Coverage for tests/test_butler.py: 13%
1337 statements
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-25 10:50 +0000
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-25 10:50 +0000
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 software is dual licensed under the GNU General Public License and also
10# under a 3-clause BSD license. Recipients may choose which of these licenses
11# to use; please see the files gpl-3.0.txt and/or bsd_license.txt,
12# respectively. If you choose the GPL option then the following text applies
13# (but note that there is still no warranty even if you opt for BSD instead):
14#
15# This program is free software: you can redistribute it and/or modify
16# it under the terms of the GNU General Public License as published by
17# the Free Software Foundation, either version 3 of the License, or
18# (at your option) any later version.
19#
20# This program is distributed in the hope that it will be useful,
21# but WITHOUT ANY WARRANTY; without even the implied warranty of
22# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23# GNU General Public License for more details.
24#
25# You should have received a copy of the GNU General Public License
26# along with this program. If not, see <http://www.gnu.org/licenses/>.
28"""Tests for Butler.
29"""
30from __future__ import annotations
32import gc
33import json
34import logging
35import os
36import pathlib
37import pickle
38import posixpath
39import random
40import shutil
41import string
42import tempfile
43import unittest
44import uuid
45from collections.abc import Mapping
46from typing import TYPE_CHECKING, Any, cast
48try:
49 import boto3
50 import botocore
51 from lsst.resources.s3utils import clean_test_environment_for_s3
52 from moto import mock_s3 # type: ignore[import]
53except ImportError:
54 boto3 = None
56 def mock_s3(*args: Any, **kwargs: Any) -> Any: # type: ignore[no-untyped-def]
57 """No-op decorator in case moto mock_s3 can not be imported."""
58 return None
61try:
62 # It's possible but silly to have testing.postgresql installed without
63 # having the postgresql server installed (because then nothing in
64 # testing.postgresql would work), so we use the presence of that module
65 # to test whether we can expect the server to be available.
66 import testing.postgresql # type: ignore[import]
67except ImportError:
68 testing = None
70import astropy.time
71import sqlalchemy
72from lsst.daf.butler import (
73 Butler,
74 ButlerConfig,
75 ButlerRepoIndex,
76 CollectionType,
77 Config,
78 DataCoordinate,
79 DatasetExistence,
80 DatasetRef,
81 DatasetType,
82 FileDataset,
83 StorageClassFactory,
84 ValidationError,
85 script,
86)
87from lsst.daf.butler.datastore import NullDatastore
88from lsst.daf.butler.datastore.file_templates import FileTemplate, FileTemplateValidationError
89from lsst.daf.butler.datastores.fileDatastore import FileDatastore
90from lsst.daf.butler.direct_butler import DirectButler
91from lsst.daf.butler.registry import (
92 CollectionError,
93 CollectionTypeError,
94 ConflictingDefinitionError,
95 DataIdValueError,
96 MissingCollectionError,
97 OrphanedRecordError,
98)
99from lsst.daf.butler.registry.sql_registry import SqlRegistry
100from lsst.daf.butler.repo_relocation import BUTLER_ROOT_TAG
101from lsst.daf.butler.tests import MetricsExample, MultiDetectorFormatter
102from lsst.daf.butler.tests.utils import TestCaseMixin, makeTestTempDir, removeTestTempDir, safeTestTempDir
103from lsst.resources import ResourcePath
104from lsst.utils import doImportType
105from lsst.utils.introspection import get_full_type_name
107if TYPE_CHECKING:
108 import types
110 from lsst.daf.butler import Datastore, DimensionGroup, Registry, StorageClass
112TESTDIR = os.path.abspath(os.path.dirname(__file__))
115def clean_environment() -> None:
116 """Remove external environment variables that affect the tests."""
117 for k in ("DAF_BUTLER_REPOSITORY_INDEX",):
118 os.environ.pop(k, None)
121def makeExampleMetrics() -> MetricsExample:
122 """Return example dataset suitable for tests."""
123 return MetricsExample(
124 {"AM1": 5.2, "AM2": 30.6},
125 {"a": [1, 2, 3], "b": {"blue": 5, "red": "green"}},
126 [563, 234, 456.7, 752, 8, 9, 27],
127 )
130class TransactionTestError(Exception):
131 """Specific error for testing transactions, to prevent misdiagnosing
132 that might otherwise occur when a standard exception is used.
133 """
135 pass
138class ButlerConfigTests(unittest.TestCase):
139 """Simple tests for ButlerConfig that are not tested in any other test
140 cases.
141 """
143 def testSearchPath(self) -> None:
144 configFile = os.path.join(TESTDIR, "config", "basic", "butler.yaml")
145 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
146 config1 = ButlerConfig(configFile)
147 self.assertNotIn("testConfigs", "\n".join(cm.output))
149 overrideDirectory = os.path.join(TESTDIR, "config", "testConfigs")
150 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
151 config2 = ButlerConfig(configFile, searchPaths=[overrideDirectory])
152 self.assertIn("testConfigs", "\n".join(cm.output))
154 key = ("datastore", "records", "table")
155 self.assertNotEqual(config1[key], config2[key])
156 self.assertEqual(config2[key], "override_record")
159class ButlerPutGetTests(TestCaseMixin):
160 """Helper method for running a suite of put/get tests from different
161 butler configurations.
162 """
164 root: str
165 default_run = "ingésτ😺"
166 storageClassFactory: StorageClassFactory
167 configFile: str
168 tmpConfigFile: str
170 @staticmethod
171 def addDatasetType(
172 datasetTypeName: str, dimensions: DimensionGroup, storageClass: StorageClass | str, registry: Registry
173 ) -> DatasetType:
174 """Create a DatasetType and register it"""
175 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
176 registry.registerDatasetType(datasetType)
177 return datasetType
179 @classmethod
180 def setUpClass(cls) -> None:
181 cls.storageClassFactory = StorageClassFactory()
182 cls.storageClassFactory.addFromConfig(cls.configFile)
184 def assertGetComponents(
185 self,
186 butler: Butler,
187 datasetRef: DatasetRef,
188 components: tuple[str, ...],
189 reference: Any,
190 collections: Any = None,
191 ) -> None:
192 datasetType = datasetRef.datasetType
193 dataId = datasetRef.dataId
194 deferred = butler.getDeferred(datasetRef)
196 for component in components:
197 compTypeName = datasetType.componentTypeName(component)
198 result = butler.get(compTypeName, dataId, collections=collections)
199 self.assertEqual(result, getattr(reference, component))
200 result_deferred = deferred.get(component=component)
201 self.assertEqual(result_deferred, result)
203 def tearDown(self) -> None:
204 removeTestTempDir(self.root)
206 def create_butler(
207 self, run: str, storageClass: StorageClass | str, datasetTypeName: str
208 ) -> tuple[DirectButler, DatasetType]:
209 butler = Butler.from_config(self.tmpConfigFile, run=run)
210 assert isinstance(butler, DirectButler), "Expect DirectButler in configuration"
212 collections = set(butler.registry.queryCollections())
213 self.assertEqual(collections, {run})
215 # Create and register a DatasetType
216 dimensions = butler.dimensions.conform(["instrument", "visit"])
218 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
220 # Add needed Dimensions
221 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
222 butler.registry.insertDimensionData(
223 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
224 )
225 butler.registry.insertDimensionData(
226 "visit_system", {"instrument": "DummyCamComp", "id": 1, "name": "default"}
227 )
228 visit_start = astropy.time.Time("2020-01-01 08:00:00.123456789", scale="tai")
229 visit_end = astropy.time.Time("2020-01-01 08:00:36.66", scale="tai")
230 butler.registry.insertDimensionData(
231 "visit",
232 {
233 "instrument": "DummyCamComp",
234 "id": 423,
235 "name": "fourtwentythree",
236 "physical_filter": "d-r",
237 "datetime_begin": visit_start,
238 "datetime_end": visit_end,
239 },
240 )
242 # Add more visits for some later tests
243 for visit_id in (424, 425):
244 butler.registry.insertDimensionData(
245 "visit",
246 {
247 "instrument": "DummyCamComp",
248 "id": visit_id,
249 "name": f"fourtwentyfour_{visit_id}",
250 "physical_filter": "d-r",
251 },
252 )
253 return butler, datasetType
255 def runPutGetTest(self, storageClass: StorageClass, datasetTypeName: str) -> DirectButler:
256 # New datasets will be added to run and tag, but we will only look in
257 # tag when looking up datasets.
258 run = self.default_run
259 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
260 assert butler.run is not None
262 # Create and store a dataset
263 metric = makeExampleMetrics()
264 dataId = butler.registry.expandDataId({"instrument": "DummyCamComp", "visit": 423})
266 # Put and remove the dataset once as a DatasetRef, once as a dataId,
267 # and once with a DatasetType
269 # Keep track of any collections we add and do not clean up
270 expected_collections = {run}
272 counter = 0
273 ref = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run="put_run_1")
274 args = tuple[DatasetRef] | tuple[str | DatasetType, DataCoordinate]
275 for args in ((ref,), (datasetTypeName, dataId), (datasetType, dataId)):
276 # Since we are using subTest we can get cascading failures
277 # here with the first attempt failing and the others failing
278 # immediately because the dataset already exists. Work around
279 # this by using a distinct run collection each time
280 counter += 1
281 this_run = f"put_run_{counter}"
282 butler.registry.registerCollection(this_run, type=CollectionType.RUN)
283 expected_collections.update({this_run})
285 with self.subTest(args=args):
286 kwargs: dict[str, Any] = {}
287 if not isinstance(args[0], DatasetRef): # type: ignore
288 kwargs["run"] = this_run
289 ref = butler.put(metric, *args, **kwargs)
290 self.assertIsInstance(ref, DatasetRef)
292 # Test get of a ref.
293 metricOut = butler.get(ref)
294 self.assertEqual(metric, metricOut)
295 # Test get
296 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
297 self.assertEqual(metric, metricOut)
298 # Test get with a datasetRef
299 metricOut = butler.get(ref)
300 self.assertEqual(metric, metricOut)
301 # Test getDeferred with dataId
302 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
303 self.assertEqual(metric, metricOut)
304 # Test getDeferred with a ref
305 metricOut = butler.getDeferred(ref).get()
306 self.assertEqual(metric, metricOut)
308 # Check we can get components
309 if storageClass.isComposite():
310 self.assertGetComponents(
311 butler, ref, ("summary", "data", "output"), metric, collections=this_run
312 )
314 primary_uri, secondary_uris = butler.getURIs(ref)
315 n_uris = len(secondary_uris)
316 if primary_uri:
317 n_uris += 1
319 # Can the artifacts themselves be retrieved?
320 if not butler._datastore.isEphemeral:
321 # Create a temporary directory to hold the retrieved
322 # artifacts.
