Coverage for tests/test_butler.py: 13%
1311 statements
« prev ^ index » next coverage.py v7.3.2, created at 2023-10-27 09:44 +0000
« prev ^ index » next coverage.py v7.3.2, created at 2023-10-27 09:44 +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 setAwsEnvCredentials, unsetAwsEnvCredentials
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, DimensionGraph, 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 (
118 "DAF_BUTLER_REPOSITORY_INDEX",
119 "S3_ENDPOINT_URL",
120 "AWS_ACCESS_KEY_ID",
121 "AWS_SECRET_ACCESS_KEY",
122 "AWS_SHARED_CREDENTIALS_FILE",
123 ):
124 os.environ.pop(k, None)
127def makeExampleMetrics() -> MetricsExample:
128 """Return example dataset suitable for tests."""
129 return MetricsExample(
130 {"AM1": 5.2, "AM2": 30.6},
131 {"a": [1, 2, 3], "b": {"blue": 5, "red": "green"}},
132 [563, 234, 456.7, 752, 8, 9, 27],
133 )
136class TransactionTestError(Exception):
137 """Specific error for testing transactions, to prevent misdiagnosing
138 that might otherwise occur when a standard exception is used.
139 """
141 pass
144class ButlerConfigTests(unittest.TestCase):
145 """Simple tests for ButlerConfig that are not tested in any other test
146 cases.
147 """
149 def testSearchPath(self) -> None:
150 configFile = os.path.join(TESTDIR, "config", "basic", "butler.yaml")
151 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
152 config1 = ButlerConfig(configFile)
153 self.assertNotIn("testConfigs", "\n".join(cm.output))
155 overrideDirectory = os.path.join(TESTDIR, "config", "testConfigs")
156 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
157 config2 = ButlerConfig(configFile, searchPaths=[overrideDirectory])
158 self.assertIn("testConfigs", "\n".join(cm.output))
160 key = ("datastore", "records", "table")
161 self.assertNotEqual(config1[key], config2[key])
162 self.assertEqual(config2[key], "override_record")
165class ButlerPutGetTests(TestCaseMixin):
166 """Helper method for running a suite of put/get tests from different
167 butler configurations.
168 """
170 root: str
171 default_run = "ingésτ😺"
172 storageClassFactory: StorageClassFactory
173 configFile: str
174 tmpConfigFile: str
176 @staticmethod
177 def addDatasetType(
178 datasetTypeName: str, dimensions: DimensionGraph, storageClass: StorageClass | str, registry: Registry
179 ) -> DatasetType:
180 """Create a DatasetType and register it"""
181 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
182 registry.registerDatasetType(datasetType)
183 return datasetType
185 @classmethod
186 def setUpClass(cls) -> None:
187 cls.storageClassFactory = StorageClassFactory()
188 cls.storageClassFactory.addFromConfig(cls.configFile)
190 def assertGetComponents(
191 self,
192 butler: Butler,
193 datasetRef: DatasetRef,
194 components: tuple[str, ...],
195 reference: Any,
196 collections: Any = None,
197 ) -> None:
198 datasetType = datasetRef.datasetType
199 dataId = datasetRef.dataId
200 deferred = butler.getDeferred(datasetRef)
202 for component in components:
203 compTypeName = datasetType.componentTypeName(component)
204 result = butler.get(compTypeName, dataId, collections=collections)
205 self.assertEqual(result, getattr(reference, component))
206 result_deferred = deferred.get(component=component)
207 self.assertEqual(result_deferred, result)
209 def tearDown(self) -> None:
210 removeTestTempDir(self.root)
212 def create_butler(
213 self, run: str, storageClass: StorageClass | str, datasetTypeName: str
214 ) -> tuple[DirectButler, DatasetType]:
215 butler = Butler.from_config(self.tmpConfigFile, run=run)
216 assert isinstance(butler, DirectButler), "Expect DirectButler in configuration"
218 collections = set(butler.registry.queryCollections())
219 self.assertEqual(collections, {run})
221 # Create and register a DatasetType
222 dimensions = butler.dimensions.extract(["instrument", "visit"])
224 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
226 # Add needed Dimensions
227 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
228 butler.registry.insertDimensionData(
229 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
230 )
231 butler.registry.insertDimensionData(
232 "visit_system", {"instrument": "DummyCamComp", "id": 1, "name": "default"}
233 )
234 visit_start = astropy.time.Time("2020-01-01 08:00:00.123456789", scale="tai")
235 visit_end = astropy.time.Time("2020-01-01 08:00:36.66", scale="tai")
236 butler.registry.insertDimensionData(
237 "visit",
238 {
239 "instrument": "DummyCamComp",
240 "id": 423,
241 "name": "fourtwentythree",
242 "physical_filter": "d-r",
243 "visit_system": 1,
244 "datetime_begin": visit_start,
245 "datetime_end": visit_end,
246 },
247 )
249 # Add more visits for some later tests
250 for visit_id in (424, 425):
251 butler.registry.insertDimensionData(
252 "visit",
253 {
254 "instrument": "DummyCamComp",
255 "id": visit_id,
256 "name": f"fourtwentyfour_{visit_id}",
257 "physical_filter": "d-r",
258 "visit_system": 1,
259 },
260 )
261 return butler, datasetType
263 def runPutGetTest(self, storageClass: StorageClass, datasetTypeName: str) -> DirectButler:
264 # New datasets will be added to run and tag, but we will only look in
265 # tag when looking up datasets.
266 run = self.default_run
267 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
268 assert butler.run is not None
270 # Create and store a dataset
271 metric = makeExampleMetrics()
272 dataId = butler.registry.expandDataId({"instrument": "DummyCamComp", "visit": 423})
274 # Put and remove the dataset once as a DatasetRef, once as a dataId,
275 # and once with a DatasetType
277 # Keep track of any collections we add and do not clean up
278 expected_collections = {run}
280 counter = 0
281 ref = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run="put_run_1")
282 args = tuple[DatasetRef] | tuple[str | DatasetType, DataCoordinate]
283 for args in ((ref,), (datasetTypeName, dataId), (datasetType, dataId)):
284 # Since we are using subTest we can get cascading failures
285 # here with the first attempt failing and the others failing
286 # immediately because the dataset already exists. Work around
287 # this by using a distinct run collection each time
288 counter += 1
289 this_run = f"put_run_{counter}"
290 butler.registry.registerCollection(this_run, type=CollectionType.RUN)
291 expected_collections.update({this_run})
293 with self.subTest(args=args):
294 kwargs: dict[str, Any] = {}
295 if not isinstance(args[0], DatasetRef): # type: ignore
296 kwargs["run"] = this_run
297 ref = butler.put(metric, *args, **kwargs)
298 self.assertIsInstance(ref, DatasetRef)
300 # Test getDirect
301 metricOut = butler.get(ref)
302 self.assertEqual(metric, metricOut)
303 # Test get
304 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
305 self.assertEqual(metric, metricOut)
306 # Test get with a datasetRef
307 metricOut = butler.get(ref)
308 self.assertEqual(metric, metricOut)
309 # Test getDeferred with dataId
310 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
311 self.assertEqual(metric, metricOut)
312 # Test getDeferred with a ref
313 metricOut = butler.getDeferred(ref).get()
314 self.assertEqual(metric, metricOut)
316 # Check we can get components
317 if storageClass.isComposite():
318 self.assertGetComponents(
319 butler, ref, ("summary", "data", "output"), metric, collections=this_run
320 )
322 # Can the artifacts themselves be retrieved?
323 if not butler._datastore.isEphemeral:
324 root_uri = ResourcePath(self.root)
326 for preserve_path in (True, False):
327 destination = root_uri.join(f"artifacts/{preserve_path}_{counter}/")
328 # Use copy so that we can test that overwrite
329 # protection works (using "auto" for File URIs would
330 # use hard links and subsequent transfer would work
331 # because it knows they are the same file).
332 transferred = butler.retrieveArtifacts(
333 [ref], destination, preserve_path=preserve_path, transfer="copy"
334 )
335 self.assertGreater(len(transferred), 0)
336 artifacts = list(ResourcePath.findFileResources([destination]))
337 self.assertEqual(set(transferred), set(artifacts))
339 for artifact in transferred:
340 path_in_destination = artifact.relative_to(destination)
341 self.assertIsNotNone(path_in_destination)
342 assert path_in_destination is not None
344 # when path is not preserved there should not be
345 # any path separators.
346 num_seps = path_in_destination.count("/")
347 if preserve_path:
348 self.assertGreater(num_seps, 0)
349 else:
350 self.assertEqual(num_seps, 0)
352 primary_uri, secondary_uris = butler.getURIs(ref)
353 n_uris = len(secondary_uris)
354 if primary_uri:
355 n_uris += 1
356 self.assertEqual(
357 len(artifacts),
358 n_uris,
359 "Comparing expected artifacts vs actual:"
360 f" {artifacts} vs {primary_uri} and {secondary_uris}",
361 )
363 if preserve_path:
364 # No need to run these twice
365 with self.assertRaises(ValueError):
366 butler.retrieveArtifacts([ref], destination, transfer="move")
368 with self.assertRaises(FileExistsError):
369 butler.retrieveArtifacts([ref], destination)
371 transferred_again = butler.retrieveArtifacts(
372 [ref], destination, preserve_path=preserve_path, overwrite=True
373 )
374 self.assertEqual(set(transferred_again), set(transferred))
376 # Now remove the dataset completely.
377 butler.pruneDatasets([ref], purge=True, unstore=True)
378 # Lookup with original args should still fail.
379 kwargs = {"collections": this_run}
380 if isinstance(args[0], DatasetRef):
381 kwargs = {} # Prevent warning from being issued.
382 self.assertFalse(butler.exists(*args, **kwargs))
383 # get() should still fail.
384 with self.assertRaises(FileNotFoundError):
385 butler.get(ref)
386 # Registry shouldn't be able to find it by dataset_id anymore.
