1764 -------
1765 uris : `DatasetRefURIs`
1766 The URI to the primary artifact associated with this dataset (if
1767 the dataset was disassembled within the datastore this may be
1768 `None`), and the URIs to any components associated with the dataset
1769 artifact. (can be empty if there are no components).
1770 """
1771 many = self.getManyURIs([ref], predict=predict, allow_missing=False)
1772 return many[ref]
1774 def getURI(self, ref: DatasetRef, predict: bool = False) -> ResourcePath:
1775 """URI to the Dataset.
1777 Parameters
1778 ----------
1779 ref : `DatasetRef`
1780 Reference to the required Dataset.
1781 predict : `bool`
1782 If `True`, allow URIs to be returned of datasets that have not
1783 been written.
1785 Returns
1786 -------
1787 uri : `str`
1788 URI pointing to the dataset within the datastore. If the
1789 dataset does not exist in the datastore, and if ``predict`` is
1790 `True`, the URI will be a prediction and will include a URI
1791 fragment "#predicted".
1792 If the datastore does not have entities that relate well
1793 to the concept of a URI the returned URI will be
1794 descriptive. The returned URI is not guaranteed to be obtainable.
1796 Raises
1797 ------
1798 FileNotFoundError
1799 Raised if a URI has been requested for a dataset that does not
1800 exist and guessing is not allowed.
1801 RuntimeError
1802 Raised if a request is made for a single URI but multiple URIs
1803 are associated with this dataset.
1805 Notes
1806 -----
1807 When a predicted URI is requested an attempt will be made to form
1808 a reasonable URI based on file templates and the expected formatter.
1809 """
1810 primary, components = self.getURIs(ref, predict)
1811 if primary is None or components:
1812 raise RuntimeError(
1813 f"Dataset ({ref}) includes distinct URIs for components. Use Datastore.getURIs() instead."
1814 )
1815 return primary
1817 def _predict_URIs(
1818 self,
1819 ref: DatasetRef,
1820 ) -> DatasetRefURIs:
1821 """Predict the URIs of a dataset ref.
1823 Parameters
1824 ----------
1825 ref : `DatasetRef`
1826 Reference to the required Dataset.
1828 Returns
1829 -------
1830 URI : DatasetRefUris
1831 Primary and component URIs. URIs will contain a URI fragment
1832 "#predicted".
1833 """
1834 uris = DatasetRefURIs()
1836 if self.composites.shouldBeDisassembled(ref):
1837 for component, _ in ref.datasetType.storageClass.components.items():
1838 comp_ref = ref.makeComponentRef(component)
1839 comp_location, _ = self._determine_put_formatter_location(comp_ref)
1841 # Add the "#predicted" URI fragment to indicate this is a
1842 # guess
1843 uris.componentURIs[component] = ResourcePath(comp_location.uri.geturl() + "#predicted")
1845 else:
1846 location, _ = self._determine_put_formatter_location(ref)
1848 # Add the "#predicted" URI fragment to indicate this is a guess
1849 uris.primaryURI = ResourcePath(location.uri.geturl() + "#predicted")
1851 return uris
1853 def getManyURIs(
1854 self,
1855 refs: Iterable[DatasetRef],
1856 predict: bool = False,
1857 allow_missing: bool = False,
1858 ) -> Dict[DatasetRef, DatasetRefURIs]:
1859 # Docstring inherited
1861 uris: Dict[DatasetRef, DatasetRefURIs] = {}
1863 records = self._get_stored_records_associated_with_refs(refs)
1864 records_keys = records.keys()
1866 existing_refs = tuple(ref for ref in refs if ref.id in records_keys)
1867 missing_refs = tuple(ref for ref in refs if ref.id not in records_keys)
1869 # Have to handle trustGetRequest mode by checking for the existence
1870 # of the missing refs on disk.
1871 if missing_refs:
1872 dataset_existence = self._mexists_check_expected(missing_refs, None)
1873 really_missing = set()
1874 not_missing = set()
1875 for ref, exists in dataset_existence.items():
1876 if exists:
1877 not_missing.add(ref)
1878 else:
1879 really_missing.add(ref)
1881 if not_missing:
1882 # Need to recalculate the missing/existing split.
1883 existing_refs = existing_refs + tuple(not_missing)
1884 missing_refs = tuple(really_missing)
1886 for ref in missing_refs:
1887 # if this has never been written then we have to guess
1888 if not predict:
1889 if not allow_missing:
1890 raise FileNotFoundError("Dataset {} not in this datastore.".format(ref))
1891 else:
1892 uris[ref] = self._predict_URIs(ref)
1894 for ref in existing_refs:
1895 file_infos = records[ref.id]
1896 file_locations = [(i.file_location(self.locationFactory), i) for i in file_infos]
1897 uris[ref] = self._locations_to_URI(ref, file_locations)
1899 return uris
1901 def _locations_to_URI(
1902 self,
1903 ref: DatasetRef,
1904 file_locations: Sequence[Tuple[Location, StoredFileInfo]],
1905 ) -> DatasetRefURIs:
1906 """Convert one or more file locations associated with a DatasetRef
1907 to a DatasetRefURIs.
1909 Parameters
1910 ----------
1911 ref : `DatasetRef`
1912 Reference to the dataset.
1913 file_locations : Sequence[Tuple[Location, StoredFileInfo]]
1914 Each item in the sequence is the location of the dataset within the
1915 datastore and stored information about the file and its formatter.
1916 If there is only one item in the sequence then it is treated as the
1917 primary URI. If there is more than one item then they are treated
1918 as component URIs. If there are no items then an error is raised
1919 unless ``self.trustGetRequest`` is `True`.
1921 Returns
1922 -------
1923 uris: DatasetRefURIs
1924 Represents the primary URI or component URIs described by the
1925 inputs.