323 with tempfile.TemporaryDirectory(
324 prefix="butler-artifacts-", ignore_cleanup_errors=True
325 ) as artifact_root:
326 root_uri = ResourcePath(artifact_root, forceDirectory=True)
328 for preserve_path in (True, False):
329 destination = root_uri.join(f"{preserve_path}_{counter}/")
330 log = logging.getLogger("lsst.x")
331 log.warning("Using destination %s for args %s", destination, args)
332 # Use copy so that we can test that overwrite
333 # protection works (using "auto" for File URIs
334 # would use hard links and subsequent transfer
335 # would work because it knows they are the same
336 # file).
337 transferred = butler.retrieveArtifacts(
338 [ref], destination, preserve_path=preserve_path, transfer="copy"
339 )
340 self.assertGreater(len(transferred), 0)
341 artifacts = list(ResourcePath.findFileResources([destination]))
342 self.assertEqual(set(transferred), set(artifacts))
344 for artifact in transferred:
345 path_in_destination = artifact.relative_to(destination)
346 self.assertIsNotNone(path_in_destination)
347 assert path_in_destination is not None
349 # When path is not preserved there should not
350 # be any path separators.
351 num_seps = path_in_destination.count("/")
352 if preserve_path:
353 self.assertGreater(num_seps, 0)
354 else:
355 self.assertEqual(num_seps, 0)
357 self.assertEqual(
358 len(artifacts),
359 n_uris,
360 "Comparing expected artifacts vs actual:"
361 f" {artifacts} vs {primary_uri} and {secondary_uris}",
362 )
364 if preserve_path:
365 # No need to run these twice
366 with self.assertRaises(ValueError):
367 butler.retrieveArtifacts([ref], destination, transfer="move")
369 with self.assertRaises(FileExistsError):
370 butler.retrieveArtifacts([ref], destination)
372 transferred_again = butler.retrieveArtifacts(
373 [ref], destination, preserve_path=preserve_path, overwrite=True
374 )
375 self.assertEqual(set(transferred_again), set(transferred))
377 # Now remove the dataset completely.
378 butler.pruneDatasets([ref], purge=True, unstore=True)
379 # Lookup with original args should still fail.
380 kwargs = {"collections": this_run}
381 if isinstance(args[0], DatasetRef):
382 kwargs = {} # Prevent warning from being issued.
383 self.assertFalse(butler.exists(*args, **kwargs))
384 # get() should still fail.
385 with self.assertRaises(FileNotFoundError):
386 butler.get(ref)
387 # Registry shouldn't be able to find it by dataset_id anymore.
388 self.assertIsNone(butler.get_dataset(ref.id))
390 # Do explicit registry removal since we know they are
391 # empty
392 butler.registry.removeCollection(this_run)
393 expected_collections.remove(this_run)
395 # Create DatasetRef for put using default run.
396 refIn = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run=butler.run)
398 # Check that getDeferred fails with standalone ref.
399 with self.assertRaises(LookupError):
400 butler.getDeferred(refIn)
402 # Put the dataset again, since the last thing we did was remove it
403 # and we want to use the default collection.
404 ref = butler.put(metric, refIn)
406 # Get with parameters
407 stop = 4
408 sliced = butler.get(ref, parameters={"slice": slice(stop)})
409 self.assertNotEqual(metric, sliced)
410 self.assertEqual(metric.summary, sliced.summary)
411 self.assertEqual(metric.output, sliced.output)
412 assert metric.data is not None # for mypy
413 self.assertEqual(metric.data[:stop], sliced.data)
414 # getDeferred with parameters
415 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
416 self.assertNotEqual(metric, sliced)
417 self.assertEqual(metric.summary, sliced.summary)
418 self.assertEqual(metric.output, sliced.output)
419 self.assertEqual(metric.data[:stop], sliced.data)
420 # getDeferred with deferred parameters
421 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
422 self.assertNotEqual(metric, sliced)
423 self.assertEqual(metric.summary, sliced.summary)
424 self.assertEqual(metric.output, sliced.output)
425 self.assertEqual(metric.data[:stop], sliced.data)
427 if storageClass.isComposite():
428 # Check that components can be retrieved
429 metricOut = butler.get(ref.datasetType.name, dataId)
430 compNameS = ref.datasetType.componentTypeName("summary")
431 compNameD = ref.datasetType.componentTypeName("data")
432 summary = butler.get(compNameS, dataId)
433 self.assertEqual(summary, metric.summary)
434 data = butler.get(compNameD, dataId)
435 self.assertEqual(data, metric.data)
437 if "counter" in storageClass.derivedComponents:
438 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
439 self.assertEqual(count, len(data))
441 count = butler.get(
442 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
443 )
444 self.assertEqual(count, stop)
446 compRef = butler.find_dataset(compNameS, dataId, collections=butler.collections)
447 assert compRef is not None
448 summary = butler.get(compRef)
449 self.assertEqual(summary, metric.summary)
451 # Create a Dataset type that has the same name but is inconsistent.
452 inconsistentDatasetType = DatasetType(
453 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
454 )
456 # Getting with a dataset type that does not match registry fails
457 with self.assertRaisesRegex(ValueError, "Supplied dataset type .* inconsistent with registry"):
458 butler.get(inconsistentDatasetType, dataId)
460 # Combining a DatasetRef with a dataId should fail
461 with self.assertRaisesRegex(ValueError, "DatasetRef given, cannot use dataId as well"):
462 butler.get(ref, dataId)
463 # Getting with an explicit ref should fail if the id doesn't match.
464 with self.assertRaises(FileNotFoundError):
465 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=uuid.UUID(int=101), run=butler.run))
467 # Getting a dataset with unknown parameters should fail
468 with self.assertRaisesRegex(KeyError, "Parameter 'unsupported' not understood"):
469 butler.get(ref, parameters={"unsupported": True})
471 # Check we have a collection
472 collections = set(butler.registry.queryCollections())
473 self.assertEqual(collections, expected_collections)
475 # Clean up to check that we can remove something that may have
476 # already had a component removed
477 butler.pruneDatasets([ref], unstore=True, purge=True)
479 # Add the same ref again, so we can check that duplicate put fails.
480 ref = butler.put(metric, datasetType, dataId)
482 # Repeat put will fail.
483 with self.assertRaisesRegex(
484 ConflictingDefinitionError, "A database constraint failure was triggered"
485 ):
486 butler.put(metric, datasetType, dataId)
488 # Remove the datastore entry.
489 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
491 # Put will still fail
492 with self.assertRaisesRegex(
493 ConflictingDefinitionError, "A database constraint failure was triggered"
494 ):
495 butler.put(metric, datasetType, dataId)
497 # Repeat the same sequence with resolved ref.
498 butler.pruneDatasets([ref], unstore=True, purge=True)
499 ref = butler.put(metric, refIn)
501 # Repeat put will fail.
502 with self.assertRaisesRegex(ConflictingDefinitionError, "Datastore already contains dataset"):
503 butler.put(metric, refIn)
505 # Remove the datastore entry.
506 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
508 # In case of resolved ref this write will succeed.
509 ref = butler.put(metric, refIn)
511 # Leave the dataset in place since some downstream tests require
512 # something to be present
514 return butler
516 def testDeferredCollectionPassing(self) -> None:
517 # Construct a butler with no run or collection, but make it writeable.
518 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
519 # Create and register a DatasetType
520 dimensions = butler.dimensions.conform(["instrument", "visit"])
521 datasetType = self.addDatasetType(
522 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
523 )
524 # Add needed Dimensions
525 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
526 butler.registry.insertDimensionData(
527 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
528 )
529 butler.registry.insertDimensionData(
530 "visit",
531 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
532 )
533 dataId = {"instrument": "DummyCamComp", "visit": 423}
534 # Create dataset.
535 metric = makeExampleMetrics()
536 # Register a new run and put dataset.
537 run = "deferred"
538 self.assertTrue(butler.registry.registerRun(run))
539 # Second time it will be allowed but indicate no-op
540 self.assertFalse(butler.registry.registerRun(run))
541 ref = butler.put(metric, datasetType, dataId, run=run)
542 # Putting with no run should fail with TypeError.
543 with self.assertRaises(CollectionError):
544 butler.put(metric, datasetType, dataId)
545 # Dataset should exist.
546 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
547 # We should be able to get the dataset back, but with and without
548 # a deferred dataset handle.
549 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
550 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
551 # Trying to find the dataset without any collection is a TypeError.
552 self.assertFalse(butler.exists(datasetType, dataId))
553 with self.assertRaises(CollectionError):
554 butler.get(datasetType, dataId)
555 # Associate the dataset with a different collection.
556 butler.registry.registerCollection("tagged")
557 butler.registry.associate("tagged", [ref])
558 # Deleting the dataset from the new collection should make it findable
559 # in the original collection.
560 butler.pruneDatasets([ref], tags=["tagged"])
561 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
564class ButlerTests(ButlerPutGetTests):
565 """Tests for Butler."""
567 useTempRoot = True
568 validationCanFail: bool
569 fullConfigKey: str | None
570 registryStr: str | None
571 datastoreName: list[str] | None
572 datastoreStr: list[str]
574 def setUp(self) -> None:
575 """Create a new butler root for each test."""
576 self.root = makeTestTempDir(TESTDIR)
577 Butler.makeRepo(self.root, config=Config(self.configFile))
578 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
580 def testConstructor(self) -> None:
581 """Independent test of constructor."""
582 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
583 self.assertIsInstance(butler, Butler)
585 # Check that butler.yaml is added automatically.
586 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
587 config_dir = self.tmpConfigFile[: -len(end)]
588 butler = Butler.from_config(config_dir, run=self.default_run)
589 self.assertIsInstance(butler, Butler)
591 # Even with a ResourcePath.
592 butler = Butler.from_config(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
593 self.assertIsInstance(butler, Butler)
595 collections = set(butler.registry.queryCollections())
596 self.assertEqual(collections, {self.default_run})
598 # Check that some special characters can be included in run name.
599 special_run = "u@b.c-A"
600 butler_special = Butler.from_config(butler=butler, run=special_run)
601 collections = set(butler_special.registry.queryCollections("*@*"))
602 self.assertEqual(collections, {special_run})
604 butler2 = Butler.from_config(butler=butler, collections=["other"])
605 self.assertEqual(butler2.collections, ("other",))
606 self.assertIsNone(butler2.run)
607 self.assertEqual(type(butler._datastore), type(butler2._datastore))
608 self.assertEqual(butler._datastore.config, butler2._datastore.config)
610 # Test that we can use an environment variable to find this
611 # repository.
612 butler_index = Config()
613 butler_index["label"] = self.tmpConfigFile
614 for suffix in (".yaml", ".json"):
615 # Ensure that the content differs so that we know that
616 # we aren't reusing the cache.