387 self.assertIsNone(butler.registry.getDataset(ref.id))
389 # Do explicit registry removal since we know they are
390 # empty
391 butler.registry.removeCollection(this_run)
392 expected_collections.remove(this_run)
394 # Create DatasetRef for put using default run.
395 refIn = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run=butler.run)
397 # Check that getDeferred fails with standalone ref.
398 with self.assertRaises(LookupError):
399 butler.getDeferred(refIn)
401 # Put the dataset again, since the last thing we did was remove it
402 # and we want to use the default collection.
403 ref = butler.put(metric, refIn)
405 # Get with parameters
406 stop = 4
407 sliced = butler.get(ref, parameters={"slice": slice(stop)})
408 self.assertNotEqual(metric, sliced)
409 self.assertEqual(metric.summary, sliced.summary)
410 self.assertEqual(metric.output, sliced.output)
411 assert metric.data is not None # for mypy
412 self.assertEqual(metric.data[:stop], sliced.data)
413 # getDeferred with parameters
414 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
415 self.assertNotEqual(metric, sliced)
416 self.assertEqual(metric.summary, sliced.summary)
417 self.assertEqual(metric.output, sliced.output)
418 self.assertEqual(metric.data[:stop], sliced.data)
419 # getDeferred with deferred parameters
420 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
421 self.assertNotEqual(metric, sliced)
422 self.assertEqual(metric.summary, sliced.summary)
423 self.assertEqual(metric.output, sliced.output)
424 self.assertEqual(metric.data[:stop], sliced.data)
426 if storageClass.isComposite():
427 # Check that components can be retrieved
428 metricOut = butler.get(ref.datasetType.name, dataId)
429 compNameS = ref.datasetType.componentTypeName("summary")
430 compNameD = ref.datasetType.componentTypeName("data")
431 summary = butler.get(compNameS, dataId)
432 self.assertEqual(summary, metric.summary)
433 data = butler.get(compNameD, dataId)
434 self.assertEqual(data, metric.data)
436 if "counter" in storageClass.derivedComponents:
437 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
438 self.assertEqual(count, len(data))
440 count = butler.get(
441 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
442 )
443 self.assertEqual(count, stop)
445 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections)
446 assert compRef is not None
447 summary = butler.get(compRef)
448 self.assertEqual(summary, metric.summary)
450 # Create a Dataset type that has the same name but is inconsistent.
451 inconsistentDatasetType = DatasetType(
452 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
453 )
455 # Getting with a dataset type that does not match registry fails
456 with self.assertRaisesRegex(ValueError, "Supplied dataset type .* inconsistent with registry"):
457 butler.get(inconsistentDatasetType, dataId)
459 # Combining a DatasetRef with a dataId should fail
460 with self.assertRaisesRegex(ValueError, "DatasetRef given, cannot use dataId as well"):
461 butler.get(ref, dataId)
462 # Getting with an explicit ref should fail if the id doesn't match.
463 with self.assertRaises(FileNotFoundError):
464 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=uuid.UUID(int=101), run=butler.run))
466 # Getting a dataset with unknown parameters should fail
467 with self.assertRaisesRegex(KeyError, "Parameter 'unsupported' not understood"):
468 butler.get(ref, parameters={"unsupported": True})
470 # Check we have a collection
471 collections = set(butler.registry.queryCollections())
472 self.assertEqual(collections, expected_collections)
474 # Clean up to check that we can remove something that may have
475 # already had a component removed
476 butler.pruneDatasets([ref], unstore=True, purge=True)
478 # Add the same ref again, so we can check that duplicate put fails.
479 ref = butler.put(metric, datasetType, dataId)
481 # Repeat put will fail.
482 with self.assertRaisesRegex(
483 ConflictingDefinitionError, "A database constraint failure was triggered"
484 ):
485 butler.put(metric, datasetType, dataId)
487 # Remove the datastore entry.
488 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
490 # Put will still fail
491 with self.assertRaisesRegex(
492 ConflictingDefinitionError, "A database constraint failure was triggered"
493 ):
494 butler.put(metric, datasetType, dataId)
496 # Repeat the same sequence with resolved ref.
497 butler.pruneDatasets([ref], unstore=True, purge=True)
498 ref = butler.put(metric, refIn)
500 # Repeat put will fail.
501 with self.assertRaisesRegex(ConflictingDefinitionError, "Datastore already contains dataset"):
502 butler.put(metric, refIn)
504 # Remove the datastore entry.
505 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
507 # In case of resolved ref this write will succeed.
508 ref = butler.put(metric, refIn)
510 # Leave the dataset in place since some downstream tests require
511 # something to be present
513 return butler
515 def testDeferredCollectionPassing(self) -> None:
516 # Construct a butler with no run or collection, but make it writeable.
517 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
518 # Create and register a DatasetType
519 dimensions = butler.dimensions.extract(["instrument", "visit"])
520 datasetType = self.addDatasetType(
521 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
522 )
523 # Add needed Dimensions
524 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
525 butler.registry.insertDimensionData(
526 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
527 )
528 butler.registry.insertDimensionData(
529 "visit",
530 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
531 )
532 dataId = {"instrument": "DummyCamComp", "visit": 423}
533 # Create dataset.
534 metric = makeExampleMetrics()
535 # Register a new run and put dataset.
536 run = "deferred"
537 self.assertTrue(butler.registry.registerRun(run))
538 # Second time it will be allowed but indicate no-op
539 self.assertFalse(butler.registry.registerRun(run))
540 ref = butler.put(metric, datasetType, dataId, run=run)
541 # Putting with no run should fail with TypeError.
542 with self.assertRaises(CollectionError):
543 butler.put(metric, datasetType, dataId)
544 # Dataset should exist.
545 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
546 # We should be able to get the dataset back, but with and without
547 # a deferred dataset handle.
548 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
549 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
550 # Trying to find the dataset without any collection is a TypeError.
551 self.assertFalse(butler.exists(datasetType, dataId))
552 with self.assertRaises(CollectionError):
553 butler.get(datasetType, dataId)
554 # Associate the dataset with a different collection.
555 butler.registry.registerCollection("tagged")
556 butler.registry.associate("tagged", [ref])
557 # Deleting the dataset from the new collection should make it findable
558 # in the original collection.
559 butler.pruneDatasets([ref], tags=["tagged"])
560 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
563class ButlerTests(ButlerPutGetTests):
564 """Tests for Butler."""
566 useTempRoot = True
567 validationCanFail: bool
568 fullConfigKey: str | None
569 registryStr: str | None
570 datastoreName: list[str] | None
571 datastoreStr: list[str]
573 def setUp(self) -> None:
574 """Create a new butler root for each test."""
575 self.root = makeTestTempDir(TESTDIR)
576 Butler.makeRepo(self.root, config=Config(self.configFile))
577 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
579 def testConstructor(self) -> None:
580 """Independent test of constructor."""
581 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
582 self.assertIsInstance(butler, Butler)
584 # Check that butler.yaml is added automatically.
585 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
586 config_dir = self.tmpConfigFile[: -len(end)]
587 butler = Butler.from_config(config_dir, run=self.default_run)
588 self.assertIsInstance(butler, Butler)
590 # Even with a ResourcePath.
591 butler = Butler.from_config(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
592 self.assertIsInstance(butler, Butler)
594 collections = set(butler.registry.queryCollections())
595 self.assertEqual(collections, {self.default_run})
597 # Check that some special characters can be included in run name.
598 special_run = "u@b.c-A"
599 butler_special = Butler.from_config(butler=butler, run=special_run)
600 collections = set(butler_special.registry.queryCollections("*@*"))
601 self.assertEqual(collections, {special_run})
603 butler2 = Butler.from_config(butler=butler, collections=["other"])
604 self.assertEqual(butler2.collections, ("other",))
605 self.assertIsNone(butler2.run)
606 self.assertIs(butler._datastore, butler2._datastore)
608 # Test that we can use an environment variable to find this
609 # repository.
610 butler_index = Config()
611 butler_index["label"] = self.tmpConfigFile
612 for suffix in (".yaml", ".json"):
613 # Ensure that the content differs so that we know that
614 # we aren't reusing the cache.
615 bad_label = f"file://bucket/not_real{suffix}"
616 butler_index["bad_label"] = bad_label
617 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
618 butler_index.dumpToUri(temp_file)
619 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
620 self.assertEqual(Butler.get_known_repos(), {"label", "bad_label"})
621 uri = Butler.get_repo_uri("bad_label")
622 self.assertEqual(uri, ResourcePath(bad_label))
623 uri = Butler.get_repo_uri("label")
624 butler = Butler.from_config(uri, writeable=False)
625 self.assertIsInstance(butler, Butler)
626 butler = Butler.from_config("label", writeable=False)
627 self.assertIsInstance(butler, Butler)
628 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
629 Butler.from_config("not_there", writeable=False)
630 with self.assertRaisesRegex(FileNotFoundError, "resolved from alias 'bad_label'"):
631 Butler.from_config("bad_label")
632 with self.assertRaises(FileNotFoundError):
633 # Should ignore aliases.
634 Butler.from_config(ResourcePath("label", forceAbsolute=False))
635 with self.assertRaises(KeyError) as cm:
636 Butler.get_repo_uri("missing")
637 self.assertEqual(
638 Butler.get_repo_uri("missing", True), ResourcePath("missing", forceAbsolute=False)
639 )
640 self.assertIn("not known to", str(cm.exception))
641 # Should report no failure.
642 self.assertEqual(ButlerRepoIndex.get_failure_reason(), "")
643 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
644 # Now with empty configuration.
645 butler_index = Config()
646 butler_index.dumpToUri(temp_file)
647 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
648 with self.assertRaisesRegex(FileNotFoundError, "(no known aliases)"):
649 Butler.from_config("label")
650 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
651 # Now with bad contents.
652 with open(temp_file.ospath, "w") as fh:
653 print("'", file=fh)
654 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
655 with self.assertRaisesRegex(FileNotFoundError, "(no known aliases:.*could not be read)"):
656 Butler.from_config("label")
657 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
658 with self.assertRaises(FileNotFoundError):
659 Butler.get_repo_uri("label")
660 self.assertEqual(Butler.get_known_repos(), set())
662 with self.assertRaisesRegex(FileNotFoundError, "index file not found"):
663 Butler.from_config("label")
665 # Check that we can create Butler when the alias file is not found.
