Coverage for tests/test_butler.py: 13%
1174 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-02 02:16 -0700
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-02 02:16 -0700
1# This file is part of daf_butler.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <http://www.gnu.org/licenses/>.
22"""Tests for Butler.
23"""
24from __future__ import annotations
26import gc
27import json
28import logging
29import os
30import pathlib
31import pickle
32import posixpath
33import random
34import shutil
35import string
36import tempfile
37import unittest
38import uuid
39from collections.abc import Mapping
40from typing import TYPE_CHECKING, Any, cast
42try:
43 import boto3
44 import botocore
45 from moto import mock_s3 # type: ignore[import]
46except ImportError:
47 boto3 = None
49 def mock_s3(cls):
50 """A no-op decorator in case moto mock_s3 can not be imported."""
51 return cls
54try:
55 # It's possible but silly to have testing.postgresql installed without
56 # having the postgresql server installed (because then nothing in
57 # testing.postgresql would work), so we use the presence of that module
58 # to test whether we can expect the server to be available.
59 import testing.postgresql # type: ignore[import]
60except ImportError:
61 testing = None
63import astropy.time
64import sqlalchemy
65from lsst.daf.butler import (
66 Butler,
67 ButlerConfig,
68 CollectionType,
69 Config,
70 DataCoordinate,
71 DatasetRef,
72 DatasetType,
73 FileDataset,
74 FileTemplate,
75 FileTemplateValidationError,
76 StorageClassFactory,
77 ValidationError,
78 script,
79)
80from lsst.daf.butler.core.repoRelocation import BUTLER_ROOT_TAG
81from lsst.daf.butler.datastores.fileDatastore import FileDatastore
82from lsst.daf.butler.registries.sql import SqlRegistry
83from lsst.daf.butler.registry import (
84 CollectionError,
85 CollectionTypeError,
86 ConflictingDefinitionError,
87 DataIdValueError,
88 MissingCollectionError,
89 OrphanedRecordError,
90)
91from lsst.daf.butler.tests import MetricsExample, MultiDetectorFormatter
92from lsst.daf.butler.tests.utils import TestCaseMixin, makeTestTempDir, removeTestTempDir, safeTestTempDir
93from lsst.resources import ResourcePath
94from lsst.resources.s3utils import setAwsEnvCredentials, unsetAwsEnvCredentials
95from lsst.utils import doImportType
96from lsst.utils.ellipsis import Ellipsis
97from lsst.utils.introspection import get_full_type_name
99if TYPE_CHECKING:
100 from lsst.daf.butler import Datastore, DimensionGraph, Registry, StorageClass
102TESTDIR = os.path.abspath(os.path.dirname(__file__))
105def makeExampleMetrics():
106 return MetricsExample(
107 {"AM1": 5.2, "AM2": 30.6},
108 {"a": [1, 2, 3], "b": {"blue": 5, "red": "green"}},
109 [563, 234, 456.7, 752, 8, 9, 27],
110 )
113class TransactionTestError(Exception):
114 """Specific error for testing transactions, to prevent misdiagnosing
115 that might otherwise occur when a standard exception is used.
116 """
118 pass
121class ButlerConfigTests(unittest.TestCase):
122 """Simple tests for ButlerConfig that are not tested in any other test
123 cases."""
125 def testSearchPath(self):
126 configFile = os.path.join(TESTDIR, "config", "basic", "butler.yaml")
127 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
128 config1 = ButlerConfig(configFile)
129 self.assertNotIn("testConfigs", "\n".join(cm.output))
131 overrideDirectory = os.path.join(TESTDIR, "config", "testConfigs")
132 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
133 config2 = ButlerConfig(configFile, searchPaths=[overrideDirectory])
134 self.assertIn("testConfigs", "\n".join(cm.output))
136 key = ("datastore", "records", "table")
137 self.assertNotEqual(config1[key], config2[key])
138 self.assertEqual(config2[key], "override_record")
141class ButlerPutGetTests(TestCaseMixin):
142 """Helper method for running a suite of put/get tests from different
143 butler configurations."""
145 root: str
146 default_run = "ingésτ😺"
147 storageClassFactory: StorageClassFactory
148 configFile: str
149 tmpConfigFile: str
151 @staticmethod
152 def addDatasetType(
153 datasetTypeName: str, dimensions: DimensionGraph, storageClass: StorageClass | str, registry: Registry
154 ) -> DatasetType:
155 """Create a DatasetType and register it"""
156 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
157 registry.registerDatasetType(datasetType)
158 return datasetType
160 @classmethod
161 def setUpClass(cls) -> None:
162 cls.storageClassFactory = StorageClassFactory()
163 cls.storageClassFactory.addFromConfig(cls.configFile)
165 def assertGetComponents(self, butler, datasetRef, components, reference, collections=None) -> None:
166 datasetType = datasetRef.datasetType
167 dataId = datasetRef.dataId
168 deferred = butler.getDeferred(datasetRef)
170 for component in components:
171 compTypeName = datasetType.componentTypeName(component)
172 result = butler.get(compTypeName, dataId, collections=collections)
173 self.assertEqual(result, getattr(reference, component))
174 result_deferred = deferred.get(component=component)
175 self.assertEqual(result_deferred, result)
177 def tearDown(self) -> None:
178 removeTestTempDir(self.root)
180 def create_butler(
181 self, run: str, storageClass: StorageClass | str, datasetTypeName: str
182 ) -> tuple[Butler, DatasetType]:
183 butler = Butler(self.tmpConfigFile, run=run)
185 collections = set(butler.registry.queryCollections())
186 self.assertEqual(collections, set([run]))
188 # Create and register a DatasetType
189 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
191 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
193 # Add needed Dimensions
194 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
195 butler.registry.insertDimensionData(
196 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
197 )
198 butler.registry.insertDimensionData(
199 "visit_system", {"instrument": "DummyCamComp", "id": 1, "name": "default"}
200 )
201 visit_start = astropy.time.Time("2020-01-01 08:00:00.123456789", scale="tai")
202 visit_end = astropy.time.Time("2020-01-01 08:00:36.66", scale="tai")
203 butler.registry.insertDimensionData(
204 "visit",
205 {
206 "instrument": "DummyCamComp",
207 "id": 423,
208 "name": "fourtwentythree",
209 "physical_filter": "d-r",
210 "visit_system": 1,
211 "datetime_begin": visit_start,
212 "datetime_end": visit_end,
213 },
214 )
216 # Add more visits for some later tests
217 for visit_id in (424, 425):
218 butler.registry.insertDimensionData(
219 "visit",
220 {
221 "instrument": "DummyCamComp",
222 "id": visit_id,
223 "name": f"fourtwentyfour_{visit_id}",
224 "physical_filter": "d-r",
225 "visit_system": 1,
226 },
227 )
228 return butler, datasetType
230 def runPutGetTest(self, storageClass: StorageClass, datasetTypeName: str) -> Butler:
231 # New datasets will be added to run and tag, but we will only look in
232 # tag when looking up datasets.