1927 Raises
1928 ------
1929 RuntimeError
1930 If no file locations are passed in and ``self.trustGetRequest`` is
1931 `False`.
1932 FileNotFoundError
1933 If the a passed-in URI does not exist, and ``self.trustGetRequest``
1934 is `False`.
1935 RuntimeError
1936 If a passed in `StoredFileInfo`'s ``component`` is `None` (this is
1937 unexpected).
1938 """
1940 guessing = False
1941 uris = DatasetRefURIs()
1943 if not file_locations:
1944 if not self.trustGetRequest:
1945 raise RuntimeError(f"Unexpectedly got no artifacts for dataset {ref}")
1946 file_locations = self._get_expected_dataset_locations_info(ref)
1947 guessing = True
1949 if len(file_locations) == 1:
1950 # No disassembly so this is the primary URI
1951 uris.primaryURI = file_locations[0][0].uri
1952 if guessing and not uris.primaryURI.exists():
1953 raise FileNotFoundError(f"Expected URI ({uris.primaryURI}) does not exist")
1954 else:
1955 for location, file_info in file_locations:
1956 if file_info.component is None:
1957 raise RuntimeError(f"Unexpectedly got no component name for a component at {location}")
1958 if guessing and not location.uri.exists():
1959 # If we are trusting then it is entirely possible for
1960 # some components to be missing. In that case we skip
1961 # to the next component.
1962 if self.trustGetRequest:
1963 continue
1964 raise FileNotFoundError(f"Expected URI ({location.uri}) does not exist")
1965 uris.componentURIs[file_info.component] = location.uri
1967 return uris
1969 def retrieveArtifacts(
1970 self,
1971 refs: Iterable[DatasetRef],
1972 destination: ResourcePath,
1973 transfer: str = "auto",
1974 preserve_path: bool = True,
1975 overwrite: bool = False,
1976 ) -> List[ResourcePath]:
1977 """Retrieve the file artifacts associated with the supplied refs.
1979 Parameters
1980 ----------
1981 refs : iterable of `DatasetRef`
1982 The datasets for which file artifacts are to be retrieved.
1983 A single ref can result in multiple files. The refs must
1984 be resolved.
1985 destination : `lsst.resources.ResourcePath`
1986 Location to write the file artifacts.
1987 transfer : `str`, optional
1988 Method to use to transfer the artifacts. Must be one of the options
1989 supported by `lsst.resources.ResourcePath.transfer_from()`.
1990 "move" is not allowed.
1991 preserve_path : `bool`, optional
1992 If `True` the full path of the file artifact within the datastore
1993 is preserved. If `False` the final file component of the path
1994 is used.
1995 overwrite : `bool`, optional
1996 If `True` allow transfers to overwrite existing files at the
1997 destination.
1999 Returns
2000 -------
2001 targets : `list` of `lsst.resources.ResourcePath`
2002 URIs of file artifacts in destination location. Order is not
2003 preserved.
2004 """
2005 if not destination.isdir():
2006 raise ValueError(f"Destination location must refer to a directory. Given {destination}")
2008 if transfer == "move":
2009 raise ValueError("Can not move artifacts out of datastore. Use copy instead.")
2011 # Source -> Destination
2012 # This also helps filter out duplicate DatasetRef in the request
2013 # that will map to the same underlying file transfer.
2014 to_transfer: Dict[ResourcePath, ResourcePath] = {}
2016 for ref in refs:
2017 locations = self._get_dataset_locations_info(ref)
2018 for location, _ in locations:
2019 source_uri = location.uri
2020 target_path: ResourcePathExpression
2021 if preserve_path:
2022 target_path = location.pathInStore
2023 if target_path.isabs():
2024 # This is an absolute path to an external file.
2025 # Use the full path.
2026 target_path = target_path.relativeToPathRoot
2027 else:
2028 target_path = source_uri.basename()
2029 target_uri = destination.join(target_path)
2030 to_transfer[source_uri] = target_uri
2032 # In theory can now parallelize the transfer
2033 log.debug("Number of artifacts to transfer to %s: %d", str(destination), len(to_transfer))
2034 for source_uri, target_uri in to_transfer.items():
2035 target_uri.transfer_from(source_uri, transfer=transfer, overwrite=overwrite)
2037 return list(to_transfer.values())
2039 def get(
2040 self,
2041 ref: DatasetRef,
2042 parameters: Optional[Mapping[str, Any]] = None,
2043 storageClass: Optional[Union[StorageClass, str]] = None,
2044 ) -> Any:
2045 """Load an InMemoryDataset from the store.
2047 Parameters
2048 ----------
2049 ref : `DatasetRef`
2050 Reference to the required Dataset.
2051 parameters : `dict`
2052 `StorageClass`-specific parameters that specify, for example,
2053 a slice of the dataset to be loaded.
2054 storageClass : `StorageClass` or `str`, optional
2055 The storage class to be used to override the Python type
2056 returned by this method. By default the returned type matches
2057 the dataset type definition for this dataset. Specifying a
2058 read `StorageClass` can force a different type to be returned.
2059 This type must be compatible with the original type.
2061 Returns
2062 -------
2063 inMemoryDataset : `object`
2064 Requested dataset or slice thereof as an InMemoryDataset.
2066 Raises
2067 ------
2068 FileNotFoundError
2069 Requested dataset can not be retrieved.
2070 TypeError
2071 Return value from formatter has unexpected type.
2072 ValueError
2073 Formatter failed to process the dataset.
2074 """
2075 # Supplied storage class for the component being read is either
2076 # from the ref itself or some an override if we want to force
2077 # type conversion.
2078 if storageClass is not None:
2079 ref = ref.overrideStorageClass(storageClass)
2080 refStorageClass = ref.datasetType.storageClass
2082 allGetInfo = self._prepare_for_get(ref, parameters)
2083 refComponent = ref.datasetType.component()
2085 # Create mapping from component name to related info
2086 allComponents = {i.component: i for i in allGetInfo}
2088 # By definition the dataset is disassembled if we have more
2089 # than one record for it.