666 butler = Butler.from_config(self.tmpConfigFile, writeable=False)
667 self.assertIsInstance(butler, Butler)
668 with self.assertRaises(KeyError) as cm:
669 # No environment variable set.
670 Butler.get_repo_uri("label")
671 self.assertEqual(Butler.get_repo_uri("label", True), ResourcePath("label", forceAbsolute=False))
672 self.assertIn("No repository index defined", str(cm.exception))
673 with self.assertRaisesRegex(FileNotFoundError, "no known aliases.*No repository index"):
674 # No aliases registered.
675 Butler.from_config("not_there")
676 self.assertEqual(Butler.get_known_repos(), set())
678 def testBasicPutGet(self) -> None:
679 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
680 self.runPutGetTest(storageClass, "test_metric")
682 def testCompositePutGetConcrete(self) -> None:
683 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
684 butler = self.runPutGetTest(storageClass, "test_metric")
686 # Should *not* be disassembled
687 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
688 self.assertEqual(len(datasets), 1)
689 uri, components = butler.getURIs(datasets[0])
690 self.assertIsInstance(uri, ResourcePath)
691 self.assertFalse(components)
692 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
693 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
695 # Predicted dataset
696 dataId = {"instrument": "DummyCamComp", "visit": 424}
697 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
698 self.assertFalse(components)
699 self.assertIsInstance(uri, ResourcePath)
700 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
701 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
703 def testCompositePutGetVirtual(self) -> None:
704 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
705 butler = self.runPutGetTest(storageClass, "test_metric_comp")
707 # Should be disassembled
708 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
709 self.assertEqual(len(datasets), 1)
710 uri, components = butler.getURIs(datasets[0])
712 if butler._datastore.isEphemeral:
713 # Never disassemble in-memory datastore
714 self.assertIsInstance(uri, ResourcePath)
715 self.assertFalse(components)
716 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
717 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
718 else:
719 self.assertIsNone(uri)
720 self.assertEqual(set(components), set(storageClass.components))
721 for compuri in components.values():
722 self.assertIsInstance(compuri, ResourcePath)
723 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
724 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
726 # Predicted dataset
727 dataId = {"instrument": "DummyCamComp", "visit": 424}
728 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
730 if butler._datastore.isEphemeral:
731 # Never disassembled
732 self.assertIsInstance(uri, ResourcePath)
733 self.assertFalse(components)
734 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
735 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
736 else:
737 self.assertIsNone(uri)
738 self.assertEqual(set(components), set(storageClass.components))
739 for compuri in components.values():
740 self.assertIsInstance(compuri, ResourcePath)
741 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
742 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
744 def testStorageClassOverrideGet(self) -> None:
745 """Test storage class conversion on get with override."""
746 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
747 datasetTypeName = "anything"
748 run = self.default_run
750 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
752 # Create and store a dataset.
753 metric = makeExampleMetrics()
754 dataId = {"instrument": "DummyCamComp", "visit": 423}
756 ref = butler.put(metric, datasetType, dataId)
758 # Return native type.
759 retrieved = butler.get(ref)
760 self.assertEqual(retrieved, metric)
762 # Specify an override.
763 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
764 model = butler.get(ref, storageClass=new_sc)
765 self.assertNotEqual(type(model), type(retrieved))
766 self.assertIs(type(model), new_sc.pytype)
767 self.assertEqual(retrieved, model)
769 # Defer but override later.
770 deferred = butler.getDeferred(ref)
771 model = deferred.get(storageClass=new_sc)
772 self.assertIs(type(model), new_sc.pytype)
773 self.assertEqual(retrieved, model)
775 # Defer but override up front.
776 deferred = butler.getDeferred(ref, storageClass=new_sc)
777 model = deferred.get()
778 self.assertIs(type(model), new_sc.pytype)
779 self.assertEqual(retrieved, model)
781 # Retrieve a component. Should be a tuple.
782 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
783 self.assertIs(type(data), tuple)
784 self.assertEqual(data, tuple(retrieved.data))
786 # Parameter on the write storage class should work regardless
787 # of read storage class.
788 data = butler.get(
789 "anything.data",
790 dataId,
791 storageClass="StructuredDataDataTestTuple",
792 parameters={"slice": slice(2, 4)},
793 )
794 self.assertEqual(len(data), 2)
796 # Try a parameter that is known to the read storage class but not
797 # the write storage class.
798 with self.assertRaises(KeyError):
799 butler.get(
800 "anything.data",
801 dataId,
802 storageClass="StructuredDataDataTestTuple",
803 parameters={"xslice": slice(2, 4)},
804 )
806 def testPytypePutCoercion(self) -> None:
807 """Test python type coercion on Butler.get and put."""
808 # Store some data with the normal example storage class.
809 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
810 datasetTypeName = "test_metric"
811 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
813 dataId = {"instrument": "DummyCamComp", "visit": 423}
815 # Put a dict and this should coerce to a MetricsExample
816 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
817 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
818 test_metric = butler.get(metric_ref)
819 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
820 self.assertEqual(test_metric.summary, test_dict["summary"])
821 self.assertEqual(test_metric.output, test_dict["output"])
823 # Check that the put still works if a DatasetType is given with
824 # a definition matching this python type.
825 registry_type = butler.registry.getDatasetType(datasetTypeName)
826 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
827 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
828 self.assertEqual(metric2_ref.datasetType, registry_type)
830 # The get will return the type expected by registry.
831 test_metric2 = butler.get(metric2_ref)
832 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
834 # Make a new DatasetRef with the compatible but different DatasetType.
835 # This should now return a dict.
836 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
837 test_dict2 = butler.get(new_ref)
838 self.assertEqual(get_full_type_name(test_dict2), "dict")
840 # Get it again with the wrong dataset type definition using get()
841 # rather than get(). This should be consistent with get()
842 # behavior and return the type of the DatasetType.
843 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
844 self.assertEqual(get_full_type_name(test_dict3), "dict")
846 def testIngest(self) -> None:
847 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
849 # Create and register a DatasetType
850 dimensions = butler.dimensions.extract(["instrument", "visit", "detector"])
852 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
853 datasetTypeName = "metric"
855 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
857 # Add needed Dimensions
858 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
859 butler.registry.insertDimensionData(
860 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
861 )
862 for detector in (1, 2):
863 butler.registry.insertDimensionData(
864 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
865 )
867 butler.registry.insertDimensionData(
868 "visit",
869 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
870 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
871 )
873 formatter = doImportType("lsst.daf.butler.formatters.yaml.YamlFormatter")
874 dataRoot = os.path.join(TESTDIR, "data", "basic")
875 datasets = []
876 for detector in (1, 2):
877 detector_name = f"detector_{detector}"
878 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
879 dataId = butler.registry.expandDataId(
880 {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
881 )
882 # Create a DatasetRef for ingest
883 refIn = DatasetRef(datasetType, dataId, run=self.default_run)
885 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
887 butler.ingest(*datasets, transfer="copy")
889 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
890 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
892 metrics1 = butler.get(datasetTypeName, dataId1)
893 metrics2 = butler.get(datasetTypeName, dataId2)
894 self.assertNotEqual(metrics1, metrics2)
896 # Compare URIs
897 uri1 = butler.getURI(datasetTypeName, dataId1)
898 uri2 = butler.getURI(datasetTypeName, dataId2)
899 self.assertNotEqual(uri1, uri2)
901 # Now do a multi-dataset but single file ingest
902 metricFile = os.path.join(dataRoot, "detectors.yaml")
903 refs = []
904 for detector in (1, 2):
905 detector_name = f"detector_{detector}"
906 dataId = butler.registry.expandDataId(
907 {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
908 )
909 # Create a DatasetRef for ingest
910 refs.append(DatasetRef(datasetType, dataId, run=self.default_run))
912 # Test "move" transfer to ensure that the files themselves
913 # have disappeared following ingest.
914 with ResourcePath.temporary_uri(suffix=".yaml") as tempFile:
915 tempFile.transfer_from(ResourcePath(metricFile), transfer="copy")
917 datasets = []
918 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter))
920 # For first ingest use copy.
921 butler.ingest(*datasets, transfer="copy", record_validation_info=False)
923 # Now try to ingest again in "execution butler" mode where
924 # the registry entries exist but the datastore does not have
925 # the files. We also need to strip the dimension records to ensure
926 # that they will be re-added by the ingest.
927 ref = datasets[0].refs[0]
928 datasets[0].refs = [
929 cast(
930 DatasetRef,
931 butler.registry.findDataset(ref.datasetType, dataId=ref.dataId, collections=ref.run),
932 )
933 for ref in datasets[0].refs
934 ]
935 all_refs = []
936 for dataset in datasets:
937 refs = []
938 for ref in dataset.refs:
939 # Create a dict from the dataId to drop the records.
940 new_data_id = {str(k): v for k, v in ref.dataId.items()}
941 new_ref = butler.registry.findDataset(ref.datasetType, new_data_id, collections=ref.run)
942 assert new_ref is not None
943 self.assertFalse(new_ref.dataId.hasRecords())
944 refs.append(new_ref)
945 dataset.refs = refs
946 all_refs.extend(dataset.refs)
947 butler.pruneDatasets(all_refs, disassociate=False, unstore=True, purge=False)
949 # Use move mode to test that the file is deleted. Also
950 # disable recording of file size.
951 butler.ingest(*datasets, transfer="move", record_validation_info=False)
953 # Check that every ref now has records.
954 for dataset in datasets:
955 for ref in dataset.refs:
956 self.assertTrue(ref.dataId.hasRecords())
958 # Ensure that the file has disappeared.
959 self.assertFalse(tempFile.exists())
961 # Check that the datastore recorded no file size.
962 # Not all datastores can support this.