233 run = self.default_run
234 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
235 assert butler.run is not None
237 # Create and store a dataset
238 metric = makeExampleMetrics()
239 dataId = butler.registry.expandDataId({"instrument": "DummyCamComp", "visit": 423})
241 # Put and remove the dataset once as a DatasetRef, once as a dataId,
242 # and once with a DatasetType
244 # Keep track of any collections we add and do not clean up
245 expected_collections = {run}
247 counter = 0
248 ref = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run="put_run_1")
249 args = tuple[DatasetRef] | tuple[str | DatasetType, DataCoordinate]
250 for args in ((ref,), (datasetTypeName, dataId), (datasetType, dataId)):
251 # Since we are using subTest we can get cascading failures
252 # here with the first attempt failing and the others failing
253 # immediately because the dataset already exists. Work around
254 # this by using a distinct run collection each time
255 counter += 1
256 this_run = f"put_run_{counter}"
257 butler.registry.registerCollection(this_run, type=CollectionType.RUN)
258 expected_collections.update({this_run})
260 with self.subTest(args=args):
261 ref = butler.put(metric, *args, run=this_run)
262 self.assertIsInstance(ref, DatasetRef)
264 # Test getDirect
265 metricOut = butler.get(ref)
266 self.assertEqual(metric, metricOut)
267 # Test get
268 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
269 self.assertEqual(metric, metricOut)
270 # Test get with a datasetRef
271 metricOut = butler.get(ref, collections=this_run)
272 self.assertEqual(metric, metricOut)
273 # Test getDeferred with dataId
274 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
275 self.assertEqual(metric, metricOut)
276 # Test getDeferred with a datasetRef
277 metricOut = butler.getDeferred(ref, collections=this_run).get()
278 self.assertEqual(metric, metricOut)
279 # and deferred direct with ref
280 metricOut = butler.getDeferred(ref).get()
281 self.assertEqual(metric, metricOut)
283 # Check we can get components
284 if storageClass.isComposite():
285 self.assertGetComponents(
286 butler, ref, ("summary", "data", "output"), metric, collections=this_run
287 )
289 # Can the artifacts themselves be retrieved?
290 if not butler.datastore.isEphemeral:
291 root_uri = ResourcePath(self.root)
293 for preserve_path in (True, False):
294 destination = root_uri.join(f"artifacts/{preserve_path}_{counter}/")
295 # Use copy so that we can test that overwrite
296 # protection works (using "auto" for File URIs would
297 # use hard links and subsequent transfer would work
298 # because it knows they are the same file).
299 transferred = butler.retrieveArtifacts(
300 [ref], destination, preserve_path=preserve_path, transfer="copy"
301 )
302 self.assertGreater(len(transferred), 0)
303 artifacts = list(ResourcePath.findFileResources([destination]))
304 self.assertEqual(set(transferred), set(artifacts))
306 for artifact in transferred:
307 path_in_destination = artifact.relative_to(destination)
308 self.assertIsNotNone(path_in_destination)
309 assert path_in_destination is not None
311 # when path is not preserved there should not be
312 # any path separators.
313 num_seps = path_in_destination.count("/")
314 if preserve_path:
315 self.assertGreater(num_seps, 0)
316 else:
317 self.assertEqual(num_seps, 0)
319 primary_uri, secondary_uris = butler.datastore.getURIs(ref)
320 n_uris = len(secondary_uris)
321 if primary_uri:
322 n_uris += 1
323 self.assertEqual(
324 len(artifacts),
325 n_uris,
326 "Comparing expected artifacts vs actual:"
327 f" {artifacts} vs {primary_uri} and {secondary_uris}",
328 )
330 if preserve_path:
331 # No need to run these twice
332 with self.assertRaises(ValueError):
333 butler.retrieveArtifacts([ref], destination, transfer="move")
335 with self.assertRaises(FileExistsError):
336 butler.retrieveArtifacts([ref], destination)
338 transferred_again = butler.retrieveArtifacts(
339 [ref], destination, preserve_path=preserve_path, overwrite=True
340 )
341 self.assertEqual(set(transferred_again), set(transferred))
343 # Now remove the dataset completely.
344 butler.pruneDatasets([ref], purge=True, unstore=True)
345 # Lookup with original args should still fail.
346 with self.assertRaises(LookupError):
347 butler.datasetExists(*args, collections=this_run)
348 # get() should still fail.
349 with self.assertRaises(FileNotFoundError):
350 butler.get(ref)
351 # Registry shouldn't be able to find it by dataset_id anymore.
352 self.assertIsNone(butler.registry.getDataset(ref.id))
354 # Do explicit registry removal since we know they are
355 # empty
356 butler.registry.removeCollection(this_run)
357 expected_collections.remove(this_run)
359 # Create DatasetRef for put using default run.
360 refIn = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run=butler.run)
362 # Put the dataset again, since the last thing we did was remove it
363 # and we want to use the default collection.
364 ref = butler.put(metric, refIn)
366 # Get with parameters
367 stop = 4
368 sliced = butler.get(ref, parameters={"slice": slice(stop)})
369 self.assertNotEqual(metric, sliced)
370 self.assertEqual(metric.summary, sliced.summary)
371 self.assertEqual(metric.output, sliced.output)
372 self.assertEqual(metric.data[:stop], sliced.data)
373 # getDeferred with parameters
374 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
375 self.assertNotEqual(metric, sliced)
376 self.assertEqual(metric.summary, sliced.summary)
377 self.assertEqual(metric.output, sliced.output)
378 self.assertEqual(metric.data[:stop], sliced.data)
379 # getDeferred with deferred parameters
380 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
381 self.assertNotEqual(metric, sliced)
382 self.assertEqual(metric.summary, sliced.summary)
383 self.assertEqual(metric.output, sliced.output)
384 self.assertEqual(metric.data[:stop], sliced.data)
386 if storageClass.isComposite():
387 # Check that components can be retrieved
388 metricOut = butler.get(ref.datasetType.name, dataId)
389 compNameS = ref.datasetType.componentTypeName("summary")
390 compNameD = ref.datasetType.componentTypeName("data")
391 summary = butler.get(compNameS, dataId)
392 self.assertEqual(summary, metric.summary)
393 data = butler.get(compNameD, dataId)
394 self.assertEqual(data, metric.data)
396 if "counter" in storageClass.derivedComponents:
397 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
398 self.assertEqual(count, len(data))
400 count = butler.get(
401 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
402 )
403 self.assertEqual(count, stop)
405 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections)
406 assert compRef is not None
407 summary = butler.get(compRef)
408 self.assertEqual(summary, metric.summary)
410 # Create a Dataset type that has the same name but is inconsistent.
411 inconsistentDatasetType = DatasetType(
412 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
413 )
415 # Getting with a dataset type that does not match registry fails
416 with self.assertRaisesRegex(ValueError, "Supplied dataset type .* inconsistent with registry"):
417 butler.get(inconsistentDatasetType, dataId)
419 # Combining a DatasetRef with a dataId should fail
420 with self.assertRaisesRegex(ValueError, "DatasetRef given, cannot use dataId as well"):
421 butler.get(ref, dataId)
422 # Getting with an explicit ref should fail if the id doesn't match.