2090 isDisassembled = len(allGetInfo) > 1
2092 # Look for the special case where we are disassembled but the
2093 # component is a derived component that was not written during
2094 # disassembly. For this scenario we need to check that the
2095 # component requested is listed as a derived component for the
2096 # composite storage class
2097 isDisassembledReadOnlyComponent = False
2098 if isDisassembled and refComponent:
2099 # The composite storage class should be accessible through
2100 # the component dataset type
2101 compositeStorageClass = ref.datasetType.parentStorageClass
2103 # In the unlikely scenario where the composite storage
2104 # class is not known, we can only assume that this is a
2105 # normal component. If that assumption is wrong then the
2106 # branch below that reads a persisted component will fail
2107 # so there is no need to complain here.
2108 if compositeStorageClass is not None:
2109 isDisassembledReadOnlyComponent = refComponent in compositeStorageClass.derivedComponents
2111 if isDisassembled and not refComponent:
2112 # This was a disassembled dataset spread over multiple files
2113 # and we need to put them all back together again.
2114 # Read into memory and then assemble
2116 # Check that the supplied parameters are suitable for the type read
2117 refStorageClass.validateParameters(parameters)
2119 # We want to keep track of all the parameters that were not used
2120 # by formatters. We assume that if any of the component formatters
2121 # use a parameter that we do not need to apply it again in the
2122 # assembler.
2123 usedParams = set()
2125 components: Dict[str, Any] = {}
2126 for getInfo in allGetInfo:
2127 # assemblerParams are parameters not understood by the
2128 # associated formatter.
2129 usedParams.update(set(getInfo.formatterParams))
2131 component = getInfo.component
2133 if component is None:
2134 raise RuntimeError(f"Internal error in datastore assembly of {ref}")
2136 # We do not want the formatter to think it's reading
2137 # a component though because it is really reading a
2138 # standalone dataset -- always tell reader it is not a
2139 # component.
2140 components[component] = self._read_artifact_into_memory(
2141 getInfo, ref.makeComponentRef(component), isComponent=False
2142 )
2144 inMemoryDataset = ref.datasetType.storageClass.delegate().assemble(components)
2146 # Any unused parameters will have to be passed to the assembler
2147 if parameters:
2148 unusedParams = {k: v for k, v in parameters.items() if k not in usedParams}
2149 else:
2150 unusedParams = {}
2152 # Process parameters
2153 return ref.datasetType.storageClass.delegate().handleParameters(
2154 inMemoryDataset, parameters=unusedParams
2155 )
2157 elif isDisassembledReadOnlyComponent:
2158 compositeStorageClass = ref.datasetType.parentStorageClass
2159 if compositeStorageClass is None:
2160 raise RuntimeError(
2161 f"Unable to retrieve derived component '{refComponent}' since"
2162 "no composite storage class is available."
2163 )
2165 if refComponent is None:
2166 # Mainly for mypy
2167 raise RuntimeError(f"Internal error in datastore {self.name}: component can not be None here")
2169 # Assume that every derived component can be calculated by
2170 # forwarding the request to a single read/write component.
2171 # Rather than guessing which rw component is the right one by
2172 # scanning each for a derived component of the same name,
2173 # we ask the storage class delegate directly which one is best to
2174 # use.
2175 compositeDelegate = compositeStorageClass.delegate()
2176 forwardedComponent = compositeDelegate.selectResponsibleComponent(
2177 refComponent, set(allComponents)
2178 )
2180 # Select the relevant component
2181 rwInfo = allComponents[forwardedComponent]
2183 # For now assume that read parameters are validated against
2184 # the real component and not the requested component
2185 forwardedStorageClass = rwInfo.formatter.fileDescriptor.readStorageClass
2186 forwardedStorageClass.validateParameters(parameters)
2188 # The reference to use for the caching must refer to the forwarded
2189 # component and not the derived component.
2190 cache_ref = ref.makeCompositeRef().makeComponentRef(forwardedComponent)
2192 # Unfortunately the FileDescriptor inside the formatter will have
2193 # the wrong write storage class so we need to create a new one
2194 # given the immutability constraint.
2195 writeStorageClass = rwInfo.info.storageClass
2197 # We may need to put some thought into parameters for read
2198 # components but for now forward them on as is
2199 readFormatter = type(rwInfo.formatter)(
2200 FileDescriptor(
2201 rwInfo.location,
2202 readStorageClass=refStorageClass,
2203 storageClass=writeStorageClass,
2204 parameters=parameters,
2205 ),
2206 ref.dataId,
2207 )
2209 # The assembler can not receive any parameter requests for a
2210 # derived component at this time since the assembler will
2211 # see the storage class of the derived component and those
2212 # parameters will have to be handled by the formatter on the
2213 # forwarded storage class.
2214 assemblerParams: Dict[str, Any] = {}
2216 # Need to created a new info that specifies the derived
2217 # component and associated storage class
2218 readInfo = DatastoreFileGetInformation(
2219 rwInfo.location,
2220 readFormatter,
2221 rwInfo.info,
2222 assemblerParams,
2223 {},
2224 refComponent,
2225 refStorageClass,
2226 )
2228 return self._read_artifact_into_memory(readInfo, ref, isComponent=True, cache_ref=cache_ref)
2230 else:
2231 # Single file request or component from that composite file
2232 for lookup in (refComponent, None):
2233 if lookup in allComponents:
2234 getInfo = allComponents[lookup]
2235 break
2236 else:
2237 raise FileNotFoundError(
2238 f"Component {refComponent} not found for ref {ref} in datastore {self.name}"
2239 )
2241 # Do not need the component itself if already disassembled
2242 if isDisassembled:
2243 isComponent = False
2244 else:
2245 isComponent = getInfo.component is not None
2247 # For a component read of a composite we want the cache to
2248 # be looking at the composite ref itself.