963 try:
964 infos = butler._datastore.getStoredItemsInfo(datasets[0].refs[0]) # type: ignore[attr-defined]
965 self.assertEqual(infos[0].file_size, -1)
966 except AttributeError:
967 pass
969 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
970 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
972 multi1 = butler.get(datasetTypeName, dataId1)
973 multi2 = butler.get(datasetTypeName, dataId2)
975 self.assertEqual(multi1, metrics1)
976 self.assertEqual(multi2, metrics2)
978 # Compare URIs
979 uri1 = butler.getURI(datasetTypeName, dataId1)
980 uri2 = butler.getURI(datasetTypeName, dataId2)
981 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
983 # Test that removing one does not break the second
984 # This line will issue a warning log message for a ChainedDatastore
985 # that uses an InMemoryDatastore since in-memory can not ingest
986 # files.
987 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
988 self.assertFalse(butler.exists(datasetTypeName, dataId1))
989 self.assertTrue(butler.exists(datasetTypeName, dataId2))
990 multi2b = butler.get(datasetTypeName, dataId2)
991 self.assertEqual(multi2, multi2b)
993 # Ensure we can ingest 0 datasets
994 datasets = []
995 butler.ingest(*datasets)
997 def testPickle(self) -> None:
998 """Test pickle support."""
999 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1000 assert isinstance(butler, DirectButler), "Expect DirectButler in configuration"
1001 butlerOut = pickle.loads(pickle.dumps(butler))
1002 self.assertIsInstance(butlerOut, Butler)
1003 self.assertEqual(butlerOut._config, butler._config)
1004 self.assertEqual(butlerOut.collections, butler.collections)
1005 self.assertEqual(butlerOut.run, butler.run)
1007 def testGetDatasetTypes(self) -> None:
1008 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1009 dimensions = butler.dimensions.extract(["instrument", "visit", "physical_filter"])
1010 dimensionEntries: list[tuple[str, list[Mapping[str, Any]]]] = [
1011 (
1012 "instrument",
1013 [
1014 {"instrument": "DummyCam"},
1015 {"instrument": "DummyHSC"},
1016 {"instrument": "DummyCamComp"},
1017 ],
1018 ),
1019 ("physical_filter", [{"instrument": "DummyCam", "name": "d-r", "band": "R"}]),
1020 ("visit", [{"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}]),
1021 ]
1022 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1023 # Add needed Dimensions
1024 for element, data in dimensionEntries:
1025 butler.registry.insertDimensionData(element, *data)
1027 # When a DatasetType is added to the registry entries are not created
1028 # for components but querying them can return the components.
1029 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
1030 components = set()
1031 for datasetTypeName in datasetTypeNames:
1032 # Create and register a DatasetType
1033 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1035 for componentName in storageClass.components:
1036 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
1038 fromRegistry: set[DatasetType] = set()
1039 for parent_dataset_type in butler.registry.queryDatasetTypes():
1040 fromRegistry.add(parent_dataset_type)
1041 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
1042 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
1044 # Now that we have some dataset types registered, validate them
1045 butler.validateConfiguration(
1046 ignore=[
1047 "test_metric_comp",
1048 "metric3",
1049 "metric5",
1050 "calexp",
1051 "DummySC",
1052 "datasetType.component",
1053 "random_data",
1054 "random_data_2",
1055 ]
1056 )
1058 # Add a new datasetType that will fail template validation
1059 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
1060 if self.validationCanFail:
1061 with self.assertRaises(ValidationError):
1062 butler.validateConfiguration()
1064 # Rerun validation but with a subset of dataset type names
1065 butler.validateConfiguration(datasetTypeNames=["metric4"])
1067 # Rerun validation but ignore the bad datasetType
1068 butler.validateConfiguration(
1069 ignore=[
1070 "test_metric_comp",
1071 "metric3",
1072 "metric5",
1073 "calexp",
1074 "DummySC",
1075 "datasetType.component",
1076 "random_data",
1077 "random_data_2",
1078 ]
1079 )
1081 def testTransaction(self) -> None:
1082 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1083 datasetTypeName = "test_metric"
1084 dimensions = butler.dimensions.extract(["instrument", "visit"])
1085 dimensionEntries: tuple[tuple[str, Mapping[str, Any]], ...] = (
1086 ("instrument", {"instrument": "DummyCam"}),
1087 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1088 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1089 )
1090 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1091 metric = makeExampleMetrics()
1092 dataId = {"instrument": "DummyCam", "visit": 42}
1093 # Create and register a DatasetType
1094 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1095 with self.assertRaises(TransactionTestError):
1096 with butler.transaction():
1097 # Add needed Dimensions
1098 for args in dimensionEntries:
1099 butler.registry.insertDimensionData(*args)
1100 # Store a dataset
1101 ref = butler.put(metric, datasetTypeName, dataId)
1102 self.assertIsInstance(ref, DatasetRef)
1103 # Test getDirect
1104 metricOut = butler.get(ref)
1105 self.assertEqual(metric, metricOut)
1106 # Test get
1107 metricOut = butler.get(datasetTypeName, dataId)
1108 self.assertEqual(metric, metricOut)
1109 # Check we can get components
1110 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1111 raise TransactionTestError("This should roll back the entire transaction")
1112 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1113 butler.registry.expandDataId(dataId)
1114 # Should raise LookupError for missing data ID value
1115 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1116 butler.get(datasetTypeName, dataId)
1117 # Also check explicitly if Dataset entry is missing
1118 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections))
1119 # Direct retrieval should not find the file in the Datastore
1120 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1121 butler.get(ref)
1123 def testMakeRepo(self) -> None:
1124 """Test that we can write butler configuration to a new repository via
1125 the Butler.makeRepo interface and then instantiate a butler from the
1126 repo root.
1127 """
1128 # Do not run the test if we know this datastore configuration does
1129 # not support a file system root
1130 if self.fullConfigKey is None:
1131 return
1133 # create two separate directories
1134 root1 = tempfile.mkdtemp(dir=self.root)
1135 root2 = tempfile.mkdtemp(dir=self.root)
1137 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1138 limited = Config(self.configFile)
1139 butler1 = Butler.from_config(butlerConfig)
1140 assert isinstance(butler1, DirectButler), "Expect DirectButler in configuration"
1141 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1142 full = Config(self.tmpConfigFile)
1143 butler2 = Butler.from_config(butlerConfig)
1144 assert isinstance(butler2, DirectButler), "Expect DirectButler in configuration"
1145 # Butlers should have the same configuration regardless of whether
1146 # defaults were expanded.
1147 self.assertEqual(butler1._config, butler2._config)
1148 # Config files loaded directly should not be the same.
1149 self.assertNotEqual(limited, full)
1150 # Make sure "limited" doesn't have a few keys we know it should be
1151 # inheriting from defaults.
1152 self.assertIn(self.fullConfigKey, full)
1153 self.assertNotIn(self.fullConfigKey, limited)
1155 # Collections don't appear until something is put in them
1156 collections1 = set(butler1.registry.queryCollections())
1157 self.assertEqual(collections1, set())
1158 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1160 # Check that a config with no associated file name will not
1161 # work properly with relocatable Butler repo
1162 butlerConfig.configFile = None
1163 with self.assertRaises(ValueError):
1164 Butler.from_config(butlerConfig)
1166 with self.assertRaises(FileExistsError):
1167 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1169 def testStringification(self) -> None:
1170 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1171 butlerStr = str(butler)
1173 if self.datastoreStr is not None:
1174 for testStr in self.datastoreStr:
1175 self.assertIn(testStr, butlerStr)
1176 if self.registryStr is not None:
1177 self.assertIn(self.registryStr, butlerStr)
1179 datastoreName = butler._datastore.name
1180 if self.datastoreName is not None:
1181 for testStr in self.datastoreName:
1182 self.assertIn(testStr, datastoreName)
1184 def testButlerRewriteDataId(self) -> None:
1185 """Test that dataIds can be rewritten based on dimension records."""
1186 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1188 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1189 datasetTypeName = "random_data"
1191 # Create dimension records.
1192 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1193 butler.registry.insertDimensionData(
1194 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1195 )
1196 butler.registry.insertDimensionData(
1197 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1198 )
1200 dimensions = butler.dimensions.extract(["instrument", "exposure"])
1201 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1202 butler.registry.registerDatasetType(datasetType)
1204 n_exposures = 5
1205 dayobs = 20210530
1207 for i in range(n_exposures):
1208 butler.registry.insertDimensionData(
1209 "exposure",
1210 {
1211 "instrument": "DummyCamComp",
1212 "id": i,
1213 "obs_id": f"exp{i}",
1214 "seq_num": i,
1215 "day_obs": dayobs,
1216 "physical_filter": "d-r",
1217 },
1218 )
1220 # Write some data.
1221 for i in range(n_exposures):
1222 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1224 # Use the seq_num for the put to test rewriting.
1225 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1226 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1228 # Check that the exposure is correct in the dataId
1229 self.assertEqual(ref.dataId["exposure"], i)
1231 # and check that we can get the dataset back with the same dataId
1232 new_metric = butler.get(datasetTypeName, dataId=dataId)
1233 self.assertEqual(new_metric, metric)
1236class FileDatastoreButlerTests(ButlerTests):
1237 """Common tests and specialization of ButlerTests for butlers backed
1238 by datastores that inherit from FileDatastore.
1239 """
1241 def checkFileExists(self, root: str | ResourcePath, relpath: str | ResourcePath) -> bool:
1242 """Check if file exists at a given path (relative to root).
1244 Test testPutTemplates verifies actual physical existance of the files
1245 in the requested location.