423 with self.assertRaises(FileNotFoundError):
424 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=uuid.UUID(int=101), run=butler.run))
426 # Getting a dataset with unknown parameters should fail
427 with self.assertRaisesRegex(KeyError, "Parameter 'unsupported' not understood"):
428 butler.get(ref, parameters={"unsupported": True})
430 # Check we have a collection
431 collections = set(butler.registry.queryCollections())
432 self.assertEqual(collections, expected_collections)
434 # Clean up to check that we can remove something that may have
435 # already had a component removed
436 butler.pruneDatasets([ref], unstore=True, purge=True)
438 # Add the same ref again, so we can check that duplicate put fails.
439 ref = butler.put(metric, datasetType, dataId)
441 # Repeat put will fail.
442 with self.assertRaisesRegex(
443 ConflictingDefinitionError, "A database constraint failure was triggered"
444 ):
445 butler.put(metric, datasetType, dataId)
447 # Remove the datastore entry.
448 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
450 # Put will still fail
451 with self.assertRaisesRegex(
452 ConflictingDefinitionError, "A database constraint failure was triggered"
453 ):
454 butler.put(metric, datasetType, dataId)
456 # Repeat the same sequence with resolved ref.
457 butler.pruneDatasets([ref], unstore=True, purge=True)
458 ref = butler.put(metric, refIn)
460 # Repeat put will fail.
461 with self.assertRaisesRegex(ConflictingDefinitionError, "Datastore already contains dataset"):
462 butler.put(metric, refIn)
464 # Remove the datastore entry.
465 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
467 # In case of resolved ref this write will succeed.
468 ref = butler.put(metric, refIn)
470 # Leave the dataset in place since some downstream tests require
471 # something to be present
473 return butler
475 def testDeferredCollectionPassing(self) -> None:
476 # Construct a butler with no run or collection, but make it writeable.
477 butler = Butler(self.tmpConfigFile, writeable=True)
478 # Create and register a DatasetType
479 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
480 datasetType = self.addDatasetType(
481 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
482 )
483 # Add needed Dimensions
484 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
485 butler.registry.insertDimensionData(
486 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
487 )
488 butler.registry.insertDimensionData(
489 "visit",
490 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
491 )
492 dataId = {"instrument": "DummyCamComp", "visit": 423}
493 # Create dataset.
494 metric = makeExampleMetrics()
495 # Register a new run and put dataset.
496 run = "deferred"
497 self.assertTrue(butler.registry.registerRun(run))
498 # Second time it will be allowed but indicate no-op
499 self.assertFalse(butler.registry.registerRun(run))
500 ref = butler.put(metric, datasetType, dataId, run=run)
501 # Putting with no run should fail with TypeError.
502 with self.assertRaises(CollectionError):
503 butler.put(metric, datasetType, dataId)
504 # Dataset should exist.
505 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
506 # We should be able to get the dataset back, but with and without
507 # a deferred dataset handle.
508 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
509 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
510 # Trying to find the dataset without any collection is a TypeError.
511 with self.assertRaises(CollectionError):
512 butler.datasetExists(datasetType, dataId)
513 with self.assertRaises(CollectionError):
514 butler.get(datasetType, dataId)
515 # Associate the dataset with a different collection.
516 butler.registry.registerCollection("tagged")
517 butler.registry.associate("tagged", [ref])
518 # Deleting the dataset from the new collection should make it findable
519 # in the original collection.
520 butler.pruneDatasets([ref], tags=["tagged"])
521 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
524class ButlerTests(ButlerPutGetTests):
525 """Tests for Butler."""
527 useTempRoot = True
528 validationCanFail: bool
529 fullConfigKey: str | None
530 registryStr: str | None
531 datastoreName: list[str] | None
532 datastoreStr: list[str]
534 def setUp(self) -> None:
535 """Create a new butler root for each test."""
536 self.root = makeTestTempDir(TESTDIR)
537 Butler.makeRepo(self.root, config=Config(self.configFile))
538 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
540 def testConstructor(self) -> None:
541 """Independent test of constructor."""
542 butler = Butler(self.tmpConfigFile, run=self.default_run)
543 self.assertIsInstance(butler, Butler)
545 # Check that butler.yaml is added automatically.
546 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
547 config_dir = self.tmpConfigFile[: -len(end)]
548 butler = Butler(config_dir, run=self.default_run)
549 self.assertIsInstance(butler, Butler)
551 # Even with a ResourcePath.
552 butler = Butler(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
553 self.assertIsInstance(butler, Butler)
555 collections = set(butler.registry.queryCollections())
556 self.assertEqual(collections, {self.default_run})
558 # Check that some special characters can be included in run name.
559 special_run = "u@b.c-A"
560 butler_special = Butler(butler=butler, run=special_run)
561 collections = set(butler_special.registry.queryCollections("*@*"))
562 self.assertEqual(collections, {special_run})
564 butler2 = Butler(butler=butler, collections=["other"])
565 self.assertEqual(butler2.collections, ("other",))
566 self.assertIsNone(butler2.run)
567 self.assertIs(butler.datastore, butler2.datastore)
569 # Test that we can use an environment variable to find this
570 # repository.
571 butler_index = Config()
572 butler_index["label"] = self.tmpConfigFile
573 for suffix in (".yaml", ".json"):
574 # Ensure that the content differs so that we know that
575 # we aren't reusing the cache.
576 bad_label = f"s3://bucket/not_real{suffix}"
577 butler_index["bad_label"] = bad_label
578 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
579 butler_index.dumpToUri(temp_file)
580 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
581 self.assertEqual(Butler.get_known_repos(), set(("label", "bad_label")))
582 uri = Butler.get_repo_uri("bad_label")
583 self.assertEqual(uri, ResourcePath(bad_label))
584 uri = Butler.get_repo_uri("label")
585 butler = Butler(uri, writeable=False)
586 self.assertIsInstance(butler, Butler)
587 butler = Butler("label", writeable=False)
588 self.assertIsInstance(butler, Butler)
589 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
590 Butler("not_there", writeable=False)
591 with self.assertRaises(KeyError) as cm:
592 Butler.get_repo_uri("missing")
593 self.assertEqual(Butler.get_repo_uri("missing", True), ResourcePath("missing"))
594 self.assertIn("not known to", str(cm.exception))
595 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
596 with self.assertRaises(FileNotFoundError):
597 Butler.get_repo_uri("label")
598 self.assertEqual(Butler.get_known_repos(), set())
599 with self.assertRaises(KeyError) as cm:
600 # No environment variable set.
601 Butler.get_repo_uri("label")
602 self.assertEqual(Butler.get_repo_uri("label", True), ResourcePath("label"))
603 self.assertIn("No repository index defined", str(cm.exception))
604 with self.assertRaisesRegex(FileNotFoundError, "no known aliases"):
605 # No aliases registered.