617 bad_label = f"file://bucket/not_real{suffix}"
618 butler_index["bad_label"] = bad_label
619 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
620 butler_index.dumpToUri(temp_file)
621 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
622 self.assertEqual(Butler.get_known_repos(), {"label", "bad_label"})
623 uri = Butler.get_repo_uri("bad_label")
624 self.assertEqual(uri, ResourcePath(bad_label))
625 uri = Butler.get_repo_uri("label")
626 butler = Butler.from_config(uri, writeable=False)
627 self.assertIsInstance(butler, Butler)
628 butler = Butler.from_config("label", writeable=False)
629 self.assertIsInstance(butler, Butler)
630 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
631 Butler.from_config("not_there", writeable=False)
632 with self.assertRaisesRegex(FileNotFoundError, "resolved from alias 'bad_label'"):
633 Butler.from_config("bad_label")
634 with self.assertRaises(FileNotFoundError):
635 # Should ignore aliases.
636 Butler.from_config(ResourcePath("label", forceAbsolute=False))
637 with self.assertRaises(KeyError) as cm:
638 Butler.get_repo_uri("missing")
639 self.assertEqual(
640 Butler.get_repo_uri("missing", True), ResourcePath("missing", forceAbsolute=False)
641 )
642 self.assertIn("not known to", str(cm.exception))
643 # Should report no failure.
644 self.assertEqual(ButlerRepoIndex.get_failure_reason(), "")
645 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
646 # Now with empty configuration.
647 butler_index = Config()
648 butler_index.dumpToUri(temp_file)
649 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
650 with self.assertRaisesRegex(FileNotFoundError, "(no known aliases)"):
651 Butler.from_config("label")
652 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
653 # Now with bad contents.
654 with open(temp_file.ospath, "w") as fh:
655 print("'", file=fh)
656 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
657 with self.assertRaisesRegex(FileNotFoundError, "(no known aliases:.*could not be read)"):
658 Butler.from_config("label")
659 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
660 with self.assertRaises(FileNotFoundError):
661 Butler.get_repo_uri("label")
662 self.assertEqual(Butler.get_known_repos(), set())
664 with self.assertRaisesRegex(FileNotFoundError, "index file not found"):
665 Butler.from_config("label")
667 # Check that we can create Butler when the alias file is not found.
668 butler = Butler.from_config(self.tmpConfigFile, writeable=False)
669 self.assertIsInstance(butler, Butler)
670 with self.assertRaises(KeyError) as cm:
671 # No environment variable set.
672 Butler.get_repo_uri("label")
673 self.assertEqual(Butler.get_repo_uri("label", True), ResourcePath("label", forceAbsolute=False))
674 self.assertIn("No repository index defined", str(cm.exception))
675 with self.assertRaisesRegex(FileNotFoundError, "no known aliases.*No repository index"):
676 # No aliases registered.
677 Butler.from_config("not_there")
678 self.assertEqual(Butler.get_known_repos(), set())
680 def testBasicPutGet(self) -> None:
681 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
682 self.runPutGetTest(storageClass, "test_metric")
684 def testCompositePutGetConcrete(self) -> None:
685 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
686 butler = self.runPutGetTest(storageClass, "test_metric")
688 # Should *not* be disassembled
689 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
690 self.assertEqual(len(datasets), 1)
691 uri, components = butler.getURIs(datasets[0])
692 self.assertIsInstance(uri, ResourcePath)
693 self.assertFalse(components)
694 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
695 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
697 # Predicted dataset
698 dataId = {"instrument": "DummyCamComp", "visit": 424}
699 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
700 self.assertFalse(components)
701 self.assertIsInstance(uri, ResourcePath)
702 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
703 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
705 def testCompositePutGetVirtual(self) -> None:
706 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
707 butler = self.runPutGetTest(storageClass, "test_metric_comp")
709 # Should be disassembled
710 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
711 self.assertEqual(len(datasets), 1)
712 uri, components = butler.getURIs(datasets[0])
714 if butler._datastore.isEphemeral:
715 # Never disassemble in-memory datastore
716 self.assertIsInstance(uri, ResourcePath)
717 self.assertFalse(components)
718 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
719 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
720 else:
721 self.assertIsNone(uri)
722 self.assertEqual(set(components), set(storageClass.components))
723 for compuri in components.values():
724 self.assertIsInstance(compuri, ResourcePath)
725 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
726 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
728 # Predicted dataset
729 dataId = {"instrument": "DummyCamComp", "visit": 424}
730 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
732 if butler._datastore.isEphemeral:
733 # Never disassembled
734 self.assertIsInstance(uri, ResourcePath)
735 self.assertFalse(components)
736 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
737 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
738 else:
739 self.assertIsNone(uri)
740 self.assertEqual(set(components), set(storageClass.components))
741 for compuri in components.values():
742 self.assertIsInstance(compuri, ResourcePath)
743 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
744 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
746 def testStorageClassOverrideGet(self) -> None:
747 """Test storage class conversion on get with override."""
748 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
749 datasetTypeName = "anything"
750 run = self.default_run
752 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
754 # Create and store a dataset.
755 metric = makeExampleMetrics()
756 dataId = {"instrument": "DummyCamComp", "visit": 423}
758 ref = butler.put(metric, datasetType, dataId)
760 # Return native type.
761 retrieved = butler.get(ref)
762 self.assertEqual(retrieved, metric)
764 # Specify an override.
765 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
766 model = butler.get(ref, storageClass=new_sc)
767 self.assertNotEqual(type(model), type(retrieved))
768 self.assertIs(type(model), new_sc.pytype)
769 self.assertEqual(retrieved, model)
771 # Defer but override later.
772 deferred = butler.getDeferred(ref)
773 model = deferred.get(storageClass=new_sc)
774 self.assertIs(type(model), new_sc.pytype)
775 self.assertEqual(retrieved, model)
777 # Defer but override up front.
778 deferred = butler.getDeferred(ref, storageClass=new_sc)
779 model = deferred.get()
780 self.assertIs(type(model), new_sc.pytype)
781 self.assertEqual(retrieved, model)
783 # Retrieve a component. Should be a tuple.
784 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
785 self.assertIs(type(data), tuple)
786 self.assertEqual(data, tuple(retrieved.data))
788 # Parameter on the write storage class should work regardless
789 # of read storage class.
790 data = butler.get(
791 "anything.data",
792 dataId,
793 storageClass="StructuredDataDataTestTuple",
794 parameters={"slice": slice(2, 4)},
795 )
796 self.assertEqual(len(data), 2)
798 # Try a parameter that is known to the read storage class but not
799 # the write storage class.
800 with self.assertRaises(KeyError):
801 butler.get(
802 "anything.data",
803 dataId,
804 storageClass="StructuredDataDataTestTuple",
805 parameters={"xslice": slice(2, 4)},
806 )
808 def testPytypePutCoercion(self) -> None:
809 """Test python type coercion on Butler.get and put."""
810 # Store some data with the normal example storage class.
811 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
812 datasetTypeName = "test_metric"
813 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
815 dataId = {"instrument": "DummyCamComp", "visit": 423}
817 # Put a dict and this should coerce to a MetricsExample
818 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
819 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
820 test_metric = butler.get(metric_ref)
821 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
822 self.assertEqual(test_metric.summary, test_dict["summary"])
823 self.assertEqual(test_metric.output, test_dict["output"])
825 # Check that the put still works if a DatasetType is given with
826 # a definition matching this python type.
827 registry_type = butler.get_dataset_type(datasetTypeName)
828 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
829 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
830 self.assertEqual(metric2_ref.datasetType, registry_type)
832 # The get will return the type expected by registry.
833 test_metric2 = butler.get(metric2_ref)
834 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
836 # Make a new DatasetRef with the compatible but different DatasetType.
837 # This should now return a dict.
838 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
839 test_dict2 = butler.get(new_ref)
840 self.assertEqual(get_full_type_name(test_dict2), "dict")
842 # Get it again with the wrong dataset type definition using get()
843 # rather than get(). This should be consistent with get()
844 # behavior and return the type of the DatasetType.
845 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
846 self.assertEqual(get_full_type_name(test_dict3), "dict")
848 def testIngest(self) -> None:
849 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
851 # Create and register a DatasetType
852 dimensions = butler.dimensions.conform(["instrument", "visit", "detector"])
854 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
855 datasetTypeName = "metric"
857 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
859 # Add needed Dimensions
860 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
861 butler.registry.insertDimensionData(
862 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
863 )
864 for detector in (1, 2):
865 butler.registry.insertDimensionData(
866 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
867 )
869 butler.registry.insertDimensionData(
870 "visit",
871 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
872 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
873 )
875 formatter = doImportType("lsst.daf.butler.formatters.yaml.YamlFormatter")
876 dataRoot = os.path.join(TESTDIR, "data", "basic")
877 datasets = []
878 for detector in (1, 2):
879 detector_name = f"detector_{detector}"
880 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
881 dataId = butler.registry.expandDataId(
882 {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
883 )
884 # Create a DatasetRef for ingest
885 refIn = DatasetRef(datasetType, dataId, run=self.default_run)
887 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
889 butler.ingest(*datasets, transfer="copy")
891 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
892 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
894 metrics1 = butler.get(datasetTypeName, dataId1)
895 metrics2 = butler.get(datasetTypeName, dataId2)
896 self.assertNotEqual(metrics1, metrics2)
898 # Compare URIs
899 uri1 = butler.getURI(datasetTypeName, dataId1)
900 uri2 = butler.getURI(datasetTypeName, dataId2)
901 self.assertNotEqual(uri1, uri2)
903 # Now do a multi-dataset but single file ingest
904 metricFile = os.path.join(dataRoot, "detectors.yaml")
905 refs = []
906 for detector in (1, 2):
907 detector_name = f"detector_{detector}"
908 dataId = butler.registry.expandDataId(
909 {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
910 )
911 # Create a DatasetRef for ingest
912 refs.append(DatasetRef(datasetType, dataId, run=self.default_run))
914 # Test "move" transfer to ensure that the files themselves
915 # have disappeared following ingest.
916 with ResourcePath.temporary_uri(suffix=".yaml") as tempFile:
917 tempFile.transfer_from(ResourcePath(metricFile), transfer="copy")
919 datasets = []
920 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter))
922 # For first ingest use copy.
923 butler.ingest(*datasets, transfer="copy", record_validation_info=False)
925 # Now try to ingest again in "execution butler" mode where
926 # the registry entries exist but the datastore does not have
927 # the files. We also need to strip the dimension records to ensure
928 # that they will be re-added by the ingest.