1246 """
1247 uri = ResourcePath(root, forceDirectory=True)
1248 return uri.join(relpath).exists()
1250 def testPutTemplates(self) -> None:
1251 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1252 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1254 # Add needed Dimensions
1255 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1256 butler.registry.insertDimensionData(
1257 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1258 )
1259 butler.registry.insertDimensionData(
1260 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1261 )
1262 butler.registry.insertDimensionData(
1263 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1264 )
1266 # Create and store a dataset
1267 metric = makeExampleMetrics()
1269 # Create two almost-identical DatasetTypes (both will use default
1270 # template)
1271 dimensions = butler.dimensions.extract(["instrument", "visit"])
1272 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1273 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1274 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1276 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1277 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1279 # Put with exactly the data ID keys needed
1280 ref = butler.put(metric, "metric1", dataId1)
1281 uri = butler.getURI(ref)
1282 self.assertTrue(uri.exists())
1283 self.assertTrue(
1284 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle")
1285 )
1287 # Check the template based on dimensions
1288 if hasattr(butler._datastore, "templates"):
1289 butler._datastore.templates.validateTemplates([ref])
1291 # Put with extra data ID keys (physical_filter is an optional
1292 # dependency); should not change template (at least the way we're
1293 # defining them to behave now; the important thing is that they
1294 # must be consistent).
1295 ref = butler.put(metric, "metric2", dataId2)
1296 uri = butler.getURI(ref)
1297 self.assertTrue(uri.exists())
1298 self.assertTrue(
1299 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle")
1300 )
1302 # Check the template based on dimensions
1303 if hasattr(butler._datastore, "templates"):
1304 butler._datastore.templates.validateTemplates([ref])
1306 # Use a template that has a typo in dimension record metadata.
1307 # Easier to test with a butler that has a ref with records attached.
1308 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1309 with self.assertLogs("lsst.daf.butler.datastore.file_templates", "INFO"):
1310 path = template.format(ref)
1311 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1313 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1314 with self.assertRaises(KeyError):
1315 with self.assertLogs("lsst.daf.butler.datastore.file_templates", "INFO"):
1316 template.format(ref)
1318 # Now use a file template that will not result in unique filenames
1319 with self.assertRaises(FileTemplateValidationError):
1320 butler.put(metric, "metric3", dataId1)
1322 def testImportExport(self) -> None:
1323 # Run put/get tests just to create and populate a repo.
1324 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1325 self.runImportExportTest(storageClass)
1327 @unittest.expectedFailure
1328 def testImportExportVirtualComposite(self) -> None:
1329 # Run put/get tests just to create and populate a repo.
1330 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1331 self.runImportExportTest(storageClass)
1333 def runImportExportTest(self, storageClass: StorageClass) -> None:
1334 """Test exporting and importing.
1336 This test does an export to a temp directory and an import back
1337 into a new temp directory repo. It does not assume a posix datastore.
1338 """
1339 exportButler = self.runPutGetTest(storageClass, "test_metric")
1341 # Test that we must have a file extension.
1342 with self.assertRaises(ValueError):
1343 with exportButler.export(filename="dump", directory=".") as export:
1344 pass
1346 # Test that unknown format is not allowed.
1347 with self.assertRaises(ValueError):
1348 with exportButler.export(filename="dump.fits", directory=".") as export:
1349 pass
1351 # Test that the repo actually has at least one dataset.
1352 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1353 self.assertGreater(len(datasets), 0)
1354 # Add a DimensionRecord that's unused by those datasets.
1355 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1356 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1357 # Export and then import datasets.
1358 with safeTestTempDir(TESTDIR) as exportDir:
1359 exportFile = os.path.join(exportDir, "exports.yaml")
1360 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1361 export.saveDatasets(datasets)
1362 # Export the same datasets again. This should quietly do
1363 # nothing because of internal deduplication, and it shouldn't
1364 # complain about being asked to export the "htm7" elements even
1365 # though there aren't any in these datasets or in the database.
1366 export.saveDatasets(datasets, elements=["htm7"])
1367 # Save one of the data IDs again; this should be harmless
1368 # because of internal deduplication.
1369 export.saveDataIds([datasets[0].dataId])
1370 # Save some dimension records directly.
1371 export.saveDimensionData("skymap", [skymapRecord])
1372 self.assertTrue(os.path.exists(exportFile))
1373 with safeTestTempDir(TESTDIR) as importDir:
1374 # We always want this to be a local posix butler
1375 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1376 # Calling script.butlerImport tests the implementation of the
1377 # butler command line interface "import" subcommand. Functions
1378 # in the script folder are generally considered protected and
1379 # should not be used as public api.
1380 with open(exportFile) as f:
1381 script.butlerImport(
1382 importDir,
1383 export_file=f,
1384 directory=exportDir,
1385 transfer="auto",
1386 skip_dimensions=None,
1387 )
1388 importButler = Butler.from_config(importDir, run=self.default_run)
1389 for ref in datasets:
1390 with self.subTest(ref=ref):
1391 # Test for existence by passing in the DatasetType and
1392 # data ID separately, to avoid lookup by dataset_id.
1393 self.assertTrue(importButler.exists(ref.datasetType, ref.dataId))
1394 self.assertEqual(
1395 list(importButler.registry.queryDimensionRecords("skymap")),
1396 [importButler.dimensions["skymap"].RecordClass(**skymapRecord)],
1397 )
1399 def testRemoveRuns(self) -> None:
1400 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1401 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
1402 # Load registry data with dimensions to hang datasets off of.
1403 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1404 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1405 # Add some RUN-type collection.
1406 run1 = "run1"
1407 butler.registry.registerRun(run1)
1408 run2 = "run2"
1409 butler.registry.registerRun(run2)
1410 # put a dataset in each
1411 metric = makeExampleMetrics()
1412 dimensions = butler.dimensions.extract(["instrument", "physical_filter"])
1413 datasetType = self.addDatasetType(
1414 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1415 )
1416 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1417 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1418 uri1 = butler.getURI(ref1)
1419 uri2 = butler.getURI(ref2)
1421 with self.assertRaises(OrphanedRecordError):
1422 butler.registry.removeDatasetType(datasetType.name)
1424 # Remove from both runs with different values for unstore.
1425 butler.removeRuns([run1], unstore=True)
1426 butler.removeRuns([run2], unstore=False)
1427 # Should be nothing in registry for either one, and datastore should
1428 # not think either exists.
1429 with self.assertRaises(MissingCollectionError):
1430 butler.registry.getCollectionType(run1)
1431 with self.assertRaises(MissingCollectionError):
1432 butler.registry.getCollectionType(run2)
1433 self.assertFalse(butler.stored(ref1))
1434 self.assertFalse(butler.stored(ref2))
1435 # The ref we unstored should be gone according to the URI, but the
1436 # one we forgot should still be around.
1437 self.assertFalse(uri1.exists())
1438 self.assertTrue(uri2.exists())
1440 # Now that the collections have been pruned we can remove the
1441 # dataset type
1442 butler.registry.removeDatasetType(datasetType.name)
1444 with self.assertLogs("lsst.daf.butler.registry", "INFO") as cm:
1445 butler.registry.removeDatasetType(("test*", "test*"))
1446 self.assertIn("not defined", "\n".join(cm.output))
1449class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1450 """PosixDatastore specialization of a butler"""
1452 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1453 fullConfigKey: str | None = ".datastore.formatters"
1454 validationCanFail = True
1455 datastoreStr = ["/tmp"]
1456 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1457 registryStr = "/gen3.sqlite3"
1459 def testPathConstructor(self) -> None:
1460 """Independent test of constructor using PathLike."""
1461 butler = Butler.from_config(self.tmpConfigFile, run=self.default_run)
1462 self.assertIsInstance(butler, Butler)
1464 # And again with a Path object with the butler yaml
1465 path = pathlib.Path(self.tmpConfigFile)
1466 butler = Butler.from_config(path, writeable=False)
1467 self.assertIsInstance(butler, Butler)
1469 # And again with a Path object without the butler yaml
1470 # (making sure we skip it if the tmp config doesn't end
1471 # in butler.yaml -- which is the case for a subclass)
1472 if self.tmpConfigFile.endswith("butler.yaml"):
1473 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1474 butler = Butler.from_config(path, writeable=False)
1475 self.assertIsInstance(butler, Butler)
1477 def testExportTransferCopy(self) -> None:
1478 """Test local export using all transfer modes"""
1479 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1480 exportButler = self.runPutGetTest(storageClass, "test_metric")
1481 # Test that the repo actually has at least one dataset.
1482 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1483 self.assertGreater(len(datasets), 0)
1484 uris = [exportButler.getURI(d) for d in datasets]
1485 assert isinstance(exportButler._datastore, FileDatastore)
1486 datastoreRoot = exportButler.get_datastore_roots()[exportButler.get_datastore_names()[0]]
1488 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1490 for path in pathsInStore:
1491 # Assume local file system
1492 assert path is not None
1493 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1495 for transfer in ("copy", "link", "symlink", "relsymlink"):
1496 with safeTestTempDir(TESTDIR) as exportDir:
1497 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1498 export.saveDatasets(datasets)
1499 for path in pathsInStore:
1500 assert path is not None
1501 self.assertTrue(
1502 self.checkFileExists(exportDir, path),
1503 f"Check that mode {transfer} exported files",
1504 )
1506 def testPruneDatasets(self) -> None:
1507 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1508 butler = Butler.from_config(self.tmpConfigFile, writeable=True)
1509 assert isinstance(butler._datastore, FileDatastore)
1510 # Load registry data with dimensions to hang datasets off of.
1511 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1512 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1513 # Add some RUN-type collections.
1514 run1 = "run1"
1515 butler.registry.registerRun(run1)
1516 run2 = "run2"
1517 butler.registry.registerRun(run2)
1518 # put some datasets. ref1 and ref2 have the same data ID, and are in
1519 # different runs. ref3 has a different data ID.
1520 metric = makeExampleMetrics()
1521 dimensions = butler.dimensions.extract(["instrument", "physical_filter"])
1522 datasetType = self.addDatasetType(
1523 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1524 )
1525 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1526 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1527 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1529 many_stored = butler.stored_many([ref1, ref2, ref3])
1530 for ref, stored in many_stored.items():
1531 self.assertTrue(stored, f"Ref {ref} should be stored")
1533 many_exists = butler._exists_many([ref1, ref2, ref3])
1534 for ref, exists in many_exists.items():
1535 self.assertTrue(exists, f"Checking ref {ref} exists.")