2249 cache_ref = ref.makeCompositeRef() if isComponent else ref
2251 # For a disassembled component we can validate parametersagainst
2252 # the component storage class directly
2253 if isDisassembled:
2254 refStorageClass.validateParameters(parameters)
2255 else:
2256 # For an assembled composite this could be a derived
2257 # component derived from a real component. The validity
2258 # of the parameters is not clear. For now validate against
2259 # the composite storage class
2260 getInfo.formatter.fileDescriptor.storageClass.validateParameters(parameters)
2262 return self._read_artifact_into_memory(getInfo, ref, isComponent=isComponent, cache_ref=cache_ref)
2264 @transactional
2265 def put(self, inMemoryDataset: Any, ref: DatasetRef) -> None:
2266 """Write a InMemoryDataset with a given `DatasetRef` to the store.
2268 Parameters
2269 ----------
2270 inMemoryDataset : `object`
2271 The dataset to store.
2272 ref : `DatasetRef`
2273 Reference to the associated Dataset.
2275 Raises
2276 ------
2277 TypeError
2278 Supplied object and storage class are inconsistent.
2279 DatasetTypeNotSupportedError
2280 The associated `DatasetType` is not handled by this datastore.
2282 Notes
2283 -----
2284 If the datastore is configured to reject certain dataset types it
2285 is possible that the put will fail and raise a
2286 `DatasetTypeNotSupportedError`. The main use case for this is to
2287 allow `ChainedDatastore` to put to multiple datastores without
2288 requiring that every datastore accepts the dataset.
2289 """
2291 doDisassembly = self.composites.shouldBeDisassembled(ref)
2292 # doDisassembly = True
2294 artifacts = []
2295 if doDisassembly:
2296 components = ref.datasetType.storageClass.delegate().disassemble(inMemoryDataset)
2297 if components is None:
2298 raise RuntimeError(
2299 f"Inconsistent configuration: dataset type {ref.datasetType.name} "
2300 f"with storage class {ref.datasetType.storageClass.name} "
2301 "is configured to be disassembled, but cannot be."
2302 )
2303 for component, componentInfo in components.items():
2304 # Don't recurse because we want to take advantage of
2305 # bulk insert -- need a new DatasetRef that refers to the
2306 # same dataset_id but has the component DatasetType
2307 # DatasetType does not refer to the types of components
2308 # So we construct one ourselves.
2309 compRef = ref.makeComponentRef(component)
2310 storedInfo = self._write_in_memory_to_artifact(componentInfo.component, compRef)
2311 artifacts.append((compRef, storedInfo))
2312 else:
2313 # Write the entire thing out
2314 storedInfo = self._write_in_memory_to_artifact(inMemoryDataset, ref)
2315 artifacts.append((ref, storedInfo))
2317 self._register_datasets(artifacts)
2319 @transactional
2320 def trash(self, ref: Union[DatasetRef, Iterable[DatasetRef]], ignore_errors: bool = True) -> None:
2321 # At this point can safely remove these datasets from the cache
2322 # to avoid confusion later on. If they are not trashed later
2323 # the cache will simply be refilled.
2324 self.cacheManager.remove_from_cache(ref)
2326 # If we are in trust mode there will be nothing to move to
2327 # the trash table and we will have to try to delete the file
2328 # immediately.
2329 if self.trustGetRequest:
2330 # Try to keep the logic below for a single file trash.
2331 if isinstance(ref, DatasetRef):
2332 refs = {ref}
2333 else:
2334 # Will recreate ref at the end of this branch.
2335 refs = set(ref)
2337 # Determine which datasets are known to datastore directly.
2338 id_to_ref = {ref.id: ref for ref in refs}
2339 existing_ids = self._get_stored_records_associated_with_refs(refs)
2340 existing_refs = {id_to_ref[ref_id] for ref_id in existing_ids}
2342 missing = refs - existing_refs
2343 if missing:
2344 # Do an explicit existence check on these refs.
2345 # We only care about the artifacts at this point and not
2346 # the dataset existence.
2347 artifact_existence: Dict[ResourcePath, bool] = {}
2348 _ = self.mexists(missing, artifact_existence)
2349 uris = [uri for uri, exists in artifact_existence.items() if exists]
2351 # FUTURE UPGRADE: Implement a parallelized bulk remove.
2352 log.debug("Removing %d artifacts from datastore that are unknown to datastore", len(uris))
2353 for uri in uris:
2354 try:
2355 uri.remove()
2356 except Exception as e:
2357 if ignore_errors:
2358 log.debug("Artifact %s could not be removed: %s", uri, e)
2359 continue
2360 raise
2362 # There is no point asking the code below to remove refs we
2363 # know are missing so update it with the list of existing
2364 # records. Try to retain one vs many logic.
2365 if not existing_refs:
2366 # Nothing more to do since none of the datasets were
2367 # known to the datastore record table.
2368 return
2369 ref = list(existing_refs)
2370 if len(ref) == 1:
2371 ref = ref[0]
2373 # Get file metadata and internal metadata
2374 if not isinstance(ref, DatasetRef):
2375 log.debug("Doing multi-dataset trash in datastore %s", self.name)
2376 # Assumed to be an iterable of refs so bulk mode enabled.