606 Butler("not_there")
607 self.assertEqual(Butler.get_known_repos(), set())
609 def testBasicPutGet(self) -> None:
610 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
611 self.runPutGetTest(storageClass, "test_metric")
613 def testCompositePutGetConcrete(self) -> None:
614 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
615 butler = self.runPutGetTest(storageClass, "test_metric")
617 # Should *not* be disassembled
618 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
619 self.assertEqual(len(datasets), 1)
620 uri, components = butler.getURIs(datasets[0])
621 self.assertIsInstance(uri, ResourcePath)
622 self.assertFalse(components)
623 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
624 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
626 # Predicted dataset
627 dataId = {"instrument": "DummyCamComp", "visit": 424}
628 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
629 self.assertFalse(components)
630 self.assertIsInstance(uri, ResourcePath)
631 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
632 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
634 def testCompositePutGetVirtual(self) -> None:
635 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
636 butler = self.runPutGetTest(storageClass, "test_metric_comp")
638 # Should be disassembled
639 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
640 self.assertEqual(len(datasets), 1)
641 uri, components = butler.getURIs(datasets[0])
643 if butler.datastore.isEphemeral:
644 # Never disassemble in-memory datastore
645 self.assertIsInstance(uri, ResourcePath)
646 self.assertFalse(components)
647 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
648 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
649 else:
650 self.assertIsNone(uri)
651 self.assertEqual(set(components), set(storageClass.components))
652 for compuri in components.values():
653 self.assertIsInstance(compuri, ResourcePath)
654 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
655 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
657 # Predicted dataset
658 dataId = {"instrument": "DummyCamComp", "visit": 424}
659 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
661 if butler.datastore.isEphemeral:
662 # Never disassembled
663 self.assertIsInstance(uri, ResourcePath)
664 self.assertFalse(components)
665 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
666 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
667 else:
668 self.assertIsNone(uri)
669 self.assertEqual(set(components), set(storageClass.components))
670 for compuri in components.values():
671 self.assertIsInstance(compuri, ResourcePath)
672 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
673 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
675 def testStorageClassOverrideGet(self) -> None:
676 """Test storage class conversion on get with override."""
677 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
678 datasetTypeName = "anything"
679 run = self.default_run
681 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
683 # Create and store a dataset.
684 metric = makeExampleMetrics()
685 dataId = {"instrument": "DummyCamComp", "visit": 423}
687 ref = butler.put(metric, datasetType, dataId)
689 # Return native type.
690 retrieved = butler.get(ref)
691 self.assertEqual(retrieved, metric)
693 # Specify an override.
694 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
695 model = butler.get(ref, storageClass=new_sc)
696 self.assertNotEqual(type(model), type(retrieved))
697 self.assertIs(type(model), new_sc.pytype)
698 self.assertEqual(retrieved, model)
700 # Defer but override later.
701 deferred = butler.getDeferred(ref)
702 model = deferred.get(storageClass=new_sc)
703 self.assertIs(type(model), new_sc.pytype)
704 self.assertEqual(retrieved, model)
706 # Defer but override up front.
707 deferred = butler.getDeferred(ref, storageClass=new_sc)
708 model = deferred.get()
709 self.assertIs(type(model), new_sc.pytype)
710 self.assertEqual(retrieved, model)
712 # Retrieve a component. Should be a tuple.
713 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
714 self.assertIs(type(data), tuple)
715 self.assertEqual(data, tuple(retrieved.data))
717 # Parameter on the write storage class should work regardless
718 # of read storage class.
719 data = butler.get(
720 "anything.data",
721 dataId,
722 storageClass="StructuredDataDataTestTuple",
723 parameters={"slice": slice(2, 4)},
724 )
725 self.assertEqual(len(data), 2)
727 # Try a parameter that is known to the read storage class but not
728 # the write storage class.
729 with self.assertRaises(KeyError):
730 butler.get(
731 "anything.data",
732 dataId,
733 storageClass="StructuredDataDataTestTuple",
734 parameters={"xslice": slice(2, 4)},
735 )
737 def testPytypePutCoercion(self) -> None:
738 """Test python type coercion on Butler.get and put."""
740 # Store some data with the normal example storage class.
741 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
742 datasetTypeName = "test_metric"
743 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
745 dataId = {"instrument": "DummyCamComp", "visit": 423}
747 # Put a dict and this should coerce to a MetricsExample
748 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
749 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
750 test_metric = butler.get(metric_ref)
751 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
752 self.assertEqual(test_metric.summary, test_dict["summary"])
753 self.assertEqual(test_metric.output, test_dict["output"])
755 # Check that the put still works if a DatasetType is given with
756 # a definition matching this python type.
757 registry_type = butler.registry.getDatasetType(datasetTypeName)
758 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
759 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
760 self.assertEqual(metric2_ref.datasetType, registry_type)
762 # The get will return the type expected by registry.
763 test_metric2 = butler.get(metric2_ref)
764 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
766 # Make a new DatasetRef with the compatible but different DatasetType.
767 # This should now return a dict.
768 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
769 test_dict2 = butler.get(new_ref)
770 self.assertEqual(get_full_type_name(test_dict2), "dict")
772 # Get it again with the wrong dataset type definition using get()
773 # rather than get(). This should be consistent with get()
774 # behavior and return the type of the DatasetType.
775 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
776 self.assertEqual(get_full_type_name(test_dict3), "dict")
778 def testIngest(self) -> None:
779 butler = Butler(self.tmpConfigFile, run=self.default_run)
781 # Create and register a DatasetType
782 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "detector"])
784 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
785 datasetTypeName = "metric"
787 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
789 # Add needed Dimensions
790 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
791 butler.registry.insertDimensionData(
792 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
793 )
794 for detector in (1, 2):
795 butler.registry.insertDimensionData(
796 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
797 )
799 butler.registry.insertDimensionData(
800 "visit",
801 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
802 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
803 )
805 formatter = doImportType("lsst.daf.butler.formatters.yaml.YamlFormatter")
806 dataRoot = os.path.join(TESTDIR, "data", "basic")
807 datasets = []
808 for detector in (1, 2):
809 detector_name = f"detector_{detector}"
810 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
811 dataId = butler.registry.expandDataId(
812 {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
813 )
814 # Create a DatasetRef for ingest
815 refIn = DatasetRef(datasetType, dataId, run=self.default_run)
817 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
819 butler.ingest(*datasets, transfer="copy")
821 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
822 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
824 metrics1 = butler.get(datasetTypeName, dataId1)
825 metrics2 = butler.get(datasetTypeName, dataId2)
826 self.assertNotEqual(metrics1, metrics2)
828 # Compare URIs
829 uri1 = butler.getURI(datasetTypeName, dataId1)
830 uri2 = butler.getURI(datasetTypeName, dataId2)
831 self.assertNotEqual(uri1, uri2)
833 # Now do a multi-dataset but single file ingest
834 metricFile = os.path.join(dataRoot, "detectors.yaml")
835 refs = []
836 for detector in (1, 2):
837 detector_name = f"detector_{detector}"
838 dataId = butler.registry.expandDataId(
839 {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
840 )
841 # Create a DatasetRef for ingest
842 refs.append(DatasetRef(datasetType, dataId, run=self.default_run))
844 # Test "move" transfer to ensure that the files themselves
845 # have disappeared following ingest.
846 with ResourcePath.temporary_uri(suffix=".yaml") as tempFile:
847 tempFile.transfer_from(ResourcePath(metricFile), transfer="copy")
849 datasets = []
850 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter))
852 # For first ingest use copy.
853 butler.ingest(*datasets, transfer="copy", record_validation_info=False)
855 # Now try to ingest again in "execution butler" mode where
856 # the registry entries exist but the datastore does not have
857 # the files. We also need to strip the dimension records to ensure
858 # that they will be re-added by the ingest.