1536 self.assertEqual(exists, DatasetExistence.VERIFIED, f"Ref {ref} should be stored")
1538 # Simple prune.
1539 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1540 self.assertFalse(butler.exists(ref1.datasetType, ref1.dataId, collections=run1))
1542 many_stored = butler.stored_many([ref1, ref2, ref3])
1543 for ref, stored in many_stored.items():
1544 self.assertFalse(stored, f"Ref {ref} should not be stored")
1546 many_exists = butler._exists_many([ref1, ref2, ref3])
1547 for ref, exists in many_exists.items():
1548 self.assertEqual(exists, DatasetExistence.UNRECOGNIZED, f"Ref {ref} should not be stored")
1550 # Put data back.
1551 ref1_new = butler.put(metric, ref1)
1552 self.assertEqual(ref1_new, ref1) # Reuses original ID.
1553 ref2 = butler.put(metric, ref2)
1555 many_stored = butler.stored_many([ref1, ref2, ref3])
1556 self.assertTrue(many_stored[ref1])
1557 self.assertTrue(many_stored[ref2])
1558 self.assertFalse(many_stored[ref3])
1560 ref3 = butler.put(metric, ref3)
1562 many_exists = butler._exists_many([ref1, ref2, ref3])
1563 for ref, exists in many_exists.items():
1564 self.assertTrue(exists, f"Ref {ref} should not be stored")
1566 # Clear out the datasets from registry and start again.
1567 refs = [ref1, ref2, ref3]
1568 butler.pruneDatasets(refs, purge=True, unstore=True)
1569 for ref in refs:
1570 butler.put(metric, ref)
1572 # Confirm we can retrieve deferred.
1573 dref1 = butler.getDeferred(ref1) # known and exists
1574 metric1 = dref1.get()
1575 self.assertEqual(metric1, metric)
1577 # Test different forms of file availability.
1578 # Need to be in a state where:
1579 # - one ref just has registry record.
1580 # - one ref has a missing file but a datastore record.
1581 # - one ref has a missing datastore record but file is there.
1582 # - one ref does not exist anywhere.
1583 # Do not need to test a ref that has everything since that is tested
1584 # above.
1585 ref0 = DatasetRef(
1586 datasetType,
1587 DataCoordinate.standardize(
1588 {"instrument": "Cam1", "physical_filter": "Cam1-G"}, universe=butler.dimensions
1589 ),
1590 run=run1,
1591 )
1593 # Delete from datastore and retain in Registry.
1594 butler.pruneDatasets([ref1], purge=False, unstore=True, disassociate=False)
1596 # File has been removed.
1597 uri2 = butler.getURI(ref2)
1598 uri2.remove()
1600 # Datastore has lost track.
1601 butler._datastore.forget([ref3])
1603 # First test with a standard butler.
1604 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1605 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1606 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1607 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1608 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED)
1610 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=False)
1611 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1612 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1613 self.assertEqual(exists_many[ref2], DatasetExistence.KNOWN)
1614 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1615 self.assertTrue(exists_many[ref2])
1617 # Check that per-ref query gives the same answer as many query.
1618 for ref, exists in exists_many.items():
1619 self.assertEqual(butler.exists(ref, full_check=False), exists)
1621 # Get deferred checks for existence before it allows it to be
1622 # retrieved.
1623 with self.assertRaises(LookupError):
1624 butler.getDeferred(ref3) # not known, file exists
1625 dref2 = butler.getDeferred(ref2) # known but file missing
1626 with self.assertRaises(FileNotFoundError):
1627 dref2.get()
1629 # Test again with a trusting butler.
1630 butler._datastore.trustGetRequest = True
1631 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1632 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1633 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1634 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1635 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ARTIFACT)
1637 # When trusting we can get a deferred dataset handle that is not
1638 # known but does exist.
1639 dref3 = butler.getDeferred(ref3)
1640 metric3 = dref3.get()
1641 self.assertEqual(metric3, metric)
1643 # Check that per-ref query gives the same answer as many query.
1644 for ref, exists in exists_many.items():
1645 self.assertEqual(butler.exists(ref, full_check=True), exists)
1647 # Create a ref that surprisingly has the UUID of an existing ref
1648 # but is not the same.
1649 ref_bad = DatasetRef(datasetType, dataId=ref3.dataId, run=ref3.run, id=ref2.id)
1650 with self.assertRaises(ValueError):
1651 butler.exists(ref_bad)
1653 # Create a ref that has a compatible storage class.
1654 ref_compat = ref2.overrideStorageClass("StructuredDataDict")
1655 exists = butler.exists(ref_compat)
1656 self.assertEqual(exists, exists_many[ref2])
1658 # Remove everything and start from scratch.
1659 butler._datastore.trustGetRequest = False
1660 butler.pruneDatasets(refs, purge=True, unstore=True)
1661 for ref in refs:
1662 butler.put(metric, ref)
1664 # These tests mess directly with the trash table and can leave the
1665 # datastore in an odd state. Do them at the end.
1666 # Check that in normal mode, deleting the record will lead to
1667 # trash not touching the file.
1668 uri1 = butler.getURI(ref1)
1669 butler._datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1670 butler._datastore.forget([ref1])
1671 butler._datastore.trash(ref1)
1672 butler._datastore.emptyTrash()
1673 self.assertTrue(uri1.exists())
1674 uri1.remove() # Clean it up.
1676 # Simulate execution butler setup by deleting the datastore
1677 # record but keeping the file around and trusting.
1678 butler._datastore.trustGetRequest = True
1679 uris = butler.get_many_uris([ref2, ref3])
1680 uri2 = uris[ref2].primaryURI
1681 uri3 = uris[ref3].primaryURI
1682 self.assertTrue(uri2.exists())
1683 self.assertTrue(uri3.exists())
1685 # Remove the datastore record.
1686 butler._datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1687 butler._datastore.forget([ref2])
1688 self.assertTrue(uri2.exists())
1689 butler._datastore.trash([ref2, ref3])
1690 # Immediate removal for ref2 file
1691 self.assertFalse(uri2.exists())
1692 # But ref3 has to wait for the empty.
1693 self.assertTrue(uri3.exists())
1694 butler._datastore.emptyTrash()
1695 self.assertFalse(uri3.exists())
1697 # Clear out the datasets from registry.
1698 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1700 def testPytypeCoercion(self) -> None:
1701 """Test python type coercion on Butler.get and put."""
1702 # Store some data with the normal example storage class.
1703 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1704 datasetTypeName = "test_metric"
1705 butler = self.runPutGetTest(storageClass, datasetTypeName)
1707 dataId = {"instrument": "DummyCamComp", "visit": 423}
1708 metric = butler.get(datasetTypeName, dataId=dataId)
1709 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1711 datasetType_ori = butler.registry.getDatasetType(datasetTypeName)
1712 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1714 # Now need to hack the registry dataset type definition.
1715 # There is no API for this.
1716 assert isinstance(butler._registry, SqlRegistry)
1717 manager = butler._registry._managers.datasets
1718 assert hasattr(manager, "_db") and hasattr(manager, "_static")
1719 manager._db.update(
1720 manager._static.dataset_type,
1721 {"name": datasetTypeName},
1722 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1723 )
1725 # Force reset of dataset type cache
1726 butler.registry.refresh()
1728 datasetType_new = butler.registry.getDatasetType(datasetTypeName)
1729 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1730 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1732 metric_model = butler.get(datasetTypeName, dataId=dataId)
1733 self.assertNotEqual(type(metric_model), type(metric))
1734 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1736 # Put the model and read it back to show that everything now
1737 # works as normal.
1738 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1739 metric_model_new = butler.get(metric_ref)
1740 self.assertEqual(metric_model_new, metric_model)
1742 # Hack the storage class again to something that will fail on the
1743 # get with no conversion class.
1744 manager._db.update(
1745 manager._static.dataset_type,
1746 {"name": datasetTypeName},
1747 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1748 )
1749 butler.registry.refresh()
1751 with self.assertRaises(ValueError):
1752 butler.get(datasetTypeName, dataId=dataId)
1755@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1756class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1757 """PosixDatastore specialization of a butler using Postgres"""
1759 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1760 fullConfigKey = ".datastore.formatters"
1761 validationCanFail = True
1762 datastoreStr = ["/tmp"]
1763 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1764 registryStr = "PostgreSQL@test"
1765 postgresql: Any
1767 @staticmethod
1768 def _handler(postgresql: Any) -> None:
1769 engine = sqlalchemy.engine.create_engine(postgresql.url())
1770 with engine.begin() as connection:
1771 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1773 @classmethod
1774 def setUpClass(cls) -> None:
1775 # Create the postgres test server.
1776 cls.postgresql = testing.postgresql.PostgresqlFactory(
1777 cache_initialized_db=True, on_initialized=cls._handler
1778 )
1779 super().setUpClass()
1781 @classmethod
1782 def tearDownClass(cls) -> None:
1783 # Clean up any lingering SQLAlchemy engines/connections
1784 # so they're closed before we shut down the server.
1785 gc.collect()
1786 cls.postgresql.clear_cache()
1787 super().tearDownClass()
1789 def setUp(self) -> None:
1790 self.server = self.postgresql()
1792 # Need to add a registry section to the config.
1793 self._temp_config = False
1794 config = Config(self.configFile)
1795 config["registry", "db"] = self.server.url()
1796 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1797 config.dump(fh)
1798 self.configFile = fh.name
1799 self._temp_config = True
1800 super().setUp()
1802 def tearDown(self) -> None:
1803 self.server.stop()
1804 if self._temp_config and os.path.exists(self.configFile):
1805 os.remove(self.configFile)
1806 super().tearDown()
1808 def testMakeRepo(self) -> None:
1809 # The base class test assumes that it's using sqlite and assumes
1810 # the config file is acceptable to sqlite.