929 ref = datasets[0].refs[0]
930 datasets[0].refs = [
931 cast(
932 DatasetRef,
933 butler.find_dataset(ref.datasetType, data_id=ref.dataId, collections=ref.run),
934 )
935 for ref in datasets[0].refs
936 ]
937 all_refs = []
938 for dataset in datasets:
939 refs = []
940 for ref in dataset.refs:
941 # Create a dict from the dataId to drop the records.
942 new_data_id = dict(ref.dataId.required)
943 new_ref = butler.find_dataset(ref.datasetType, new_data_id, collections=ref.run)
944 assert new_ref is not None
945 self.assertFalse(new_ref.dataId.hasRecords())
946 refs.append(new_ref)
947 dataset.refs = refs
948 all_refs.extend(dataset.refs)
949 butler.pruneDatasets(all_refs, disassociate=False, unstore=True, purge=False)
951 # Use move mode to test that the file is deleted. Also
952 # disable recording of file size.
953 butler.ingest(*datasets, transfer="move", record_validation_info=False)
955 # Check that every ref now has records.
956 for dataset in datasets:
957 for ref in dataset.refs:
958 self.assertTrue(ref.dataId.hasRecords())
960 # Ensure that the file has disappeared.
961 self.assertFalse(tempFile.exists())
963 # Check that the datastore recorded no file size.
964 # Not all datastores can support this.
965 try:
966 infos = butler._datastore.getStoredItemsInfo(datasets[0].refs[0]) # type: ignore[attr-defined]
967 self.assertEqual(infos[0].file_size, -1)
968 except AttributeError:
969 pass
971 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
972 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
974 multi1 = butler.get(datasetTypeName, dataId1)
975 multi2 = butler.get(datasetTypeName, dataId2)
977 self.assertEqual(multi1, metrics1)
978 self.assertEqual(multi2, metrics2)
980 # Compare URIs
981 uri1 = butler.getURI(datasetTypeName, dataId1)
982 uri2 = butler.getURI(datasetTypeName, dataId2)
983 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
985 # Test that removing one does not break the second
986 # This line will issue a warning log message for a ChainedDatastore
987 # that uses an InMemoryDatastore since in-memory can not ingest
988 # files.
989 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
990 self.assertFalse(butler.exists(datasetTypeName, dataId1))
991 self.assertTrue(butler.exists(datasetTypeName, dataId2))
992 multi2b = butler.get(datasetTypeName, dataId2)
993 self.assertEqual(multi2, multi2b)
995 # Ensure we can ingest 0 datasets
996 datasets = []
997 butler.ingest(*datasets)
999 def testPickle(self) -> None:
1000 """Test pickle support."""
1001 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1002 assert isinstance(butler, DirectButler), "Expect DirectButler in configuration"
1003 butlerOut = pickle.loads(pickle.dumps(butler))
1004 self.assertIsInstance(butlerOut, Butler)
1005 self.assertEqual(butlerOut._config, butler._config)
1006 self.assertEqual(butlerOut.collections, butler.collections)
1007 self.assertEqual(butlerOut.run, butler.run)
1009 def testGetDatasetTypes(self) -> None:
1010 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1011 dimensions = butler.dimensions.conform(["instrument", "visit", "physical_filter"])
1012 dimensionEntries: list[tuple[str, list[Mapping[str, Any]]]] = [
1013 (
1014 "instrument",
1015 [
1016 {"instrument": "DummyCam"},
1017 {"instrument": "DummyHSC"},
1018 {"instrument": "DummyCamComp"},
1019 ],
1020 ),
1021 ("physical_filter", [{"instrument": "DummyCam", "name": "d-r", "band": "R"}]),
1022 ("visit", [{"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}]),
1023 ]
1024 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1025 # Add needed Dimensions
1026 for element, data in dimensionEntries:
1027 butler.registry.insertDimensionData(element, *data)
1029 # When a DatasetType is added to the registry entries are not created
1030 # for components but querying them can return the components.
1031 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
1032 components = set()
1033 for datasetTypeName in datasetTypeNames:
1034 # Create and register a DatasetType
1035 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1037 for componentName in storageClass.components:
1038 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
1040 fromRegistry: set[DatasetType] = set()
1041 for parent_dataset_type in butler.registry.queryDatasetTypes():
1042 fromRegistry.add(parent_dataset_type)
1043 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
1044 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
1046 # Now that we have some dataset types registered, validate them
1047 butler.validateConfiguration(
1048 ignore=[
1049 "test_metric_comp",
1050 "metric3",
1051 "metric5",
1052 "calexp",
1053 "DummySC",
1054 "datasetType.component",
1055 "random_data",
1056 "random_data_2",
1057 ]
1058 )
1060 # Add a new datasetType that will fail template validation
1061 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
1062 if self.validationCanFail:
1063 with self.assertRaises(ValidationError):
1064 butler.validateConfiguration()
1066 # Rerun validation but with a subset of dataset type names
1067 butler.validateConfiguration(datasetTypeNames=["metric4"])
1069 # Rerun validation but ignore the bad datasetType
1070 butler.validateConfiguration(
1071 ignore=[
1072 "test_metric_comp",
1073 "metric3",
1074 "metric5",
1075 "calexp",
1076 "DummySC",
1077 "datasetType.component",
1078 "random_data",
1079 "random_data_2",
1080 ]
1081 )
1083 def testTransaction(self) -> None:
1084 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1085 datasetTypeName = "test_metric"
1086 dimensions = butler.dimensions.conform(["instrument", "visit"])
1087 dimensionEntries: tuple[tuple[str, Mapping[str, Any]], ...] = (
1088 ("instrument", {"instrument": "DummyCam"}),
1089 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1090 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1091 )
1092 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1093 metric = makeExampleMetrics()
1094 dataId = {"instrument": "DummyCam", "visit": 42}
1095 # Create and register a DatasetType
1096 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1097 with self.assertRaises(TransactionTestError):
1098 with butler.transaction():
1099 # Add needed Dimensions
1100 for args in dimensionEntries:
1101 butler.registry.insertDimensionData(*args)
1102 # Store a dataset
1103 ref = butler.put(metric, datasetTypeName, dataId)
1104 self.assertIsInstance(ref, DatasetRef)
1105 # Test get of a ref.
1106 metricOut = butler.get(ref)
1107 self.assertEqual(metric, metricOut)
1108 # Test get
1109 metricOut = butler.get(datasetTypeName, dataId)
1110 self.assertEqual(metric, metricOut)
1111 # Check we can get components
1112 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1113 raise TransactionTestError("This should roll back the entire transaction")
1114 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1115 butler.registry.expandDataId(dataId)
1116 # Should raise LookupError for missing data ID value
1117 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1118 butler.get(datasetTypeName, dataId)
1119 # Also check explicitly if Dataset entry is missing
1120 self.assertIsNone(butler.find_dataset(datasetType, dataId, collections=butler.collections))
1121 # Direct retrieval should not find the file in the Datastore
1122 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1123 butler.get(ref)
1125 def testMakeRepo(self) -> None:
1126 """Test that we can write butler configuration to a new repository via
1127 the Butler.makeRepo interface and then instantiate a butler from the
1128 repo root.
1129 """
1130 # Do not run the test if we know this datastore configuration does
1131 # not support a file system root
1132 if self.fullConfigKey is None:
1133 return
1135 # create two separate directories
1136 root1 = tempfile.mkdtemp(dir=self.root)
1137 root2 = tempfile.mkdtemp(dir=self.root)
1139 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1140 limited = Config(self.configFile)
1141 butler1 = Butler.from_config(butlerConfig)
1142 assert isinstance(butler1, DirectButler), "Expect DirectButler in configuration"
1143 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1144 full = Config(self.tmpConfigFile)
1145 butler2 = Butler.from_config(butlerConfig)
1146 assert isinstance(butler2, DirectButler), "Expect DirectButler in configuration"
1147 # Butlers should have the same configuration regardless of whether
1148 # defaults were expanded.
1149 self.assertEqual(butler1._config, butler2._config)
1150 # Config files loaded directly should not be the same.
1151 self.assertNotEqual(limited, full)
1152 # Make sure "limited" doesn't have a few keys we know it should be
1153 # inheriting from defaults.
1154 self.assertIn(self.fullConfigKey, full)
1155 self.assertNotIn(self.fullConfigKey, limited)
1157 # Collections don't appear until something is put in them
1158 collections1 = set(butler1.registry.queryCollections())
1159 self.assertEqual(collections1, set())
1160 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1162 # Check that a config with no associated file name will not
1163 # work properly with relocatable Butler repo
1164 butlerConfig.configFile = None
1165 with self.assertRaises(ValueError):
1166 Butler.from_config(butlerConfig)
1168 with self.assertRaises(FileExistsError):
1169 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1171 def testStringification(self) -> None:
1172 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1173 butlerStr = str(butler)
1175 if self.datastoreStr is not None:
1176 for testStr in self.datastoreStr:
1177 self.assertIn(testStr, butlerStr)
1178 if self.registryStr is not None:
1179 self.assertIn(self.registryStr, butlerStr)
1181 datastoreName = butler._datastore.name
1182 if self.datastoreName is not None:
1183 for testStr in self.datastoreName:
1184 self.assertIn(testStr, datastoreName)
1186 def testButlerRewriteDataId(self) -> None:
1187 """Test that dataIds can be rewritten based on dimension records."""
1188 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1190 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1191 datasetTypeName = "random_data"
1193 # Create dimension records.
1194 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1195 butler.registry.insertDimensionData(
1196 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1197 )
1198 butler.registry.insertDimensionData(
1199 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1200 )
1202 dimensions = butler.dimensions.conform(["instrument", "exposure"])
1203 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1204 butler.registry.registerDatasetType(datasetType)
1206 n_exposures = 5
1207 dayobs = 20210530
1209 for i in range(n_exposures):
1210 butler.registry.insertDimensionData(
1211 "exposure",
1212 {
1213 "instrument": "DummyCamComp",
1214 "id": i,
1215 "obs_id": f"exp{i}",
1216 "seq_num": i,
1217 "day_obs": dayobs,
1218 "physical_filter": "d-r",
1219 },
1220 )
1222 # Write some data.
1223 for i in range(n_exposures):
1224 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1226 # Use the seq_num for the put to test rewriting.
1227 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1228 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1230 # Check that the exposure is correct in the dataId
1231 self.assertEqual(ref.dataId["exposure"], i)
1233 # and check that we can get the dataset back with the same dataId
1234 new_metric = butler.get(datasetTypeName, dataId=dataId)
1235 self.assertEqual(new_metric, metric)
1237 def testGetDatasetCollectionCaching(self):
1238 # Prior to DM-41117, there was a bug where get_dataset would throw
1239 # MissingCollectionError if you tried to fetch a dataset that was added
1240 # after the collection cache was last updated.