859 ref = datasets[0].refs[0]
860 datasets[0].refs = [
861 cast(
862 DatasetRef,
863 butler.registry.findDataset(ref.datasetType, dataId=ref.dataId, collections=ref.run),
864 )
865 for ref in datasets[0].refs
866 ]
867 all_refs = []
868 for dataset in datasets:
869 refs = []
870 for ref in dataset.refs:
871 # Create a dict from the dataId to drop the records.
872 new_data_id = {str(k): v for k, v in ref.dataId.items()}
873 new_ref = butler.registry.findDataset(ref.datasetType, new_data_id, collections=ref.run)
874 assert new_ref is not None
875 self.assertFalse(new_ref.dataId.hasRecords())
876 refs.append(new_ref)
877 dataset.refs = refs
878 all_refs.extend(dataset.refs)
879 butler.pruneDatasets(all_refs, disassociate=False, unstore=True, purge=False)
881 # Use move mode to test that the file is deleted. Also
882 # disable recording of file size.
883 butler.ingest(*datasets, transfer="move", record_validation_info=False)
885 # Check that every ref now has records.
886 for dataset in datasets:
887 for ref in dataset.refs:
888 self.assertTrue(ref.dataId.hasRecords())
890 # Ensure that the file has disappeared.
891 self.assertFalse(tempFile.exists())
893 # Check that the datastore recorded no file size.
894 # Not all datastores can support this.
895 try:
896 infos = butler.datastore.getStoredItemsInfo(datasets[0].refs[0]) # type: ignore[attr-defined]
897 self.assertEqual(infos[0].file_size, -1)
898 except AttributeError:
899 pass
901 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
902 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
904 multi1 = butler.get(datasetTypeName, dataId1)
905 multi2 = butler.get(datasetTypeName, dataId2)
907 self.assertEqual(multi1, metrics1)
908 self.assertEqual(multi2, metrics2)
910 # Compare URIs
911 uri1 = butler.getURI(datasetTypeName, dataId1)
912 uri2 = butler.getURI(datasetTypeName, dataId2)
913 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
915 # Test that removing one does not break the second
916 # This line will issue a warning log message for a ChainedDatastore
917 # that uses an InMemoryDatastore since in-memory can not ingest
918 # files.
919 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
920 self.assertFalse(butler.datasetExists(datasetTypeName, dataId1))
921 self.assertTrue(butler.datasetExists(datasetTypeName, dataId2))
922 multi2b = butler.get(datasetTypeName, dataId2)
923 self.assertEqual(multi2, multi2b)
925 # Ensure we can ingest 0 datasets
926 datasets = []
927 butler.ingest(*datasets)
929 def testPickle(self) -> None:
930 """Test pickle support."""
931 butler = Butler(self.tmpConfigFile, run=self.default_run)
932 butlerOut = pickle.loads(pickle.dumps(butler))
933 self.assertIsInstance(butlerOut, Butler)
934 self.assertEqual(butlerOut._config, butler._config)
935 self.assertEqual(butlerOut.collections, butler.collections)
936 self.assertEqual(butlerOut.run, butler.run)
938 def testGetDatasetTypes(self) -> None:
939 butler = Butler(self.tmpConfigFile, run=self.default_run)
940 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "physical_filter"])
941 dimensionEntries: list[tuple[str, list[Mapping[str, Any]]]] = [
942 (
943 "instrument",
944 [
945 {"instrument": "DummyCam"},
946 {"instrument": "DummyHSC"},
947 {"instrument": "DummyCamComp"},
948 ],
949 ),
950 ("physical_filter", [{"instrument": "DummyCam", "name": "d-r", "band": "R"}]),
951 ("visit", [{"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}]),
952 ]
953 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
954 # Add needed Dimensions
955 for element, data in dimensionEntries:
956 butler.registry.insertDimensionData(element, *data)
958 # When a DatasetType is added to the registry entries are not created
959 # for components but querying them can return the components.
960 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
961 components = set()
962 for datasetTypeName in datasetTypeNames:
963 # Create and register a DatasetType
964 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
966 for componentName in storageClass.components:
967 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
969 fromRegistry: set[DatasetType] = set()
970 for parent_dataset_type in butler.registry.queryDatasetTypes():
971 fromRegistry.add(parent_dataset_type)
972 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
973 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
975 # Now that we have some dataset types registered, validate them
976 butler.validateConfiguration(
977 ignore=[
978 "test_metric_comp",
979 "metric3",
980 "metric5",
981 "calexp",
982 "DummySC",
983 "datasetType.component",
984 "random_data",
985 "random_data_2",
986 ]
987 )
989 # Add a new datasetType that will fail template validation
990 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
991 if self.validationCanFail:
992 with self.assertRaises(ValidationError):
993 butler.validateConfiguration()
995 # Rerun validation but with a subset of dataset type names
996 butler.validateConfiguration(datasetTypeNames=["metric4"])
998 # Rerun validation but ignore the bad datasetType
999 butler.validateConfiguration(
1000 ignore=[
1001 "test_metric_comp",
1002 "metric3",
1003 "metric5",
1004 "calexp",
1005 "DummySC",
1006 "datasetType.component",
1007 "random_data",
1008 "random_data_2",
1009 ]
1010 )
1012 def testTransaction(self) -> None:
1013 butler = Butler(self.tmpConfigFile, run=self.default_run)
1014 datasetTypeName = "test_metric"
1015 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1016 dimensionEntries: tuple[tuple[str, Mapping[str, Any]], ...] = (
1017 ("instrument", {"instrument": "DummyCam"}),
1018 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1019 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1020 )
1021 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1022 metric = makeExampleMetrics()
1023 dataId = {"instrument": "DummyCam", "visit": 42}
1024 # Create and register a DatasetType
1025 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1026 with self.assertRaises(TransactionTestError):
1027 with butler.transaction():
1028 # Add needed Dimensions
1029 for args in dimensionEntries:
1030 butler.registry.insertDimensionData(*args)
1031 # Store a dataset
1032 ref = butler.put(metric, datasetTypeName, dataId)
1033 self.assertIsInstance(ref, DatasetRef)
1034 # Test getDirect
1035 metricOut = butler.get(ref)
1036 self.assertEqual(metric, metricOut)
1037 # Test get
1038 metricOut = butler.get(datasetTypeName, dataId)
1039 self.assertEqual(metric, metricOut)
1040 # Check we can get components
1041 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1042 raise TransactionTestError("This should roll back the entire transaction")
1043 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1044 butler.registry.expandDataId(dataId)
1045 # Should raise LookupError for missing data ID value
1046 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1047 butler.get(datasetTypeName, dataId)
1048 # Also check explicitly if Dataset entry is missing
1049 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections))
1050 # Direct retrieval should not find the file in the Datastore
1051 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1052 butler.get(ref)
1054 def testMakeRepo(self) -> None:
1055 """Test that we can write butler configuration to a new repository via
1056 the Butler.makeRepo interface and then instantiate a butler from the
1057 repo root.
1058 """
1059 # Do not run the test if we know this datastore configuration does
1060 # not support a file system root
1061 if self.fullConfigKey is None:
1062 return
1064 # create two separate directories
1065 root1 = tempfile.mkdtemp(dir=self.root)
1066 root2 = tempfile.mkdtemp(dir=self.root)
1068 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1069 limited = Config(self.configFile)
1070 butler1 = Butler(butlerConfig)
1071 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1072 full = Config(self.tmpConfigFile)
1073 butler2 = Butler(butlerConfig)
1074 # Butlers should have the same configuration regardless of whether
1075 # defaults were expanded.