1811 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1814class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1815 """InMemoryDatastore specialization of a butler"""
1817 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1818 fullConfigKey = None
1819 useTempRoot = False
1820 validationCanFail = False
1821 datastoreStr = ["datastore='InMemory"]
1822 datastoreName = ["InMemoryDatastore@"]
1823 registryStr = "/gen3.sqlite3"
1825 def testIngest(self) -> None:
1826 pass
1829class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1830 """PosixDatastore specialization"""
1832 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1833 fullConfigKey = ".datastore.datastores.1.formatters"
1834 validationCanFail = True
1835 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1836 datastoreName = [
1837 "InMemoryDatastore@",
1838 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1839 "SecondDatastore",
1840 ]
1841 registryStr = "/gen3.sqlite3"
1844class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1845 """Test that a yaml file in one location can refer to a root in another."""
1847 datastoreStr = ["dir1"]
1848 # Disable the makeRepo test since we are deliberately not using
1849 # butler.yaml as the config name.
1850 fullConfigKey = None
1852 def setUp(self) -> None:
1853 self.root = makeTestTempDir(TESTDIR)
1855 # Make a new repository in one place
1856 self.dir1 = os.path.join(self.root, "dir1")
1857 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1859 # Move the yaml file to a different place and add a "root"
1860 self.dir2 = os.path.join(self.root, "dir2")
1861 os.makedirs(self.dir2, exist_ok=True)
1862 configFile1 = os.path.join(self.dir1, "butler.yaml")
1863 config = Config(configFile1)
1864 config["root"] = self.dir1
1865 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1866 config.dumpToUri(configFile2)
1867 os.remove(configFile1)
1868 self.tmpConfigFile = configFile2
1870 def testFileLocations(self) -> None:
1871 self.assertNotEqual(self.dir1, self.dir2)
1872 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1873 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1874 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1877class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1878 """Test that a config file created by makeRepo outside of repo works."""
1880 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1882 def setUp(self) -> None:
1883 self.root = makeTestTempDir(TESTDIR)
1884 self.root2 = makeTestTempDir(TESTDIR)
1886 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1887 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1889 def tearDown(self) -> None:
1890 if os.path.exists(self.root2):
1891 shutil.rmtree(self.root2, ignore_errors=True)
1892 super().tearDown()
1894 def testConfigExistence(self) -> None:
1895 c = Config(self.tmpConfigFile)
1896 uri_config = ResourcePath(c["root"])
1897 uri_expected = ResourcePath(self.root, forceDirectory=True)
1898 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1899 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1901 def testPutGet(self) -> None:
1902 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1903 self.runPutGetTest(storageClass, "test_metric")
1906class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase):
1907 """Test that a config file created by makeRepo outside of repo works."""
1909 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1911 def setUp(self) -> None:
1912 self.root = makeTestTempDir(TESTDIR)
1913 self.root2 = makeTestTempDir(TESTDIR)
1915 self.tmpConfigFile = self.root2
1916 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1918 def testConfigExistence(self) -> None:
1919 # Append the yaml file else Config constructor does not know the file
1920 # type.
1921 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1922 super().testConfigExistence()
1925class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1926 """Test that a config file created by makeRepo outside of repo works."""
1928 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1930 def setUp(self) -> None:
1931 self.root = makeTestTempDir(TESTDIR)
1932 self.root2 = makeTestTempDir(TESTDIR)
1934 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1935 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1938@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1939class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1940 """S3Datastore specialization of a butler; an S3 storage Datastore +
1941 a local in-memory SqlRegistry.
1942 """
1944 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1945 fullConfigKey = None
1946 validationCanFail = True
1948 bucketName = "anybucketname"
1949 """Name of the Bucket that will be used in the tests. The name is read from
1950 the config file used with the tests during set-up.
1951 """
1953 root = "butlerRoot/"
1954 """Root repository directory expected to be used in case useTempRoot=False.
1955 Otherwise the root is set to a 20 characters long randomly generated string
1956 during set-up.
1957 """
1959 datastoreStr = [f"datastore={root}"]
1960 """Contains all expected root locations in a format expected to be
1961 returned by Butler stringification.
1962 """
1964 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1965 """The expected format of the S3 Datastore string."""
1967 registryStr = "/gen3.sqlite3"
1968 """Expected format of the Registry string."""
1970 mock_s3 = mock_s3()
1971 """The mocked s3 interface from moto."""
1973 def genRoot(self) -> str:
1974 """Return a random string of len 20 to serve as a root
1975 name for the temporary bucket repo.
1977 This is equivalent to tempfile.mkdtemp as this is what self.root
1978 becomes when useTempRoot is True.
1979 """
1980 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1981 return rndstr + "/"
1983 def setUp(self) -> None:
1984 config = Config(self.configFile)
1985 uri = ResourcePath(config[".datastore.datastore.root"])
1986 self.bucketName = uri.netloc
1988 # Enable S3 mocking of tests.
1989 self.mock_s3.start()
1991 # set up some fake credentials if they do not exist
1992 self.usingDummyCredentials = setAwsEnvCredentials()
1994 if self.useTempRoot:
1995 self.root = self.genRoot()
1996 rooturi = f"s3://{self.bucketName}/{self.root}"
1997 config.update({"datastore": {"datastore": {"root": rooturi}}})
1999 # need local folder to store registry database
2000 self.reg_dir = makeTestTempDir(TESTDIR)
2001 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
2003 # MOTO needs to know that we expect Bucket bucketname to exist
2004 # (this used to be the class attribute bucketName)
2005 s3 = boto3.resource("s3")
2006 s3.create_bucket(Bucket=self.bucketName)
2008 self.datastoreStr = [f"datastore='{rooturi}'"]
2009 self.datastoreName = [f"FileDatastore@{rooturi}"]
2010 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
2011 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
2013 def tearDown(self) -> None:
2014 s3 = boto3.resource("s3")
2015 bucket = s3.Bucket(self.bucketName)
2016 try:
2017 bucket.objects.all().delete()
2018 except botocore.exceptions.ClientError as e:
2019 if e.response["Error"]["Code"] == "404":
2020 # the key was not reachable - pass
2021 pass
2022 else:
2023 raise
2025 bucket = s3.Bucket(self.bucketName)
2026 bucket.delete()
2028 # Stop the S3 mock.
2029 self.mock_s3.stop()
2031 # unset any potentially set dummy credentials
2032 if self.usingDummyCredentials:
2033 unsetAwsEnvCredentials()
2035 if self.reg_dir is not None and os.path.exists(self.reg_dir):
2036 shutil.rmtree(self.reg_dir, ignore_errors=True)
2038 if self.useTempRoot and os.path.exists(self.root):
2039 shutil.rmtree(self.root, ignore_errors=True)
2041 super().tearDown()
2044class PosixDatastoreTransfers(unittest.TestCase):
2045 """Test data transfers between butlers.
2047 Test for different managers. UUID to UUID and integer to integer are
2048 tested. UUID to integer is not supported since we do not currently
2049 want to allow that. Integer to UUID is supported with the caveat
2050 that UUID4 will be generated and this will be incorrect for raw
2051 dataset types. The test ignores that.
2052 """
2054 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
2055 storageClassFactory: StorageClassFactory
2057 @classmethod
2058 def setUpClass(cls) -> None:
2059 cls.storageClassFactory = StorageClassFactory()
2060 cls.storageClassFactory.addFromConfig(cls.configFile)
2062 def setUp(self) -> None:
2063 self.root = makeTestTempDir(TESTDIR)
2064 self.config = Config(self.configFile)
2066 def tearDown(self) -> None:
2067 removeTestTempDir(self.root)
2069 def create_butler(self, manager: str, label: str) -> Butler:
2070 config = Config(self.configFile)
2071 config["registry", "managers", "datasets"] = manager
2072 return Butler.from_config(
2073 Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True
2074 )
2076 def create_butlers(self, manager1: str | None = None, manager2: str | None = None) -> None:
2077 default = "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID"
2078 if manager1 is None:
2079 manager1 = default
2080 if manager2 is None:
2081 manager2 = default
2082 self.source_butler = self.create_butler(manager1, "1")
2083 self.target_butler = self.create_butler(manager2, "2")
2085 def testTransferUuidToUuid(self) -> None:
2086 self.create_butlers()
2087 self.assertButlerTransfers()
2089 def _enable_trust(self, datastore: Datastore) -> None:
2090 datastores = getattr(datastore, "datastores", [datastore])
2091 for this_datastore in datastores:
2092 if hasattr(this_datastore, "trustGetRequest"):
2093 this_datastore.trustGetRequest = True
2095 def testTransferMissing(self) -> None:
2096 """Test transfers where datastore records are missing.
2098 This is how execution butler works.
2099 """
2100 self.create_butlers()
2102 # Configure the source butler to allow trust.
2103 self._enable_trust(self.source_butler._datastore)
2105 self.assertButlerTransfers(purge=True)
2107 def testTransferMissingDisassembly(self) -> None:
2108 """Test transfers where datastore records are missing.
2110 This is how execution butler works.
2111 """
2112 self.create_butlers()
2114 # Configure the source butler to allow trust.
2115 self._enable_trust(self.source_butler._datastore)
2117 # Test disassembly.
2118 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
2120 def testAbsoluteURITransferDirect(self) -> None:
2121 """Test transfer using an absolute URI."""
2122 self._absolute_transfer("auto")
2124 def testAbsoluteURITransferCopy(self) -> None:
2125 """Test transfer using an absolute URI."""
2126 self._absolute_transfer("copy")
2128 def _absolute_transfer(self, transfer: str) -> None:
2129 self.create_butlers()
2131 storageClassName = "StructuredData"
2132 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2133 datasetTypeName = "random_data"
2134 run = "run1"
2135 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2137 dimensions = self.source_butler.dimensions.extract(())
2138 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2139 self.source_butler.registry.registerDatasetType(datasetType)
2141 metrics = makeExampleMetrics()
2142 with ResourcePath.temporary_uri(suffix=".json") as temp:
2143 dataId = DataCoordinate.makeEmpty(self.source_butler.dimensions)
2144 source_refs = [DatasetRef(datasetType, dataId, run=run)]
2145 temp.write(json.dumps(metrics.exportAsDict()).encode())
2146 dataset = FileDataset(path=temp, refs=source_refs)
2147 self.source_butler.ingest(dataset, transfer="direct")
2149 self.target_butler.transfer_from(
2150 self.source_butler, dataset.refs, register_dataset_types=True, transfer=transfer
2151 )
2153 uri = self.target_butler.getURI(dataset.refs[0])
2154 if transfer == "auto":
2155 self.assertEqual(uri, temp)
2156 else:
2157 self.assertNotEqual(uri, temp)
2159 def assertButlerTransfers(self, purge: bool = False, storageClassName: str = "StructuredData") -> None:
2160 """Test that a run can be transferred to another butler."""