2377 try:
2378 self.bridge.moveToTrash(ref, transaction=self._transaction)
2379 except Exception as e:
2380 if ignore_errors:
2381 log.warning("Unexpected issue moving multiple datasets to trash: %s", e)
2382 else:
2383 raise
2384 return
2386 log.debug("Trashing dataset %s in datastore %s", ref, self.name)
2388 fileLocations = self._get_dataset_locations_info(ref)
2390 if not fileLocations:
2391 err_msg = f"Requested dataset to trash ({ref}) is not known to datastore {self.name}"
2392 if ignore_errors:
2393 log.warning(err_msg)
2394 return
2395 else:
2396 raise FileNotFoundError(err_msg)
2398 for location, storedFileInfo in fileLocations:
2399 if not self._artifact_exists(location):
2400 err_msg = (
2401 f"Dataset is known to datastore {self.name} but "
2402 f"associated artifact ({location.uri}) is missing"
2403 )
2404 if ignore_errors:
2405 log.warning(err_msg)
2406 return
2407 else:
2408 raise FileNotFoundError(err_msg)
2410 # Mark dataset as trashed
2411 try:
2412 self.bridge.moveToTrash([ref], transaction=self._transaction)
2413 except Exception as e:
2414 if ignore_errors:
2415 log.warning(
2416 "Attempted to mark dataset (%s) to be trashed in datastore %s "
2417 "but encountered an error: %s",
2418 ref,
2419 self.name,
2420 e,
2421 )
2422 pass
2423 else:
2424 raise
2426 @transactional
2427 def emptyTrash(self, ignore_errors: bool = True) -> None:
2428 """Remove all datasets from the trash.
2430 Parameters
2431 ----------
2432 ignore_errors : `bool`
2433 If `True` return without error even if something went wrong.
2434 Problems could occur if another process is simultaneously trying
2435 to delete.
2436 """
2437 log.debug("Emptying trash in datastore %s", self.name)
2439 # Context manager will empty trash iff we finish it without raising.
2440 # It will also automatically delete the relevant rows from the
2441 # trash table and the records table.
2442 with self.bridge.emptyTrash(
2443 self._table, record_class=StoredFileInfo, record_column="path"
2444 ) as trash_data:
2445 # Removing the artifacts themselves requires that the files are
2446 # not also associated with refs that are not to be trashed.
2447 # Therefore need to do a query with the file paths themselves
2448 # and return all the refs associated with them. Can only delete
2449 # a file if the refs to be trashed are the only refs associated
2450 # with the file.
2451 # This requires multiple copies of the trashed items
2452 trashed, artifacts_to_keep = trash_data
2454 if artifacts_to_keep is None:
2455 # The bridge is not helping us so have to work it out
2456 # ourselves. This is not going to be as efficient.
2457 trashed = list(trashed)
2459 # The instance check is for mypy since up to this point it
2460 # does not know the type of info.
2461 path_map = self._refs_associated_with_artifacts(
2462 [info.path for _, info in trashed if isinstance(info, StoredFileInfo)]
2463 )
2465 for ref, info in trashed:
2466 # Mypy needs to know this is not the base class
2467 assert isinstance(info, StoredFileInfo), f"Unexpectedly got info of class {type(info)}"
2469 path_map[info.path].remove(ref.id)
2470 if not path_map[info.path]:
2471 del path_map[info.path]
2473 artifacts_to_keep = set(path_map)
2475 for ref, info in trashed:
2476 # Should not happen for this implementation but need
2477 # to keep mypy happy.
2478 assert info is not None, f"Internal logic error in emptyTrash with ref {ref}."
2480 # Mypy needs to know this is not the base class
2481 assert isinstance(info, StoredFileInfo), f"Unexpectedly got info of class {type(info)}"
2483 if info.path in artifacts_to_keep:
2484 # This is a multi-dataset artifact and we are not
2485 # removing all associated refs.
2486 continue
2488 # Only trashed refs still known to datastore will be returned.
2489 location = info.file_location(self.locationFactory)
2491 # Point of no return for this artifact
2492 log.debug("Removing artifact %s from datastore %s", location.uri, self.name)
2493 try:
2494 self._delete_artifact(location)
2495 except FileNotFoundError:
2496 # If the file itself has been deleted there is nothing
2497 # we can do about it. It is possible that trash has
2498 # been run in parallel in another process or someone
2499 # decided to delete the file. It is unlikely to come
2500 # back and so we should still continue with the removal
2501 # of the entry from the trash table. It is also possible
2502 # we removed it in a previous iteration if it was
2503 # a multi-dataset artifact. The delete artifact method
2504 # will log a debug message in this scenario.
2505 # Distinguishing file missing before trash started and
2506 # file already removed previously as part of this trash
2507 # is not worth the distinction with regards to potential
2508 # memory cost.
2509 pass
2510 except Exception as e:
2511 if ignore_errors:
2512 # Use a debug message here even though it's not
2513 # a good situation. In some cases this can be
2514 # caused by a race between user A and user B
2515 # and neither of them has permissions for the
2516 # other's files. Butler does not know about users
2517 # and trash has no idea what collections these
2518 # files were in (without guessing from a path).
2519 log.debug(
2520 "Encountered error removing artifact %s from datastore %s: %s",
2521 location.uri,
2522 self.name,
2523 e,
2524 )
2525 else:
2526 raise
2528 @transactional
2529 def transfer_from(
2530 self,
2531 source_datastore: Datastore,
2532 refs: Iterable[DatasetRef],
2533 transfer: str = "auto",
2534 artifact_existence: Optional[Dict[ResourcePath, bool]] = None,
2535 ) -> tuple[set[DatasetRef], set[DatasetRef]]:
2536 # Docstring inherited
2537 if type(self) is not type(source_datastore):
2538 raise TypeError(
2539 f"Datastore mismatch between this datastore ({type(self)}) and the "
2540 f"source datastore ({type(source_datastore)})."
2541 )
2543 # Be explicit for mypy
2544 if not isinstance(source_datastore, FileDatastore):
2545 raise TypeError(
2546 "Can only transfer to a FileDatastore from another FileDatastore, not"
2547 f" {type(source_datastore)}"
2548 )
2550 # Stop early if "direct" transfer mode is requested. That would
2551 # require that the URI inside the source datastore should be stored
2552 # directly in the target datastore, which seems unlikely to be useful
2553 # since at any moment the source datastore could delete the file.