1241 reader_butler, datasetType = self.create_butler(self.default_run, "int", "datasettypename")
1242 writer_butler = Butler.from_config(self.tmpConfigFile, writeable=True, run="new_run")
1243 dataId = {"instrument": "DummyCamComp", "visit": 423}
1244 put_ref = writer_butler.put(123, datasetType, dataId)
1245 get_ref = reader_butler.get_dataset(put_ref.id)
1246 self.assertEqual(get_ref.id, put_ref.id)
1249class FileDatastoreButlerTests(ButlerTests):
1250 """Common tests and specialization of ButlerTests for butlers backed
1251 by datastores that inherit from FileDatastore.
1252 """
1254 def checkFileExists(self, root: str | ResourcePath, relpath: str | ResourcePath) -> bool:
1255 """Check if file exists at a given path (relative to root).
1257 Test testPutTemplates verifies actual physical existance of the files
1258 in the requested location.
1259 """
1260 uri = ResourcePath(root, forceDirectory=True)
1261 return uri.join(relpath).exists()
1263 def testPutTemplates(self) -> None:
1264 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1265 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1267 # Add needed Dimensions
1268 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1269 butler.registry.insertDimensionData(
1270 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1271 )
1272 butler.registry.insertDimensionData(
1273 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1274 )
1275 butler.registry.insertDimensionData(
1276 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1277 )
1279 # Create and store a dataset
1280 metric = makeExampleMetrics()
1282 # Create two almost-identical DatasetTypes (both will use default
1283 # template)
1284 dimensions = butler.dimensions.conform(["instrument", "visit"])
1285 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1286 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1287 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1289 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1290 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1292 # Put with exactly the data ID keys needed
1293 ref = butler.put(metric, "metric1", dataId1)
1294 uri = butler.getURI(ref)
1295 self.assertTrue(uri.exists())
1296 self.assertTrue(
1297 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle")
1298 )
1300 # Check the template based on dimensions
1301 if hasattr(butler._datastore, "templates"):
1302 butler._datastore.templates.validateTemplates([ref])
1304 # Put with extra data ID keys (physical_filter is an optional
1305 # dependency); should not change template (at least the way we're
1306 # defining them to behave now; the important thing is that they
1307 # must be consistent).
1308 ref = butler.put(metric, "metric2", dataId2)
1309 uri = butler.getURI(ref)
1310 self.assertTrue(uri.exists())
1311 self.assertTrue(
1312 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle")
1313 )
1315 # Check the template based on dimensions
1316 if hasattr(butler._datastore, "templates"):
1317 butler._datastore.templates.validateTemplates([ref])
1319 # Use a template that has a typo in dimension record metadata.
1320 # Easier to test with a butler that has a ref with records attached.
1321 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1322 with self.assertLogs("lsst.daf.butler.datastore.file_templates", "INFO"):
1323 path = template.format(ref)
1324 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1326 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1327 with self.assertRaises(KeyError):
1328 with self.assertLogs("lsst.daf.butler.datastore.file_templates", "INFO"):
1329 template.format(ref)
1331 # Now use a file template that will not result in unique filenames
1332 with self.assertRaises(FileTemplateValidationError):
1333 butler.put(metric, "metric3", dataId1)
1335 def testImportExport(self) -> None:
1336 # Run put/get tests just to create and populate a repo.
1337 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1338 self.runImportExportTest(storageClass)
1340 @unittest.expectedFailure
1341 def testImportExportVirtualComposite(self) -> None:
1342 # Run put/get tests just to create and populate a repo.
1343 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1344 self.runImportExportTest(storageClass)
1346 def runImportExportTest(self, storageClass: StorageClass) -> None:
1347 """Test exporting and importing.
1349 This test does an export to a temp directory and an import back
1350 into a new temp directory repo. It does not assume a posix datastore.
1351 """
1352 exportButler = self.runPutGetTest(storageClass, "test_metric")
1354 # Test that we must have a file extension.
1355 with self.assertRaises(ValueError):
1356 with exportButler.export(filename="dump", directory=".") as export:
1357 pass
1359 # Test that unknown format is not allowed.
1360 with self.assertRaises(ValueError):
1361 with exportButler.export(filename="dump.fits", directory=".") as export:
1362 pass
1364 # Test that the repo actually has at least one dataset.
1365 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1366 self.assertGreater(len(datasets), 0)
1367 # Add a DimensionRecord that's unused by those datasets.
1368 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1369 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1370 # Export and then import datasets.
1371 with safeTestTempDir(TESTDIR) as exportDir:
1372 exportFile = os.path.join(exportDir, "exports.yaml")
1373 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1374 export.saveDatasets(datasets)
1375 # Export the same datasets again. This should quietly do
1376 # nothing because of internal deduplication, and it shouldn't
1377 # complain about being asked to export the "htm7" elements even
1378 # though there aren't any in these datasets or in the database.
1379 export.saveDatasets(datasets, elements=["htm7"])
1380 # Save one of the data IDs again; this should be harmless
1381 # because of internal deduplication.
1382 export.saveDataIds([datasets[0].dataId])
1383 # Save some dimension records directly.
1384 export.saveDimensionData("skymap", [skymapRecord])
1385 self.assertTrue(os.path.exists(exportFile))
1386 with safeTestTempDir(TESTDIR) as importDir:
1387 # We always want this to be a local posix butler
1388 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1389 # Calling script.butlerImport tests the implementation of the
1390 # butler command line interface "import" subcommand. Functions
1391 # in the script folder are generally considered protected and
1392 # should not be used as public api.
1393 with open(exportFile) as f:
1394 script.butlerImport(
1395 importDir,
1396 export_file=f,
1397 directory=exportDir,
1398 transfer="auto",
1399 skip_dimensions=None,
1400 )
1401 importButler = Butler.from_config(importDir, run=self.default_run)
1402 for ref in datasets:
1403 with self.subTest(ref=ref):
1404 # Test for existence by passing in the DatasetType and
1405 # data ID separately, to avoid lookup by dataset_id.
1406 self.assertTrue(importButler.exists(ref.datasetType, ref.dataId))
1407 self.assertEqual(
1408 list(importButler.registry.queryDimensionRecords("skymap")),
1409 [importButler.dimensions["skymap"].RecordClass(**skymapRecord)],
1410 )
1412 def testRemoveRuns(self) -> None:
1413 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1414 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
1415 # Load registry data with dimensions to hang datasets off of.
1416 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1417 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1418 # Add some RUN-type collection.
1419 run1 = "run1"
1420 butler.registry.registerRun(run1)
1421 run2 = "run2"
1422 butler.registry.registerRun(run2)
1423 # put a dataset in each
1424 metric = makeExampleMetrics()
1425 dimensions = butler.dimensions.conform(["instrument", "physical_filter"])
1426 datasetType = self.addDatasetType(
1427 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1428 )
1429 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1430 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1431 uri1 = butler.getURI(ref1)
1432 uri2 = butler.getURI(ref2)
1434 with self.assertRaises(OrphanedRecordError):
1435 butler.registry.removeDatasetType(datasetType.name)
1437 # Remove from both runs with different values for unstore.
1438 butler.removeRuns([run1], unstore=True)
1439 butler.removeRuns([run2], unstore=False)
1440 # Should be nothing in registry for either one, and datastore should
1441 # not think either exists.
1442 with self.assertRaises(MissingCollectionError):
1443 butler.registry.getCollectionType(run1)
1444 with self.assertRaises(MissingCollectionError):
1445 butler.registry.getCollectionType(run2)
1446 self.assertFalse(butler.stored(ref1))
1447 self.assertFalse(butler.stored(ref2))
1448 # The ref we unstored should be gone according to the URI, but the
1449 # one we forgot should still be around.
1450 self.assertFalse(uri1.exists())
1451 self.assertTrue(uri2.exists())
1453 # Now that the collections have been pruned we can remove the
1454 # dataset type
1455 butler.registry.removeDatasetType(datasetType.name)
1457 with self.assertLogs("lsst.daf.butler.registry", "INFO") as cm:
1458 butler.registry.removeDatasetType(("test*", "test*"))
1459 self.assertIn("not defined", "\n".join(cm.output))
1462class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1463 """PosixDatastore specialization of a butler"""
1465 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1466 fullConfigKey: str | None = ".datastore.formatters"
1467 validationCanFail = True
1468 datastoreStr = ["/tmp"]
1469 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1470 registryStr = "/gen3.sqlite3"
1472 def testPathConstructor(self) -> None:
1473 """Independent test of constructor using PathLike."""
1474 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1475 self.assertIsInstance(butler, Butler)
1477 # And again with a Path object with the butler yaml
1478 path = pathlib.Path(self.tmpConfigFile)
1479 butler = Butler.from_config(path, writeable=False)
1480 self.assertIsInstance(butler, Butler)
1482 # And again with a Path object without the butler yaml
1483 # (making sure we skip it if the tmp config doesn't end
1484 # in butler.yaml -- which is the case for a subclass)
1485 if self.tmpConfigFile.endswith("butler.yaml"):
1486 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1487 butler = Butler.from_config(path, writeable=False)
1488 self.assertIsInstance(butler, Butler)
1490 def testExportTransferCopy(self) -> None:
1491 """Test local export using all transfer modes"""
1492 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1493 exportButler = self.runPutGetTest(storageClass, "test_metric")
1494 # Test that the repo actually has at least one dataset.
1495 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1496 self.assertGreater(len(datasets), 0)
1497 uris = [exportButler.getURI(d) for d in datasets]
1498 assert isinstance(exportButler._datastore, FileDatastore)
1499 datastoreRoot = exportButler.get_datastore_roots()[exportButler.get_datastore_names()[0]]
1501 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1503 for path in pathsInStore:
1504 # Assume local file system
1505 assert path is not None
1506 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1508 for transfer in ("copy", "link", "symlink", "relsymlink"):
1509 with safeTestTempDir(TESTDIR) as exportDir:
1510 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1511 export.saveDatasets(datasets)
1512 for path in pathsInStore:
1513 assert path is not None
1514 self.assertTrue(
1515 self.checkFileExists(exportDir, path),
1516 f"Check that mode {transfer} exported files",
1517 )
1519 def testPruneDatasets(self) -> None:
1520 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1521 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
1522 assert isinstance(butler._datastore, FileDatastore)
1523 # Load registry data with dimensions to hang datasets off of.
1524 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1525 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1526 # Add some RUN-type collections.
1527 run1 = "run1"
1528 butler.registry.registerRun(run1)
1529 run2 = "run2"
1530 butler.registry.registerRun(run2)
1531 # put some datasets. ref1 and ref2 have the same data ID, and are in
1532 # different runs. ref3 has a different data ID.