2161 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2162 datasetTypeName = "random_data"
2164 # Test will create 3 collections and we will want to transfer
2165 # two of those three.
2166 runs = ["run1", "run2", "other"]
2168 # Also want to use two different dataset types to ensure that
2169 # grouping works.
2170 datasetTypeNames = ["random_data", "random_data_2"]
2172 # Create the run collections in the source butler.
2173 for run in runs:
2174 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2176 # Create dimensions in source butler.
2177 n_exposures = 30
2178 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
2179 self.source_butler.registry.insertDimensionData(
2180 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
2181 )
2182 self.source_butler.registry.insertDimensionData(
2183 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
2184 )
2186 for i in range(n_exposures):
2187 self.source_butler.registry.insertDimensionData(
2188 "exposure",
2189 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
2190 )
2192 # Create dataset types in the source butler.
2193 dimensions = self.source_butler.dimensions.extract(["instrument", "exposure"])
2194 for datasetTypeName in datasetTypeNames:
2195 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2196 self.source_butler.registry.registerDatasetType(datasetType)
2198 # Write a dataset to an unrelated run -- this will ensure that
2199 # we are rewriting integer dataset ids in the target if necessary.
2200 # Will not be relevant for UUID.
2201 run = "distraction"
2202 butler = Butler.from_config(butler=self.source_butler, run=run)
2203 butler.put(
2204 makeExampleMetrics(),
2205 datasetTypeName,
2206 exposure=1,
2207 instrument="DummyCamComp",
2208 physical_filter="d-r",
2209 )
2211 # Write some example metrics to the source
2212 butler = Butler.from_config(butler=self.source_butler)
2214 # Set of DatasetRefs that should be in the list of refs to transfer
2215 # but which will not be transferred.
2216 deleted: set[DatasetRef] = set()
2218 n_expected = 20 # Number of datasets expected to be transferred
2219 source_refs = []
2220 for i in range(n_exposures):
2221 # Put a third of datasets into each collection, only retain
2222 # two thirds.
2223 index = i % 3
2224 run = runs[index]
2225 datasetTypeName = datasetTypeNames[i % 2]
2227 metric = MetricsExample(
2228 summary={"counter": i}, output={"text": "metric"}, data=[2 * x for x in range(i)]
2229 )
2230 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2231 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2233 # Remove the datastore record using low-level API, but only
2234 # for a specific index.
2235 if purge and index == 1:
2236 # For one of these delete the file as well.
2237 # This allows the "missing" code to filter the
2238 # file out.
2239 # Access the individual datastores.
2240 datastores = []
2241 if hasattr(butler._datastore, "datastores"):
2242 datastores.extend(butler._datastore.datastores)
2243 else:
2244 datastores.append(butler._datastore)
2246 if not deleted:
2247 # For a chained datastore we need to remove
2248 # files in each chain.
2249 for datastore in datastores:
2250 # The file might not be known to the datastore
2251 # if constraints are used.
2252 try:
2253 primary, uris = datastore.getURIs(ref)
2254 except FileNotFoundError:
2255 continue
2256 if primary and primary.scheme != "mem":
2257 primary.remove()
2258 for uri in uris.values():
2259 if uri.scheme != "mem":
2260 uri.remove()
2261 n_expected -= 1
2262 deleted.add(ref)
2264 # Remove the datastore record.
2265 for datastore in datastores:
2266 if hasattr(datastore, "removeStoredItemInfo"):
2267 datastore.removeStoredItemInfo(ref)
2269 if index < 2:
2270 source_refs.append(ref)
2271 if ref not in deleted:
2272 new_metric = butler.get(ref)
2273 self.assertEqual(new_metric, metric)
2275 # Create some bad dataset types to ensure we check for inconsistent
2276 # definitions.
2277 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2278 for datasetTypeName in datasetTypeNames:
2279 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2280 self.target_butler.registry.registerDatasetType(datasetType)
2281 with self.assertRaises(ConflictingDefinitionError) as cm:
2282 self.target_butler.transfer_from(self.source_butler, source_refs)
2283 self.assertIn("dataset type differs", str(cm.exception))
2285 # And remove the bad definitions.
2286 for datasetTypeName in datasetTypeNames:
2287 self.target_butler.registry.removeDatasetType(datasetTypeName)
2289 # Transfer without creating dataset types should fail.
2290 with self.assertRaises(KeyError):
2291 self.target_butler.transfer_from(self.source_butler, source_refs)
2293 # Transfer without creating dimensions should fail.
2294 with self.assertRaises(ConflictingDefinitionError) as cm:
2295 self.target_butler.transfer_from(self.source_butler, source_refs, register_dataset_types=True)
2296 self.assertIn("dimension", str(cm.exception))
2298 # The failed transfer above leaves registry in an inconsistent
2299 # state because the run is created but then rolled back without
2300 # the collection cache being cleared. For now force a refresh.
2301 # Can remove with DM-35498.
2302 self.target_butler.registry.refresh()
2304 # Now transfer them to the second butler, including dimensions.
2305 with self.assertLogs(level=logging.DEBUG) as log_cm:
2306 transferred = self.target_butler.transfer_from(
2307 self.source_butler,
2308 source_refs,
2309 register_dataset_types=True,
2310 transfer_dimensions=True,
2311 )
2312 self.assertEqual(len(transferred), n_expected)
2313 log_output = ";".join(log_cm.output)
2315 # A ChainedDatastore will use the in-memory datastore for mexists
2316 # so we can not rely on the mexists log message.
2317 self.assertIn("Number of datastore records found in source", log_output)
2318 self.assertIn("Creating output run", log_output)
2320 # Do the transfer twice to ensure that it will do nothing extra.
2321 # Only do this if purge=True because it does not work for int
2322 # dataset_id.
2323 if purge:
2324 # This should not need to register dataset types.
2325 transferred = self.target_butler.transfer_from(self.source_butler, source_refs)
2326 self.assertEqual(len(transferred), n_expected)
2328 # Also do an explicit low-level transfer to trigger some
2329 # edge cases.
2330 with self.assertLogs(level=logging.DEBUG) as log_cm:
2331 self.target_butler._datastore.transfer_from(self.source_butler._datastore, source_refs)
2332 log_output = ";".join(log_cm.output)
2333 self.assertIn("no file artifacts exist", log_output)
2335 with self.assertRaises((TypeError, AttributeError)):
2336 self.target_butler._datastore.transfer_from(self.source_butler, source_refs) # type: ignore
2338 with self.assertRaises(ValueError):
2339 self.target_butler._datastore.transfer_from(
2340 self.source_butler._datastore, source_refs, transfer="split"
2341 )
2343 # Now try to get the same refs from the new butler.
2344 for ref in source_refs:
2345 if ref not in deleted:
2346 new_metric = self.target_butler.get(ref)
2347 old_metric = self.source_butler.get(ref)
2348 self.assertEqual(new_metric, old_metric)
2350 # Now prune run2 collection and create instead a CHAINED collection.
2351 # This should block the transfer.
2352 self.target_butler.removeRuns(["run2"], unstore=True)
2353 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2354 with self.assertRaises(CollectionTypeError):
2355 # Re-importing the run1 datasets can be problematic if they
2356 # use integer IDs so filter those out.
2357 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2358 self.target_butler.transfer_from(self.source_butler, to_transfer)
2361class ChainedDatastoreTransfers(PosixDatastoreTransfers):
2362 """Test transfers using a chained datastore."""
2364 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
2367class NullDatastoreTestCase(unittest.TestCase):
2368 """Test that we can fall back to a null datastore."""
2370 # Need a good config to create the repo.
2371 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
2372 storageClassFactory: StorageClassFactory
2374 @classmethod
2375 def setUpClass(cls) -> None:
2376 cls.storageClassFactory = StorageClassFactory()
2377 cls.storageClassFactory.addFromConfig(cls.configFile)
2379 def setUp(self) -> None:
2380 """Create a new butler root for each test."""
2381 self.root = makeTestTempDir(TESTDIR)
2382 Butler.makeRepo(self.root, config=Config(self.configFile))
2384 def tearDown(self) -> None:
2385 removeTestTempDir(self.root)
2387 def test_fallback(self) -> None:
2388 # Read the butler config and mess with the datastore section.
2389 bad_config = Config(os.path.join(self.root, "butler.yaml"))
2390 bad_config["datastore", "cls"] = "lsst.not.a.datastore.Datastore"
2392 with self.assertRaises(RuntimeError):
2393 Butler.from_config(bad_config)
2395 butler = Butler.from_config(bad_config, writeable=True, without_datastore=True)
2396 self.assertIsInstance(butler._datastore, NullDatastore)
2398 # Check that registry is working.
2399 butler.registry.registerRun("MYRUN")
2400 collections = butler.registry.queryCollections(...)
2401 self.assertIn("MYRUN", set(collections))
2403 # Create a ref.
2404 dimensions = butler.dimensions.extract([])
2405 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
2406 datasetTypeName = "metric"
2407 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2408 butler.registry.registerDatasetType(datasetType)
2409 ref = DatasetRef(datasetType, {}, run="MYRUN")
2411 # Check that datastore will complain.
2412 with self.assertRaises(FileNotFoundError):
2413 butler.get(ref)
2414 with self.assertRaises(FileNotFoundError):
2415 butler.getURI(ref)
2418def setup_module(module: types.ModuleType) -> None:
2419 """Set up the module for pytest."""
2420 clean_environment()
2423if __name__ == "__main__":
2424 clean_environment()
2425 unittest.main()