2554 if transfer in ("direct", "split"):
2555 raise ValueError(
2556 f"Can not transfer from a source datastore using {transfer} mode since"
2557 " those files are controlled by the other datastore."
2558 )
2560 # Empty existence lookup if none given.
2561 if artifact_existence is None:
2562 artifact_existence = {}
2564 # We will go through the list multiple times so must convert
2565 # generators to lists.
2566 refs = list(refs)
2568 # In order to handle disassembled composites the code works
2569 # at the records level since it can assume that internal APIs
2570 # can be used.
2571 # - If the record already exists in the destination this is assumed
2572 # to be okay.
2573 # - If there is no record but the source and destination URIs are
2574 # identical no transfer is done but the record is added.
2575 # - If the source record refers to an absolute URI currently assume
2576 # that that URI should remain absolute and will be visible to the
2577 # destination butler. May need to have a flag to indicate whether
2578 # the dataset should be transferred. This will only happen if
2579 # the detached Butler has had a local ingest.
2581 # What we really want is all the records in the source datastore
2582 # associated with these refs. Or derived ones if they don't exist
2583 # in the source.
2584 source_records = source_datastore._get_stored_records_associated_with_refs(refs)
2586 # The source dataset_ids are the keys in these records
2587 source_ids = set(source_records)
2588 log.debug("Number of datastore records found in source: %d", len(source_ids))
2590 requested_ids = set(ref.id for ref in refs)
2591 missing_ids = requested_ids - source_ids
2593 # Missing IDs can be okay if that datastore has allowed
2594 # gets based on file existence. Should we transfer what we can
2595 # or complain about it and warn?
2596 if missing_ids and not source_datastore.trustGetRequest:
2597 raise ValueError(
2598 f"Some datasets are missing from source datastore {source_datastore}: {missing_ids}"
2599 )
2601 # Need to map these missing IDs to a DatasetRef so we can guess
2602 # the details.
2603 if missing_ids:
2604 log.info(
2605 "Number of expected datasets missing from source datastore records: %d out of %d",
2606 len(missing_ids),
2607 len(requested_ids),
2608 )
2609 id_to_ref = {ref.id: ref for ref in refs if ref.id in missing_ids}
2611 # This should be chunked in case we end up having to check
2612 # the file store since we need some log output to show
2613 # progress.
2614 for missing_ids_chunk in chunk_iterable(missing_ids, chunk_size=10_000):
2615 records = {}
2616 for missing in missing_ids_chunk:
2617 # Ask the source datastore where the missing artifacts
2618 # should be. An execution butler might not know about the
2619 # artifacts even if they are there.
2620 expected = source_datastore._get_expected_dataset_locations_info(id_to_ref[missing])
2621 records[missing] = [info for _, info in expected]
2623 # Call the mexist helper method in case we have not already
2624 # checked these artifacts such that artifact_existence is
2625 # empty. This allows us to benefit from parallelism.
2626 # datastore.mexists() itself does not give us access to the
2627 # derived datastore record.
2628 log.verbose("Checking existence of %d datasets unknown to datastore", len(records))
2629 ref_exists = source_datastore._process_mexists_records(
2630 id_to_ref, records, False, artifact_existence=artifact_existence
2631 )
2633 # Now go through the records and propagate the ones that exist.
2634 location_factory = source_datastore.locationFactory
2635 for missing, record_list in records.items():
2636 # Skip completely if the ref does not exist.
2637 ref = id_to_ref[missing]
2638 if not ref_exists[ref]:
2639 log.warning("Asked to transfer dataset %s but no file artifacts exist for it.", ref)
2640 continue
2641 # Check for file artifact to decide which parts of a
2642 # disassembled composite do exist. If there is only a
2643 # single record we don't even need to look because it can't
2644 # be a composite and must exist.
2645 if len(record_list) == 1:
2646 dataset_records = record_list
2647 else:
2648 dataset_records = [
2649 record
2650 for record in record_list
2651 if artifact_existence[record.file_location(location_factory).uri]
2652 ]
2653 assert len(dataset_records) > 0, "Disassembled composite should have had some files."
2655 # Rely on source_records being a defaultdict.
2656 source_records[missing].extend(dataset_records)
2658 # See if we already have these records
2659 target_records = self._get_stored_records_associated_with_refs(refs)
2661 # The artifacts to register
2662 artifacts = []
2664 # Refs that already exist
2665 already_present = []
2667 # Refs that were rejected by this datastore.
2668 rejected = set()
2670 # Refs that were transferred successfully.
2671 accepted = set()
2673 # Record each time we have done a "direct" transfer.
2674 direct_transfers = []
2676 # Now can transfer the artifacts
2677 for ref in refs:
2678 if not self.constraints.isAcceptable(ref):
2679 # This datastore should not be accepting this dataset.
2680 rejected.add(ref)
2681 continue
2683 accepted.add(ref)
2685 if ref.id in target_records:
2686 # Already have an artifact for this.
2687 already_present.append(ref)
2688 continue
2690 # mypy needs to know these are always resolved refs
2691 for info in source_records[ref.id]:
2692 source_location = info.file_location(source_datastore.locationFactory)
2693 target_location = info.file_location(self.locationFactory)
2694 if source_location == target_location and not source_location.pathInStore.isabs():
2695 # Artifact is already in the target location.
2696 # (which is how execution butler currently runs)
2697 pass
2698 else:
2699 if target_location.pathInStore.isabs():
2700 # Just because we can see the artifact when running
2701 # the transfer doesn't mean it will be generally
2702 # accessible to a user of this butler. Need to decide
2703 # what to do about an absolute path.
2704 if transfer == "auto":
2705 # For "auto" transfers we allow the absolute URI
2706 # to be recorded in the target datastore.