1533 metric = makeExampleMetrics()
1534 dimensions = butler.dimensions.conform(["instrument", "physical_filter"])
1535 datasetType = self.addDatasetType(
1536 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1537 )
1538 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1539 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1540 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1542 many_stored = butler.stored_many([ref1, ref2, ref3])
1543 for ref, stored in many_stored.items():
1544 self.assertTrue(stored, f"Ref {ref} should be stored")
1546 many_exists = butler._exists_many([ref1, ref2, ref3])
1547 for ref, exists in many_exists.items():
1548 self.assertTrue(exists, f"Checking ref {ref} exists.")
1549 self.assertEqual(exists, DatasetExistence.VERIFIED, f"Ref {ref} should be stored")
1551 # Simple prune.
1552 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1553 self.assertFalse(butler.exists(ref1.datasetType, ref1.dataId, collections=run1))
1555 many_stored = butler.stored_many([ref1, ref2, ref3])
1556 for ref, stored in many_stored.items():
1557 self.assertFalse(stored, f"Ref {ref} should not be stored")
1559 many_exists = butler._exists_many([ref1, ref2, ref3])
1560 for ref, exists in many_exists.items():
1561 self.assertEqual(exists, DatasetExistence.UNRECOGNIZED, f"Ref {ref} should not be stored")
1563 # Put data back.
1564 ref1_new = butler.put(metric, ref1)
1565 self.assertEqual(ref1_new, ref1) # Reuses original ID.
1566 ref2 = butler.put(metric, ref2)
1568 many_stored = butler.stored_many([ref1, ref2, ref3])
1569 self.assertTrue(many_stored[ref1])
1570 self.assertTrue(many_stored[ref2])
1571 self.assertFalse(many_stored[ref3])
1573 ref3 = butler.put(metric, ref3)
1575 many_exists = butler._exists_many([ref1, ref2, ref3])
1576 for ref, exists in many_exists.items():
1577 self.assertTrue(exists, f"Ref {ref} should not be stored")
1579 # Clear out the datasets from registry and start again.
1580 refs = [ref1, ref2, ref3]
1581 butler.pruneDatasets(refs, purge=True, unstore=True)
1582 for ref in refs:
1583 butler.put(metric, ref)
1585 # Confirm we can retrieve deferred.
1586 dref1 = butler.getDeferred(ref1) # known and exists
1587 metric1 = dref1.get()
1588 self.assertEqual(metric1, metric)
1590 # Test different forms of file availability.
1591 # Need to be in a state where:
1592 # - one ref just has registry record.
1593 # - one ref has a missing file but a datastore record.
1594 # - one ref has a missing datastore record but file is there.
1595 # - one ref does not exist anywhere.
1596 # Do not need to test a ref that has everything since that is tested
1597 # above.
1598 ref0 = DatasetRef(
1599 datasetType,
1600 DataCoordinate.standardize(
1601 {"instrument": "Cam1", "physical_filter": "Cam1-G"}, universe=butler.dimensions
1602 ),
1603 run=run1,
1604 )
1606 # Delete from datastore and retain in Registry.
1607 butler.pruneDatasets([ref1], purge=False, unstore=True, disassociate=False)
1609 # File has been removed.
1610 uri2 = butler.getURI(ref2)
1611 uri2.remove()
1613 # Datastore has lost track.
1614 butler._datastore.forget([ref3])
1616 # First test with a standard butler.
1617 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1618 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1619 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1620 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1621 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED)
1623 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=False)
1624 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1625 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1626 self.assertEqual(exists_many[ref2], DatasetExistence.KNOWN)
1627 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1628 self.assertTrue(exists_many[ref2])
1630 # Check that per-ref query gives the same answer as many query.
1631 for ref, exists in exists_many.items():
1632 self.assertEqual(butler.exists(ref, full_check=False), exists)
1634 # Get deferred checks for existence before it allows it to be
1635 # retrieved.
1636 with self.assertRaises(LookupError):
1637 butler.getDeferred(ref3) # not known, file exists
1638 dref2 = butler.getDeferred(ref2) # known but file missing
1639 with self.assertRaises(FileNotFoundError):
1640 dref2.get()
1642 # Test again with a trusting butler.
1643 butler._datastore.trustGetRequest = True
1644 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1645 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1646 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1647 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1648 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ARTIFACT)
1650 # When trusting we can get a deferred dataset handle that is not
1651 # known but does exist.
1652 dref3 = butler.getDeferred(ref3)
1653 metric3 = dref3.get()
1654 self.assertEqual(metric3, metric)
1656 # Check that per-ref query gives the same answer as many query.
1657 for ref, exists in exists_many.items():
1658 self.assertEqual(butler.exists(ref, full_check=True), exists)
1660 # Create a ref that surprisingly has the UUID of an existing ref
1661 # but is not the same.
1662 ref_bad = DatasetRef(datasetType, dataId=ref3.dataId, run=ref3.run, id=ref2.id)
1663 with self.assertRaises(ValueError):
1664 butler.exists(ref_bad)
1666 # Create a ref that has a compatible storage class.
1667 ref_compat = ref2.overrideStorageClass("StructuredDataDict")
1668 exists = butler.exists(ref_compat)
1669 self.assertEqual(exists, exists_many[ref2])
1671 # Remove everything and start from scratch.
1672 butler._datastore.trustGetRequest = False
1673 butler.pruneDatasets(refs, purge=True, unstore=True)
1674 for ref in refs:
1675 butler.put(metric, ref)
1677 # These tests mess directly with the trash table and can leave the
1678 # datastore in an odd state. Do them at the end.
1679 # Check that in normal mode, deleting the record will lead to
1680 # trash not touching the file.
1681 uri1 = butler.getURI(ref1)
1682 butler._datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1683 butler._datastore.forget([ref1])
1684 butler._datastore.trash(ref1)
1685 butler._datastore.emptyTrash()
1686 self.assertTrue(uri1.exists())
1687 uri1.remove() # Clean it up.
1689 # Simulate execution butler setup by deleting the datastore
1690 # record but keeping the file around and trusting.
1691 butler._datastore.trustGetRequest = True
1692 uris = butler.get_many_uris([ref2, ref3])
1693 uri2 = uris[ref2].primaryURI
1694 uri3 = uris[ref3].primaryURI
1695 self.assertTrue(uri2.exists())
1696 self.assertTrue(uri3.exists())
1698 # Remove the datastore record.
1699 butler._datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1700 butler._datastore.forget([ref2])
1701 self.assertTrue(uri2.exists())
1702 butler._datastore.trash([ref2, ref3])
1703 # Immediate removal for ref2 file
1704 self.assertFalse(uri2.exists())
1705 # But ref3 has to wait for the empty.
1706 self.assertTrue(uri3.exists())
1707 butler._datastore.emptyTrash()
1708 self.assertFalse(uri3.exists())
1710 # Clear out the datasets from registry.
1711 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1713 def testPytypeCoercion(self) -> None:
1714 """Test python type coercion on Butler.get and put."""
1715 # Store some data with the normal example storage class.
1716 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1717 datasetTypeName = "test_metric"
1718 butler = self.runPutGetTest(storageClass, datasetTypeName)
1720 dataId = {"instrument": "DummyCamComp", "visit": 423}
1721 metric = butler.get(datasetTypeName, dataId=dataId)
1722 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1724 datasetType_ori = butler.get_dataset_type(datasetTypeName)
1725 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1727 # Now need to hack the registry dataset type definition.
1728 # There is no API for this.
1729 assert isinstance(butler._registry, SqlRegistry)
1730 manager = butler._registry._managers.datasets
1731 assert hasattr(manager, "_db") and hasattr(manager, "_static")
1732 manager._db.update(
1733 manager._static.dataset_type,
1734 {"name": datasetTypeName},
1735 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1736 )
1738 # Force reset of dataset type cache
1739 butler.registry.refresh()
1741 datasetType_new = butler.get_dataset_type(datasetTypeName)
1742 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1743 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1745 metric_model = butler.get(datasetTypeName, dataId=dataId)
1746 self.assertNotEqual(type(metric_model), type(metric))
1747 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1749 # Put the model and read it back to show that everything now
1750 # works as normal.
1751 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1752 metric_model_new = butler.get(metric_ref)
1753 self.assertEqual(metric_model_new, metric_model)
1755 # Hack the storage class again to something that will fail on the
1756 # get with no conversion class.
1757 manager._db.update(
1758 manager._static.dataset_type,
1759 {"name": datasetTypeName},
1760 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1761 )
1762 butler.registry.refresh()
1764 with self.assertRaises(ValueError):
1765 butler.get(datasetTypeName, dataId=dataId)
1768@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1769class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1770 """PosixDatastore specialization of a butler using Postgres"""
1772 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1773 fullConfigKey = ".datastore.formatters"
1774 validationCanFail = True
1775 datastoreStr = ["/tmp"]
1776 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1777 registryStr = "PostgreSQL@test"
1778 postgresql: Any
1780 @staticmethod
1781 def _handler(postgresql: Any) -> None:
1782 engine = sqlalchemy.engine.create_engine(postgresql.url())
1783 with engine.begin() as connection:
1784 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1786 @classmethod
1787 def setUpClass(cls) -> None:
1788 # Create the postgres test server.
1789 cls.postgresql = testing.postgresql.PostgresqlFactory(
1790 cache_initialized_db=True, on_initialized=cls._handler
1791 )
1792 super().setUpClass()
1794 @classmethod
1795 def tearDownClass(cls) -> None:
1796 # Clean up any lingering SQLAlchemy engines/connections
1797 # so they're closed before we shut down the server.
1798 gc.collect()
1799 cls.postgresql.clear_cache()
1800 super().tearDownClass()
1802 def setUp(self) -> None:
1803 self.server = self.postgresql()
1805 # Need to add a registry section to the config.
1806 self._temp_config = False
1807 config = Config(self.configFile)
1808 config["registry", "db"] = self.server.url()
1809 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1810 config.dump(fh)
1811 self.configFile = fh.name
1812 self._temp_config = True
1813 super().setUp()
1815 def tearDown(self) -> None:
1816 self.server.stop()
1817 if self._temp_config and os.path.exists(self.configFile):
1818 os.remove(self.configFile)
1819 super().tearDown()
1821 def testMakeRepo(self) -> None:
1822 # The base class test assumes that it's using sqlite and assumes
1823 # the config file is acceptable to sqlite.
1824 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1827@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1828class ClonedPostgresPosixDatastoreButlerTestCase(PostgresPosixDatastoreButlerTestCase, unittest.TestCase):
1829 """Test that Butler with a Postgres registry still works after cloning."""