1076 self.assertEqual(butler1._config, butler2._config)
1077 # Config files loaded directly should not be the same.
1078 self.assertNotEqual(limited, full)
1079 # Make sure "limited" doesn't have a few keys we know it should be
1080 # inheriting from defaults.
1081 self.assertIn(self.fullConfigKey, full)
1082 self.assertNotIn(self.fullConfigKey, limited)
1084 # Collections don't appear until something is put in them
1085 collections1 = set(butler1.registry.queryCollections())
1086 self.assertEqual(collections1, set())
1087 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1089 # Check that a config with no associated file name will not
1090 # work properly with relocatable Butler repo
1091 butlerConfig.configFile = None
1092 with self.assertRaises(ValueError):
1093 Butler(butlerConfig)
1095 with self.assertRaises(FileExistsError):
1096 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1098 def testStringification(self) -> None:
1099 butler = Butler(self.tmpConfigFile, run=self.default_run)
1100 butlerStr = str(butler)
1102 if self.datastoreStr is not None:
1103 for testStr in self.datastoreStr:
1104 self.assertIn(testStr, butlerStr)
1105 if self.registryStr is not None:
1106 self.assertIn(self.registryStr, butlerStr)
1108 datastoreName = butler.datastore.name
1109 if self.datastoreName is not None:
1110 for testStr in self.datastoreName:
1111 self.assertIn(testStr, datastoreName)
1113 def testButlerRewriteDataId(self) -> None:
1114 """Test that dataIds can be rewritten based on dimension records."""
1116 butler = Butler(self.tmpConfigFile, run=self.default_run)
1118 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1119 datasetTypeName = "random_data"
1121 # Create dimension records.
1122 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1123 butler.registry.insertDimensionData(
1124 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1125 )
1126 butler.registry.insertDimensionData(
1127 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1128 )
1130 dimensions = butler.registry.dimensions.extract(["instrument", "exposure"])
1131 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1132 butler.registry.registerDatasetType(datasetType)
1134 n_exposures = 5
1135 dayobs = 20210530
1137 for i in range(n_exposures):
1138 butler.registry.insertDimensionData(
1139 "exposure",
1140 {
1141 "instrument": "DummyCamComp",
1142 "id": i,
1143 "obs_id": f"exp{i}",
1144 "seq_num": i,
1145 "day_obs": dayobs,
1146 "physical_filter": "d-r",
1147 },
1148 )
1150 # Write some data.
1151 for i in range(n_exposures):
1152 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1154 # Use the seq_num for the put to test rewriting.
1155 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1156 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1158 # Check that the exposure is correct in the dataId
1159 self.assertEqual(ref.dataId["exposure"], i)
1161 # and check that we can get the dataset back with the same dataId
1162 new_metric = butler.get(datasetTypeName, dataId=dataId)
1163 self.assertEqual(new_metric, metric)
1166class FileDatastoreButlerTests(ButlerTests):
1167 """Common tests and specialization of ButlerTests for butlers backed
1168 by datastores that inherit from FileDatastore.
1169 """
1171 def checkFileExists(self, root: str | ResourcePath, relpath: str | ResourcePath) -> bool:
1172 """Checks if file exists at a given path (relative to root).
1174 Test testPutTemplates verifies actual physical existance of the files
1175 in the requested location.
1176 """
1177 uri = ResourcePath(root, forceDirectory=True)
1178 return uri.join(relpath).exists()
1180 def testPutTemplates(self) -> None:
1181 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1182 butler = Butler(self.tmpConfigFile, run=self.default_run)
1184 # Add needed Dimensions
1185 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1186 butler.registry.insertDimensionData(
1187 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1188 )
1189 butler.registry.insertDimensionData(
1190 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1191 )
1192 butler.registry.insertDimensionData(
1193 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1194 )
1196 # Create and store a dataset
1197 metric = makeExampleMetrics()
1199 # Create two almost-identical DatasetTypes (both will use default
1200 # template)
1201 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1202 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1203 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1204 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1206 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1207 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1209 # Put with exactly the data ID keys needed
1210 ref = butler.put(metric, "metric1", dataId1)
1211 uri = butler.getURI(ref)
1212 self.assertTrue(uri.exists())
1213 self.assertTrue(
1214 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle")
1215 )
1217 # Check the template based on dimensions
1218 if hasattr(butler.datastore, "templates"):
1219 butler.datastore.templates.validateTemplates([ref])
1221 # Put with extra data ID keys (physical_filter is an optional
1222 # dependency); should not change template (at least the way we're
1223 # defining them to behave now; the important thing is that they
1224 # must be consistent).
1225 ref = butler.put(metric, "metric2", dataId2)
1226 uri = butler.getURI(ref)
1227 self.assertTrue(uri.exists())
1228 self.assertTrue(
1229 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle")
1230 )
1232 # Check the template based on dimensions
1233 if hasattr(butler.datastore, "templates"):
1234 butler.datastore.templates.validateTemplates([ref])
1236 # Use a template that has a typo in dimension record metadata.
1237 # Easier to test with a butler that has a ref with records attached.
1238 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1239 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1240 path = template.format(ref)
1241 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1243 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1244 with self.assertRaises(KeyError):
1245 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1246 template.format(ref)
1248 # Now use a file template that will not result in unique filenames
1249 with self.assertRaises(FileTemplateValidationError):
1250 butler.put(metric, "metric3", dataId1)
1252 def testImportExport(self) -> None:
1253 # Run put/get tests just to create and populate a repo.
1254 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1255 self.runImportExportTest(storageClass)
1257 @unittest.expectedFailure
1258 def testImportExportVirtualComposite(self) -> None:
1259 # Run put/get tests just to create and populate a repo.
1260 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1261 self.runImportExportTest(storageClass)
1263 def runImportExportTest(self, storageClass: StorageClass) -> None:
1264 """This test does an export to a temp directory and an import back
1265 into a new temp directory repo. It does not assume a posix datastore"""
1266 exportButler = self.runPutGetTest(storageClass, "test_metric")
1268 # Test that we must have a file extension.
1269 with self.assertRaises(ValueError):
1270 with exportButler.export(filename="dump", directory=".") as export:
1271 pass
1273 # Test that unknown format is not allowed.
1274 with self.assertRaises(ValueError):
1275 with exportButler.export(filename="dump.fits", directory=".") as export:
1276 pass
1278 # Test that the repo actually has at least one dataset.
1279 datasets = list(exportButler.registry.queryDatasets(..., collections=Ellipsis))
1280 self.assertGreater(len(datasets), 0)
1281 # Add a DimensionRecord that's unused by those datasets.
1282 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1283 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1284 # Export and then import datasets.
1285 with safeTestTempDir(TESTDIR) as exportDir:
1286 exportFile = os.path.join(exportDir, "exports.yaml")
1287 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1288 export.saveDatasets(datasets)
1289 # Export the same datasets again. This should quietly do
1290 # nothing because of internal deduplication, and it shouldn't
1291 # complain about being asked to export the "htm7" elements even
1292 # though there aren't any in these datasets or in the database.