2707 direct_transfers.append(source_location)
2708 else:
2709 # The user is explicitly requesting a transfer
2710 # even for an absolute URI. This requires us to
2711 # calculate the target path.
2712 template_ref = ref
2713 if info.component:
2714 template_ref = ref.makeComponentRef(info.component)
2715 target_location = self._calculate_ingested_datastore_name(
2716 source_location.uri,
2717 template_ref,
2718 )
2720 info = info.update(path=target_location.pathInStore.path)
2722 # Need to transfer it to the new location.
2723 # Assume we should always overwrite. If the artifact
2724 # is there this might indicate that a previous transfer
2725 # was interrupted but was not able to be rolled back
2726 # completely (eg pre-emption) so follow Datastore default
2727 # and overwrite.
2728 target_location.uri.transfer_from(
2729 source_location.uri, transfer=transfer, overwrite=True, transaction=self._transaction
2730 )
2732 artifacts.append((ref, info))
2734 if direct_transfers:
2735 log.info(
2736 "Transfer request for an outside-datastore artifact with absolute URI done %d time%s",
2737 len(direct_transfers),
2738 "" if len(direct_transfers) == 1 else "s",
2739 )
2741 self._register_datasets(artifacts)
2743 if already_present:
2744 n_skipped = len(already_present)
2745 log.info(
2746 "Skipped transfer of %d dataset%s already present in datastore",
2747 n_skipped,
2748 "" if n_skipped == 1 else "s",
2749 )
2751 return accepted, rejected
2753 @transactional
2754 def forget(self, refs: Iterable[DatasetRef]) -> None:
2755 # Docstring inherited.
2756 refs = list(refs)
2757 self.bridge.forget(refs)
2758 self._table.delete(["dataset_id"], *[{"dataset_id": ref.id} for ref in refs])
2760 def validateConfiguration(
2761 self, entities: Iterable[Union[DatasetRef, DatasetType, StorageClass]], logFailures: bool = False
2762 ) -> None:
2763 """Validate some of the configuration for this datastore.
2765 Parameters
2766 ----------
2767 entities : iterable of `DatasetRef`, `DatasetType`, or `StorageClass`
2768 Entities to test against this configuration. Can be differing
2769 types.
2770 logFailures : `bool`, optional
2771 If `True`, output a log message for every validation error
2772 detected.
2774 Raises
2775 ------
2776 DatastoreValidationError
2777 Raised if there is a validation problem with a configuration.
2778 All the problems are reported in a single exception.
2780 Notes
2781 -----
2782 This method checks that all the supplied entities have valid file
2783 templates and also have formatters defined.
2784 """
2786 templateFailed = None
2787 try:
2788 self.templates.validateTemplates(entities, logFailures=logFailures)
2789 except FileTemplateValidationError as e:
2790 templateFailed = str(e)
2792 formatterFailed = []
2793 for entity in entities:
2794 try:
2795 self.formatterFactory.getFormatterClass(entity)
2796 except KeyError as e:
2797 formatterFailed.append(str(e))
2798 if logFailures:
2799 log.critical("Formatter failure: %s", e)
2801 if templateFailed or formatterFailed:
2802 messages = []
2803 if templateFailed:
2804 messages.append(templateFailed)
2805 if formatterFailed:
2806 messages.append(",".join(formatterFailed))
2807 msg = ";\n".join(messages)
2808 raise DatastoreValidationError(msg)
2810 def getLookupKeys(self) -> Set[LookupKey]:
2811 # Docstring is inherited from base class
2812 return (
2813 self.templates.getLookupKeys()
2814 | self.formatterFactory.getLookupKeys()
2815 | self.constraints.getLookupKeys()
2816 )
2818 def validateKey(self, lookupKey: LookupKey, entity: Union[DatasetRef, DatasetType, StorageClass]) -> None:
2819 # Docstring is inherited from base class
2820 # The key can be valid in either formatters or templates so we can
2821 # only check the template if it exists
2822 if lookupKey in self.templates:
2823 try:
2824 self.templates[lookupKey].validateTemplate(entity)
2825 except FileTemplateValidationError as e:
2826 raise DatastoreValidationError(e) from e
2828 def export(
2829 self,
2830 refs: Iterable[DatasetRef],
2831 *,
2832 directory: Optional[ResourcePathExpression] = None,
2833 transfer: Optional[str] = "auto",
2834 ) -> Iterable[FileDataset]:
2835 # Docstring inherited from Datastore.export.
2836 if transfer == "auto" and directory is None:
2837 transfer = None
2839 if transfer is not None and directory is None:
2840 raise TypeError(f"Cannot export using transfer mode {transfer} with no export directory given")
2842 if transfer == "move":
2843 raise TypeError("Can not export by moving files out of datastore.")
2844 elif transfer == "direct":
2845 # For an export, treat this as equivalent to None. We do not
2846 # want an import to risk using absolute URIs to datasets owned
2847 # by another datastore.
2848 log.info("Treating 'direct' transfer mode as in-place export.")
2849 transfer = None
2851 # Force the directory to be a URI object
2852 directoryUri: Optional[ResourcePath] = None
2853 if directory is not None:
2854 directoryUri = ResourcePath(directory, forceDirectory=True)
2856 if transfer is not None and directoryUri is not None:
2857 # mypy needs the second test
2858 if not directoryUri.exists():
2859 raise FileNotFoundError(f"Export location {directory} does not exist")
2861 progress = Progress("lsst.daf.butler.datastores.FileDatastore.export", level=logging.DEBUG)
2862 for ref in progress.wrap(refs, "Exporting dataset files"):
2863 fileLocations = self._get_dataset_locations_info(ref)
2864 if not fileLocations:
2865 raise FileNotFoundError(f"Could not retrieve dataset {ref}.")