1831 def create_butler(
1832 self, run: str, storageClass: StorageClass | str, datasetTypeName: str
1833 ) -> tuple[DirectButler, DatasetType]:
1834 butler, datasetType = super().create_butler(run, storageClass, datasetTypeName)
1835 return butler._clone(run=run), datasetType
1838class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1839 """InMemoryDatastore specialization of a butler"""
1841 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1842 fullConfigKey = None
1843 useTempRoot = False
1844 validationCanFail = False
1845 datastoreStr = ["datastore='InMemory"]
1846 datastoreName = ["InMemoryDatastore@"]
1847 registryStr = "/gen3.sqlite3"
1849 def testIngest(self) -> None:
1850 pass
1853class ClonedSqliteButlerTestCase(InMemoryDatastoreButlerTestCase, unittest.TestCase):
1854 """Test that a Butler with a Sqlite registry still works after cloning."""
1856 def create_butler(
1857 self, run: str, storageClass: StorageClass | str, datasetTypeName: str
1858 ) -> tuple[DirectButler, DatasetType]:
1859 butler, datasetType = super().create_butler(run, storageClass, datasetTypeName)
1860 return butler._clone(run=run), datasetType
1863class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1864 """PosixDatastore specialization"""
1866 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1867 fullConfigKey = ".datastore.datastores.1.formatters"
1868 validationCanFail = True
1869 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1870 datastoreName = [
1871 "InMemoryDatastore@",
1872 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1873 "SecondDatastore",
1874 ]
1875 registryStr = "/gen3.sqlite3"
1878class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1879 """Test that a yaml file in one location can refer to a root in another."""
1881 datastoreStr = ["dir1"]
1882 # Disable the makeRepo test since we are deliberately not using
1883 # butler.yaml as the config name.
1884 fullConfigKey = None
1886 def setUp(self) -> None:
1887 self.root = makeTestTempDir(TESTDIR)
1889 # Make a new repository in one place
1890 self.dir1 = os.path.join(self.root, "dir1")
1891 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1893 # Move the yaml file to a different place and add a "root"
1894 self.dir2 = os.path.join(self.root, "dir2")
1895 os.makedirs(self.dir2, exist_ok=True)
1896 configFile1 = os.path.join(self.dir1, "butler.yaml")
1897 config = Config(configFile1)
1898 config["root"] = self.dir1
1899 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1900 config.dumpToUri(configFile2)
1901 os.remove(configFile1)
1902 self.tmpConfigFile = configFile2
1904 def testFileLocations(self) -> None:
1905 self.assertNotEqual(self.dir1, self.dir2)
1906 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1907 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1908 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1911class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1912 """Test that a config file created by makeRepo outside of repo works."""
1914 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1916 def setUp(self) -> None:
1917 self.root = makeTestTempDir(TESTDIR)
1918 self.root2 = makeTestTempDir(TESTDIR)
1920 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1921 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1923 def tearDown(self) -> None:
1924 if os.path.exists(self.root2):
1925 shutil.rmtree(self.root2, ignore_errors=True)
1926 super().tearDown()
1928 def testConfigExistence(self) -> None:
1929 c = Config(self.tmpConfigFile)
1930 uri_config = ResourcePath(c["root"])
1931 uri_expected = ResourcePath(self.root, forceDirectory=True)
1932 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1933 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1935 def testPutGet(self) -> None:
1936 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1937 self.runPutGetTest(storageClass, "test_metric")
1940class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase):
1941 """Test that a config file created by makeRepo outside of repo works."""
1943 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1945 def setUp(self) -> None:
1946 self.root = makeTestTempDir(TESTDIR)
1947 self.root2 = makeTestTempDir(TESTDIR)
1949 self.tmpConfigFile = self.root2
1950 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1952 def testConfigExistence(self) -> None:
1953 # Append the yaml file else Config constructor does not know the file
1954 # type.
1955 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1956 super().testConfigExistence()
1959class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1960 """Test that a config file created by makeRepo outside of repo works."""
1962 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1964 def setUp(self) -> None:
1965 self.root = makeTestTempDir(TESTDIR)
1966 self.root2 = makeTestTempDir(TESTDIR)
1968 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1969 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1972@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1973class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1974 """S3Datastore specialization of a butler; an S3 storage Datastore +
1975 a local in-memory SqlRegistry.
1976 """
1978 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1979 fullConfigKey = None
1980 validationCanFail = True
1982 bucketName = "anybucketname"
1983 """Name of the Bucket that will be used in the tests. The name is read from
1984 the config file used with the tests during set-up.
1985 """
1987 root = "butlerRoot/"
1988 """Root repository directory expected to be used in case useTempRoot=False.
1989 Otherwise the root is set to a 20 characters long randomly generated string
1990 during set-up.
1991 """
1993 datastoreStr = [f"datastore={root}"]
1994 """Contains all expected root locations in a format expected to be
1995 returned by Butler stringification.
1996 """
1998 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1999 """The expected format of the S3 Datastore string."""
2001 registryStr = "/gen3.sqlite3"
2002 """Expected format of the Registry string."""
2004 mock_s3 = mock_s3()
2005 """The mocked s3 interface from moto."""
2007 def genRoot(self) -> str:
2008 """Return a random string of len 20 to serve as a root
2009 name for the temporary bucket repo.
2011 This is equivalent to tempfile.mkdtemp as this is what self.root
2012 becomes when useTempRoot is True.
2013 """
2014 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
2015 return rndstr + "/"
2017 def setUp(self) -> None:
2018 config = Config(self.configFile)
2019 uri = ResourcePath(config[".datastore.datastore.root"])
2020 self.bucketName = uri.netloc
2022 # Enable S3 mocking of tests.
2023 self.enterContext(clean_test_environment_for_s3())
2024 self.mock_s3.start()
2026 if self.useTempRoot:
2027 self.root = self.genRoot()
2028 rooturi = f"s3://{self.bucketName}/{self.root}"
2029 config.update({"datastore": {"datastore": {"root": rooturi}}})
2031 # need local folder to store registry database
2032 self.reg_dir = makeTestTempDir(TESTDIR)
2033 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
2035 # MOTO needs to know that we expect Bucket bucketname to exist
2036 # (this used to be the class attribute bucketName)
2037 s3 = boto3.resource("s3")
2038 s3.create_bucket(Bucket=self.bucketName)
2040 self.datastoreStr = [f"datastore='{rooturi}'"]
2041 self.datastoreName = [f"FileDatastore@{rooturi}"]
2042 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
2043 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
2045 def tearDown(self) -> None:
2046 s3 = boto3.resource("s3")
2047 bucket = s3.Bucket(self.bucketName)
2048 try:
2049 bucket.objects.all().delete()
2050 except botocore.exceptions.ClientError as e:
2051 if e.response["Error"]["Code"] == "404":
2052 # the key was not reachable - pass
2053 pass
2054 else:
2055 raise
2057 bucket = s3.Bucket(self.bucketName)
2058 bucket.delete()
2060 # Stop the S3 mock.
2061 self.mock_s3.stop()
2063 if self.reg_dir is not None and os.path.exists(self.reg_dir):
2064 shutil.rmtree(self.reg_dir, ignore_errors=True)
2066 if self.useTempRoot and os.path.exists(self.root):
2067 shutil.rmtree(self.root, ignore_errors=True)
2069 super().tearDown()
2072class PosixDatastoreTransfers(unittest.TestCase):
2073 """Test data transfers between butlers.
2075 Test for different managers. UUID to UUID and integer to integer are
2076 tested. UUID to integer is not supported since we do not currently
2077 want to allow that. Integer to UUID is supported with the caveat
2078 that UUID4 will be generated and this will be incorrect for raw
2079 dataset types. The test ignores that.
2080 """
2082 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
2083 storageClassFactory: StorageClassFactory
2085 @classmethod
2086 def setUpClass(cls) -> None:
2087 cls.storageClassFactory = StorageClassFactory()
2088 cls.storageClassFactory.addFromConfig(cls.configFile)
2090 def setUp(self) -> None:
2091 self.root = makeTestTempDir(TESTDIR)
2092 self.config = Config(self.configFile)
2094 def tearDown(self) -> None:
2095 removeTestTempDir(self.root)
2097 def create_butler(self, manager: str, label: str) -> Butler:
2098 config = Config(self.configFile)
2099 config["registry", "managers", "datasets"] = manager
2100 return Butler.from_config(
2101 Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True
2102 )
2104 def create_butlers(self, manager1: str | None = None, manager2: str | None = None) -> None:
2105 default = "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID"
2106 if manager1 is None:
2107 manager1 = default
2108 if manager2 is None:
2109 manager2 = default
2110 self.source_butler = self.create_butler(manager1, "1")
2111 self.target_butler = self.create_butler(manager2, "2")
2113 def testTransferUuidToUuid(self) -> None:
2114 self.create_butlers()
2115 self.assertButlerTransfers()
2117 def _enable_trust(self, datastore: Datastore) -> None:
2118 datastores = getattr(datastore, "datastores", [datastore])
2119 for this_datastore in datastores:
2120 if hasattr(this_datastore, "trustGetRequest"):
2121 this_datastore.trustGetRequest = True
2123 def testTransferMissing(self) -> None:
2124 """Test transfers where datastore records are missing.
2126 This is how execution butler works.
2127 """
2128 self.create_butlers()
2130 # Configure the source butler to allow trust.
2131 self._enable_trust(self.source_butler._datastore)
2133 self.assertButlerTransfers(purge=True)
2135 def testTransferMissingDisassembly(self) -> None:
2136 """Test transfers where datastore records are missing.
2138 This is how execution butler works.
2139 """
2140 self.create_butlers()
2142 # Configure the source butler to allow trust.
2143 self._enable_trust(self.source_butler._datastore)
2145 # Test disassembly.
2146 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
2148 def testAbsoluteURITransferDirect(self) -> None:
2149 """Test transfer using an absolute URI."""
2150 self._absolute_transfer("auto")
2152 def testAbsoluteURITransferCopy(self) -> None:
2153 """Test transfer using an absolute URI."""
2154 self._absolute_transfer("copy")
2156 def _absolute_transfer(self, transfer: str) -> None:
2157 self.create_butlers()
2159 storageClassName = "StructuredData"
2160 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2161 datasetTypeName = "random_data"
2162 run = "run1"
2163 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2165 dimensions = self.source_butler.dimensions.conform(())
2166 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2167 self.source_butler.registry.registerDatasetType(datasetType)
2169 metrics = makeExampleMetrics()
2170 with ResourcePath.temporary_uri(suffix=".json") as temp:
2171 dataId = DataCoordinate.make_empty(self.source_butler.dimensions)
2172 source_refs = [DatasetRef(datasetType, dataId, run=run)]
2173 temp.write(json.dumps(metrics.exportAsDict()).encode())
2174 dataset = FileDataset(path=temp, refs=source_refs)
2175 self.source_butler.ingest(dataset, transfer="direct")
2177 self.target_butler.transfer_from(
2178 self.source_butler, dataset.refs, register_dataset_types=True, transfer=transfer
2179 )
2181 uri = self.target_butler.getURI(dataset.refs[0])
2182 if transfer == "auto":
2183 self.assertEqual(uri, temp)
2184 else:
2185 self.assertNotEqual(uri, temp)
2187 def assertButlerTransfers(self, purge: bool = False, storageClassName: str = "StructuredData") -> None:
2188 """Test that a run can be transferred to another butler."""