1293 export.saveDatasets(datasets, elements=["htm7"])
1294 # Save one of the data IDs again; this should be harmless
1295 # because of internal deduplication.
1296 export.saveDataIds([datasets[0].dataId])
1297 # Save some dimension records directly.
1298 export.saveDimensionData("skymap", [skymapRecord])
1299 self.assertTrue(os.path.exists(exportFile))
1300 with safeTestTempDir(TESTDIR) as importDir:
1301 # We always want this to be a local posix butler
1302 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1303 # Calling script.butlerImport tests the implementation of the
1304 # butler command line interface "import" subcommand. Functions
1305 # in the script folder are generally considered protected and
1306 # should not be used as public api.
1307 with open(exportFile, "r") as f:
1308 script.butlerImport(
1309 importDir,
1310 export_file=f,
1311 directory=exportDir,
1312 transfer="auto",
1313 skip_dimensions=None,
1314 )
1315 importButler = Butler(importDir, run=self.default_run)
1316 for ref in datasets:
1317 with self.subTest(ref=ref):
1318 # Test for existence by passing in the DatasetType and
1319 # data ID separately, to avoid lookup by dataset_id.
1320 self.assertTrue(importButler.datasetExists(ref.datasetType, ref.dataId))
1321 self.assertEqual(
1322 list(importButler.registry.queryDimensionRecords("skymap")),
1323 [importButler.registry.dimensions["skymap"].RecordClass(**skymapRecord)],
1324 )
1326 def testRemoveRuns(self) -> None:
1327 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1328 butler = Butler(self.tmpConfigFile, writeable=True)
1329 # Load registry data with dimensions to hang datasets off of.
1330 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1331 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1332 # Add some RUN-type collection.
1333 run1 = "run1"
1334 butler.registry.registerRun(run1)
1335 run2 = "run2"
1336 butler.registry.registerRun(run2)
1337 # put a dataset in each
1338 metric = makeExampleMetrics()
1339 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1340 datasetType = self.addDatasetType(
1341 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1342 )
1343 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1344 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1345 uri1 = butler.getURI(ref1, collections=[run1])
1346 uri2 = butler.getURI(ref2, collections=[run2])
1348 with self.assertRaises(OrphanedRecordError):
1349 butler.registry.removeDatasetType(datasetType.name)
1351 # Remove from both runs with different values for unstore.
1352 butler.removeRuns([run1], unstore=True)
1353 butler.removeRuns([run2], unstore=False)
1354 # Should be nothing in registry for either one, and datastore should
1355 # not think either exists.
1356 with self.assertRaises(MissingCollectionError):
1357 butler.registry.getCollectionType(run1)
1358 with self.assertRaises(MissingCollectionError):
1359 butler.registry.getCollectionType(run2)
1360 self.assertFalse(butler.datastore.exists(ref1))
1361 self.assertFalse(butler.datastore.exists(ref2))
1362 # The ref we unstored should be gone according to the URI, but the
1363 # one we forgot should still be around.
1364 self.assertFalse(uri1.exists())
1365 self.assertTrue(uri2.exists())
1367 # Now that the collections have been pruned we can remove the
1368 # dataset type
1369 butler.registry.removeDatasetType(datasetType.name)
1371 with self.assertLogs("lsst.daf.butler.registries", "INFO") as cm:
1372 butler.registry.removeDatasetType(tuple(["test*", "test*"]))
1373 self.assertIn("not defined", "\n".join(cm.output))
1376class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1377 """PosixDatastore specialization of a butler"""
1379 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1380 fullConfigKey: str | None = ".datastore.formatters"
1381 validationCanFail = True
1382 datastoreStr = ["/tmp"]
1383 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1384 registryStr = "/gen3.sqlite3"
1386 def testPathConstructor(self) -> None:
1387 """Independent test of constructor using PathLike."""
1388 butler = Butler(self.tmpConfigFile, run=self.default_run)
1389 self.assertIsInstance(butler, Butler)
1391 # And again with a Path object with the butler yaml
1392 path = pathlib.Path(self.tmpConfigFile)
1393 butler = Butler(path, writeable=False)
1394 self.assertIsInstance(butler, Butler)
1396 # And again with a Path object without the butler yaml
1397 # (making sure we skip it if the tmp config doesn't end
1398 # in butler.yaml -- which is the case for a subclass)
1399 if self.tmpConfigFile.endswith("butler.yaml"):
1400 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1401 butler = Butler(path, writeable=False)
1402 self.assertIsInstance(butler, Butler)
1404 def testExportTransferCopy(self) -> None:
1405 """Test local export using all transfer modes"""
1406 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1407 exportButler = self.runPutGetTest(storageClass, "test_metric")
1408 # Test that the repo actually has at least one dataset.
1409 datasets = list(exportButler.registry.queryDatasets(..., collections=Ellipsis))
1410 self.assertGreater(len(datasets), 0)
1411 uris = [exportButler.getURI(d) for d in datasets]
1412 assert isinstance(exportButler.datastore, FileDatastore)
1413 datastoreRoot = exportButler.datastore.root
1415 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1417 for path in pathsInStore:
1418 # Assume local file system
1419 assert path is not None
1420 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1422 for transfer in ("copy", "link", "symlink", "relsymlink"):
1423 with safeTestTempDir(TESTDIR) as exportDir:
1424 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1425 export.saveDatasets(datasets)
1426 for path in pathsInStore:
1427 assert path is not None
1428 self.assertTrue(
1429 self.checkFileExists(exportDir, path),
1430 f"Check that mode {transfer} exported files",
1431 )
1433 def testPruneDatasets(self) -> None:
1434 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1435 butler = Butler(self.tmpConfigFile, writeable=True)
1436 assert isinstance(butler.datastore, FileDatastore)
1437 # Load registry data with dimensions to hang datasets off of.
1438 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1439 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1440 # Add some RUN-type collections.
1441 run1 = "run1"
1442 butler.registry.registerRun(run1)
1443 run2 = "run2"
1444 butler.registry.registerRun(run2)
1445 # put some datasets. ref1 and ref2 have the same data ID, and are in
1446 # different runs. ref3 has a different data ID.
1447 metric = makeExampleMetrics()
1448 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1449 datasetType = self.addDatasetType(
1450 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1451 )
1452 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1453 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1454 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1456 # Simple prune.
1457 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1458 with self.assertRaises(LookupError):
1459 butler.datasetExists(ref1.datasetType, ref1.dataId, collections=run1)
1461 # Put data back.
1462 ref1 = butler.put(metric, ref1, run=run1)
1463 ref2 = butler.put(metric, ref2, run=run2)
1464 ref3 = butler.put(metric, ref3, run=run1)
1466 # Check that in normal mode, deleting the record will lead to
1467 # trash not touching the file.
1468 uri1 = butler.datastore.getURI(ref1)
1469 butler.datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1470 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref1.id})
1471 butler.datastore.trash(ref1)
1472 butler.datastore.emptyTrash()
1473 self.assertTrue(uri1.exists())
1474 uri1.remove() # Clean it up.
1476 # Simulate execution butler setup by deleting the datastore
1477 # record but keeping the file around and trusting.