2866 # For now we can not export disassembled datasets
2867 if len(fileLocations) > 1:
2868 raise NotImplementedError(f"Can not export disassembled datasets such as {ref}")
2869 location, storedFileInfo = fileLocations[0]
2871 pathInStore = location.pathInStore.path
2872 if transfer is None:
2873 # TODO: do we also need to return the readStorageClass somehow?
2874 # We will use the path in store directly. If this is an
2875 # absolute URI, preserve it.
2876 if location.pathInStore.isabs():
2877 pathInStore = str(location.uri)
2878 elif transfer == "direct":
2879 # Use full URIs to the remote store in the export
2880 pathInStore = str(location.uri)
2881 else:
2882 # mypy needs help
2883 assert directoryUri is not None, "directoryUri must be defined to get here"
2884 storeUri = ResourcePath(location.uri)
2886 # if the datastore has an absolute URI to a resource, we
2887 # have two options:
2888 # 1. Keep the absolute URI in the exported YAML
2889 # 2. Allocate a new name in the local datastore and transfer
2890 # it.
2891 # For now go with option 2
2892 if location.pathInStore.isabs():
2893 template = self.templates.getTemplate(ref)
2894 newURI = ResourcePath(template.format(ref), forceAbsolute=False)
2895 pathInStore = str(newURI.updatedExtension(location.pathInStore.getExtension()))
2897 exportUri = directoryUri.join(pathInStore)
2898 exportUri.transfer_from(storeUri, transfer=transfer)
2900 yield FileDataset(refs=[ref], path=pathInStore, formatter=storedFileInfo.formatter)
2902 @staticmethod
2903 def computeChecksum(
2904 uri: ResourcePath, algorithm: str = "blake2b", block_size: int = 8192
2905 ) -> Optional[str]:
2906 """Compute the checksum of the supplied file.
2908 Parameters
2909 ----------
2910 uri : `lsst.resources.ResourcePath`
2911 Name of resource to calculate checksum from.
2912 algorithm : `str`, optional
2913 Name of algorithm to use. Must be one of the algorithms supported
2914 by :py:class`hashlib`.
2915 block_size : `int`
2916 Number of bytes to read from file at one time.
2918 Returns
2919 -------
2920 hexdigest : `str`
2921 Hex digest of the file.
2923 Notes
2924 -----
2925 Currently returns None if the URI is for a remote resource.
2926 """
2927 if algorithm not in hashlib.algorithms_guaranteed:
2928 raise NameError("The specified algorithm '{}' is not supported by hashlib".format(algorithm))
2930 if not uri.isLocal:
2931 return None
2933 hasher = hashlib.new(algorithm)
2935 with uri.as_local() as local_uri:
2936 with open(local_uri.ospath, "rb") as f:
2937 for chunk in iter(lambda: f.read(block_size), b""):
2938 hasher.update(chunk)
2940 return hasher.hexdigest()
2942 def needs_expanded_data_ids(
2943 self,
2944 transfer: Optional[str],
2945 entity: Optional[Union[DatasetRef, DatasetType, StorageClass]] = None,
2946 ) -> bool:
2947 # Docstring inherited.
2948 # This _could_ also use entity to inspect whether the filename template
2949 # involves placeholders other than the required dimensions for its
2950 # dataset type, but that's not necessary for correctness; it just
2951 # enables more optimizations (perhaps only in theory).
2952 return transfer not in ("direct", None)
2954 def import_records(self, data: Mapping[str, DatastoreRecordData]) -> None:
2955 # Docstring inherited from the base class.
2956 record_data = data.get(self.name)
2957 if not record_data:
2958 return
2960 self._bridge.insert(FakeDatasetRef(dataset_id) for dataset_id in record_data.records.keys())
2962 # TODO: Verify that there are no unexpected table names in the dict?
2963 unpacked_records = []
2964 for dataset_data in record_data.records.values():
2965 records = dataset_data.get(self._table.name)
2966 if records:
2967 for info in records:
2968 assert isinstance(info, StoredFileInfo), "Expecting StoredFileInfo records"
2969 unpacked_records.append(info.to_record())
2970 if unpacked_records:
2971 self._table.insert(*unpacked_records, transaction=self._transaction)
2973 def export_records(self, refs: Iterable[DatasetIdRef]) -> Mapping[str, DatastoreRecordData]:
2974 # Docstring inherited from the base class.
2975 exported_refs = list(self._bridge.check(refs))
2976 ids = {ref.id for ref in exported_refs}
2977 records: dict[DatasetId, dict[str, list[StoredDatastoreItemInfo]]] = {id: {} for id in ids}
2978 for row in self._table.fetch(dataset_id=ids):
2979 info: StoredDatastoreItemInfo = StoredFileInfo.from_record(row)
2980 dataset_records = records.setdefault(info.dataset_id, {})
2981 dataset_records.setdefault(self._table.name, []).append(info)
2983 record_data = DatastoreRecordData(records=records)
2984 return {self.name: record_data}
2986 def set_retrieve_dataset_type_method(self, method: Callable[[str], DatasetType | None] | None) -> None:
2987 # Docstring inherited from the base class.
2988 self._retrieve_dataset_method = method
2990 def _cast_storage_class(self, ref: DatasetRef) -> DatasetRef:
2991 """Update dataset reference to use the storage class from registry.
2993 This does nothing for regular datastores, and is only enabled for
2994 trusted mode where we need to use registry definition of storage class
2995 for some datastore methods. `set_retrieve_dataset_type_method` has to
2996 be called beforehand.
2997 """
2998 if self.trustGetRequest:
2999 if self._retrieve_dataset_method is None:
3000 # We could raise an exception here but unit tests do not define
3001 # this method.
3002 return ref
3003 dataset_type = self._retrieve_dataset_method(ref.datasetType.name)
3004 if dataset_type is not None:
3005 ref = ref.overrideStorageClass(dataset_type.storageClass)
3006 return ref