2189 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2190 datasetTypeName = "random_data"
2192 # Test will create 3 collections and we will want to transfer
2193 # two of those three.
2194 runs = ["run1", "run2", "other"]
2196 # Also want to use two different dataset types to ensure that
2197 # grouping works.
2198 datasetTypeNames = ["random_data", "random_data_2"]
2200 # Create the run collections in the source butler.
2201 for run in runs:
2202 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2204 # Create dimensions in source butler.
2205 n_exposures = 30
2206 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
2207 self.source_butler.registry.insertDimensionData(
2208 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
2209 )
2210 self.source_butler.registry.insertDimensionData(
2211 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
2212 )
2214 for i in range(n_exposures):
2215 self.source_butler.registry.insertDimensionData(
2216 "exposure",
2217 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
2218 )
2220 # Create dataset types in the source butler.
2221 dimensions = self.source_butler.dimensions.conform(["instrument", "exposure"])
2222 for datasetTypeName in datasetTypeNames:
2223 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2224 self.source_butler.registry.registerDatasetType(datasetType)
2226 # Write a dataset to an unrelated run -- this will ensure that
2227 # we are rewriting integer dataset ids in the target if necessary.
2228 # Will not be relevant for UUID.
2229 run = "distraction"
2230 butler = Butler.from_config(butler=self.source_butler, run=run)
2231 butler.put(
2232 makeExampleMetrics(),
2233 datasetTypeName,
2234 exposure=1,
2235 instrument="DummyCamComp",
2236 physical_filter="d-r",
2237 )
2239 # Write some example metrics to the source
2240 butler = Butler.from_config(butler=self.source_butler)
2242 # Set of DatasetRefs that should be in the list of refs to transfer
2243 # but which will not be transferred.
2244 deleted: set[DatasetRef] = set()
2246 n_expected = 20 # Number of datasets expected to be transferred
2247 source_refs = []
2248 for i in range(n_exposures):
2249 # Put a third of datasets into each collection, only retain
2250 # two thirds.
2251 index = i % 3
2252 run = runs[index]
2253 datasetTypeName = datasetTypeNames[i % 2]
2255 metric = MetricsExample(
2256 summary={"counter": i}, output={"text": "metric"}, data=[2 * x for x in range(i)]
2257 )
2258 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2259 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2261 # Remove the datastore record using low-level API, but only
2262 # for a specific index.
2263 if purge and index == 1:
2264 # For one of these delete the file as well.
2265 # This allows the "missing" code to filter the
2266 # file out.
2267 # Access the individual datastores.
2268 datastores = []
2269 if hasattr(butler._datastore, "datastores"):
2270 datastores.extend(butler._datastore.datastores)
2271 else:
2272 datastores.append(butler._datastore)
2274 if not deleted:
2275 # For a chained datastore we need to remove
2276 # files in each chain.
2277 for datastore in datastores:
2278 # The file might not be known to the datastore
2279 # if constraints are used.
2280 try:
2281 primary, uris = datastore.getURIs(ref)
2282 except FileNotFoundError:
2283 continue
2284 if primary and primary.scheme != "mem":
2285 primary.remove()
2286 for uri in uris.values():
2287 if uri.scheme != "mem":
2288 uri.remove()
2289 n_expected -= 1
2290 deleted.add(ref)
2292 # Remove the datastore record.
2293 for datastore in datastores:
2294 if hasattr(datastore, "removeStoredItemInfo"):
2295 datastore.removeStoredItemInfo(ref)
2297 if index < 2:
2298 source_refs.append(ref)
2299 if ref not in deleted:
2300 new_metric = butler.get(ref)
2301 self.assertEqual(new_metric, metric)
2303 # Create some bad dataset types to ensure we check for inconsistent
2304 # definitions.
2305 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2306 for datasetTypeName in datasetTypeNames:
2307 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2308 self.target_butler.registry.registerDatasetType(datasetType)
2309 with self.assertRaises(ConflictingDefinitionError) as cm:
2310 self.target_butler.transfer_from(self.source_butler, source_refs)
2311 self.assertIn("dataset type differs", str(cm.exception))
2313 # And remove the bad definitions.
2314 for datasetTypeName in datasetTypeNames:
2315 self.target_butler.registry.removeDatasetType(datasetTypeName)
2317 # Transfer without creating dataset types should fail.
2318 with self.assertRaises(KeyError):
2319 self.target_butler.transfer_from(self.source_butler, source_refs)
2321 # Transfer without creating dimensions should fail.
2322 with self.assertRaises(ConflictingDefinitionError) as cm:
2323 self.target_butler.transfer_from(self.source_butler, source_refs, register_dataset_types=True)
2324 self.assertIn("dimension", str(cm.exception))
2326 # The failed transfer above leaves registry in an inconsistent
2327 # state because the run is created but then rolled back without
2328 # the collection cache being cleared. For now force a refresh.
2329 # Can remove with DM-35498.
2330 self.target_butler.registry.refresh()
2332 # Do a dry run -- this should not have any effect on the target butler.
2333 self.target_butler.transfer_from(self.source_butler, source_refs, dry_run=True)
2335 # Transfer the records for one ref to test the alternative API.
2336 with self.assertLogs(logger="lsst", level=logging.DEBUG) as log_cm:
2337 self.target_butler.transfer_dimension_records_from(self.source_butler, [source_refs[0]])
2338 self.assertIn("number of records transferred: 1", ";".join(log_cm.output))
2340 # Now transfer them to the second butler, including dimensions.
2341 with self.assertLogs(logger="lsst", level=logging.DEBUG) as log_cm:
2342 transferred = self.target_butler.transfer_from(
2343 self.source_butler,
2344 source_refs,
2345 register_dataset_types=True,
2346 transfer_dimensions=True,
2347 )
2348 self.assertEqual(len(transferred), n_expected)
2349 log_output = ";".join(log_cm.output)
2351 # A ChainedDatastore will use the in-memory datastore for mexists
2352 # so we can not rely on the mexists log message.
2353 self.assertIn("Number of datastore records found in source", log_output)
2354 self.assertIn("Creating output run", log_output)
2356 # Do the transfer twice to ensure that it will do nothing extra.
2357 # Only do this if purge=True because it does not work for int
2358 # dataset_id.
2359 if purge:
2360 # This should not need to register dataset types.
2361 transferred = self.target_butler.transfer_from(self.source_butler, source_refs)
2362 self.assertEqual(len(transferred), n_expected)
2364 # Also do an explicit low-level transfer to trigger some
2365 # edge cases.
2366 with self.assertLogs(level=logging.DEBUG) as log_cm:
2367 self.target_butler._datastore.transfer_from(self.source_butler._datastore, source_refs)
2368 log_output = ";".join(log_cm.output)
2369 self.assertIn("no file artifacts exist", log_output)
2371 with self.assertRaises((TypeError, AttributeError)):
2372 self.target_butler._datastore.transfer_from(self.source_butler, source_refs) # type: ignore
2374 with self.assertRaises(ValueError):
2375 self.target_butler._datastore.transfer_from(
2376 self.source_butler._datastore, source_refs, transfer="split"
2377 )
2379 # Now try to get the same refs from the new butler.
2380 for ref in source_refs:
2381 if ref not in deleted:
2382 new_metric = self.target_butler.get(ref)
2383 old_metric = self.source_butler.get(ref)
2384 self.assertEqual(new_metric, old_metric)
2386 # Now prune run2 collection and create instead a CHAINED collection.
2387 # This should block the transfer.
2388 self.target_butler.removeRuns(["run2"], unstore=True)
2389 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2390 with self.assertRaises(CollectionTypeError):
2391 # Re-importing the run1 datasets can be problematic if they
2392 # use integer IDs so filter those out.
2393 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2394 self.target_butler.transfer_from(self.source_butler, to_transfer)
2397class ChainedDatastoreTransfers(PosixDatastoreTransfers):
2398 """Test transfers using a chained datastore."""
2400 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
2403class NullDatastoreTestCase(unittest.TestCase):
2404 """Test that we can fall back to a null datastore."""
2406 # Need a good config to create the repo.
2407 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
2408 storageClassFactory: StorageClassFactory
2410 @classmethod
2411 def setUpClass(cls) -> None:
2412 cls.storageClassFactory = StorageClassFactory()
2413 cls.storageClassFactory.addFromConfig(cls.configFile)
2415 def setUp(self) -> None:
2416 """Create a new butler root for each test."""
2417 self.root = makeTestTempDir(TESTDIR)
2418 Butler.makeRepo(self.root, config=Config(self.configFile))
2420 def tearDown(self) -> None:
2421 removeTestTempDir(self.root)
2423 def test_fallback(self) -> None:
2424 # Read the butler config and mess with the datastore section.
2425 config_path = os.path.join(self.root, "butler.yaml")
2426 bad_config = Config(config_path)
2427 bad_config["datastore", "cls"] = "lsst.not.a.datastore.Datastore"
2428 bad_config.dumpToUri(config_path)
2430 with self.assertRaises(RuntimeError):
2431 Butler(self.root, without_datastore=False)
2433 with self.assertRaises(RuntimeError):
2434 Butler.from_config(self.root, without_datastore=False)
2436 butler = Butler.from_config(self.root, writeable=True, without_datastore=True)
2437 self.assertIsInstance(butler._datastore, NullDatastore)
2439 # Check that registry is working.
2440 butler.registry.registerRun("MYRUN")
2441 collections = butler.registry.queryCollections(...)
2442 self.assertIn("MYRUN", set(collections))
2444 # Create a ref.
2445 dimensions = butler.dimensions.conform([])
2446 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
2447 datasetTypeName = "metric"
2448 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2449 butler.registry.registerDatasetType(datasetType)
2450 ref = DatasetRef(datasetType, {}, run="MYRUN")
2452 # Check that datastore will complain.
2453 with self.assertRaises(FileNotFoundError):
2454 butler.get(ref)
2455 with self.assertRaises(FileNotFoundError):
2456 butler.getURI(ref)
2459def setup_module(module: types.ModuleType) -> None:
2460 """Set up the module for pytest."""
2461 clean_environment()
2464if __name__ == "__main__":
2465 clean_environment()
2466 unittest.main()