1478 butler.datastore.trustGetRequest = True
1479 uri2 = butler.datastore.getURI(ref2)
1480 uri3 = butler.datastore.getURI(ref3)
1481 self.assertTrue(uri2.exists())
1482 self.assertTrue(uri3.exists())
1484 # Remove the datastore record.
1485 butler.datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1486 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref2.id})
1487 self.assertTrue(uri2.exists())
1488 butler.datastore.trash([ref2, ref3])
1489 # Immediate removal for ref2 file
1490 self.assertFalse(uri2.exists())
1491 # But ref3 has to wait for the empty.
1492 self.assertTrue(uri3.exists())
1493 butler.datastore.emptyTrash()
1494 self.assertFalse(uri3.exists())
1496 # Clear out the datasets from registry.
1497 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1499 def testPytypeCoercion(self) -> None:
1500 """Test python type coercion on Butler.get and put."""
1502 # Store some data with the normal example storage class.
1503 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1504 datasetTypeName = "test_metric"
1505 butler = self.runPutGetTest(storageClass, datasetTypeName)
1507 dataId = {"instrument": "DummyCamComp", "visit": 423}
1508 metric = butler.get(datasetTypeName, dataId=dataId)
1509 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1511 datasetType_ori = butler.registry.getDatasetType(datasetTypeName)
1512 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1514 # Now need to hack the registry dataset type definition.
1515 # There is no API for this.
1516 assert isinstance(butler.registry, SqlRegistry)
1517 manager = butler.registry._managers.datasets
1518 assert hasattr(manager, "_db") and hasattr(manager, "_static")
1519 manager._db.update(
1520 manager._static.dataset_type,
1521 {"name": datasetTypeName},
1522 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1523 )
1525 # Force reset of dataset type cache
1526 butler.registry.refresh()
1528 datasetType_new = butler.registry.getDatasetType(datasetTypeName)
1529 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1530 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1532 metric_model = butler.get(datasetTypeName, dataId=dataId)
1533 self.assertNotEqual(type(metric_model), type(metric))
1534 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1536 # Put the model and read it back to show that everything now
1537 # works as normal.
1538 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1539 metric_model_new = butler.get(metric_ref)
1540 self.assertEqual(metric_model_new, metric_model)
1542 # Hack the storage class again to something that will fail on the
1543 # get with no conversion class.
1544 manager._db.update(
1545 manager._static.dataset_type,
1546 {"name": datasetTypeName},
1547 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1548 )
1549 butler.registry.refresh()
1551 with self.assertRaises(ValueError):
1552 butler.get(datasetTypeName, dataId=dataId)
1555@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1556class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1557 """PosixDatastore specialization of a butler using Postgres"""
1559 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1560 fullConfigKey = ".datastore.formatters"
1561 validationCanFail = True
1562 datastoreStr = ["/tmp"]
1563 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1564 registryStr = "PostgreSQL@test"
1565 postgresql: Any
1567 @staticmethod
1568 def _handler(postgresql: Any) -> None:
1569 engine = sqlalchemy.engine.create_engine(postgresql.url())
1570 with engine.begin() as connection:
1571 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1573 @classmethod
1574 def setUpClass(cls) -> None:
1575 # Create the postgres test server.
1576 cls.postgresql = testing.postgresql.PostgresqlFactory(
1577 cache_initialized_db=True, on_initialized=cls._handler
1578 )
1579 super().setUpClass()
1581 @classmethod
1582 def tearDownClass(cls) -> None:
1583 # Clean up any lingering SQLAlchemy engines/connections
1584 # so they're closed before we shut down the server.
1585 gc.collect()
1586 cls.postgresql.clear_cache()
1587 super().tearDownClass()
1589 def setUp(self) -> None:
1590 self.server = self.postgresql()
1592 # Need to add a registry section to the config.
1593 self._temp_config = False
1594 config = Config(self.configFile)
1595 config["registry", "db"] = self.server.url()
1596 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1597 config.dump(fh)
1598 self.configFile = fh.name
1599 self._temp_config = True
1600 super().setUp()
1602 def tearDown(self) -> None:
1603 self.server.stop()
1604 if self._temp_config and os.path.exists(self.configFile):
1605 os.remove(self.configFile)
1606 super().tearDown()
1608 def testMakeRepo(self) -> None:
1609 # The base class test assumes that it's using sqlite and assumes
1610 # the config file is acceptable to sqlite.
1611 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1614class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1615 """InMemoryDatastore specialization of a butler"""
1617 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1618 fullConfigKey = None
1619 useTempRoot = False
1620 validationCanFail = False
1621 datastoreStr = ["datastore='InMemory"]
1622 datastoreName = ["InMemoryDatastore@"]
1623 registryStr = "/gen3.sqlite3"
1625 def testIngest(self) -> None:
1626 pass
1629class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1630 """PosixDatastore specialization"""
1632 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1633 fullConfigKey = ".datastore.datastores.1.formatters"
1634 validationCanFail = True
1635 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1636 datastoreName = [
1637 "InMemoryDatastore@",
1638 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1639 "SecondDatastore",
1640 ]
1641 registryStr = "/gen3.sqlite3"
1644class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1645 """Test that a yaml file in one location can refer to a root in another."""
1647 datastoreStr = ["dir1"]
1648 # Disable the makeRepo test since we are deliberately not using
1649 # butler.yaml as the config name.
1650 fullConfigKey = None
1652 def setUp(self) -> None:
1653 self.root = makeTestTempDir(TESTDIR)
1655 # Make a new repository in one place
1656 self.dir1 = os.path.join(self.root, "dir1")
1657 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1659 # Move the yaml file to a different place and add a "root"
1660 self.dir2 = os.path.join(self.root, "dir2")
1661 os.makedirs(self.dir2, exist_ok=True)
1662 configFile1 = os.path.join(self.dir1, "butler.yaml")
1663 config = Config(configFile1)
1664 config["root"] = self.dir1
1665 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1666 config.dumpToUri(configFile2)
1667 os.remove(configFile1)
1668 self.tmpConfigFile = configFile2
1670 def testFileLocations(self) -> None:
1671 self.assertNotEqual(self.dir1, self.dir2)
1672 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1673 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1674 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1677class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1678 """Test that a config file created by makeRepo outside of repo works."""
1680 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1682 def setUp(self) -> None:
1683 self.root = makeTestTempDir(TESTDIR)
1684 self.root2 = makeTestTempDir(TESTDIR)
1686 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1687 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1689 def tearDown(self) -> None:
1690 if os.path.exists(self.root2):
1691 shutil.rmtree(self.root2, ignore_errors=True)
1692 super().tearDown()
1694 def testConfigExistence(self) -> None:
1695 c = Config(self.tmpConfigFile)
1696 uri_config = ResourcePath(c["root"])
1697 uri_expected = ResourcePath(self.root, forceDirectory=True)
1698 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1699 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1701 def testPutGet(self) -> None:
1702 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1703 self.runPutGetTest(storageClass, "test_metric")
1706class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase):
1707 """Test that a config file created by makeRepo outside of repo works."""
1709 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1711 def setUp(self) -> None:
1712 self.root = makeTestTempDir(TESTDIR)
1713 self.root2 = makeTestTempDir(TESTDIR)
1715 self.tmpConfigFile = self.root2
1716 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1718 def testConfigExistence(self) -> None:
1719 # Append the yaml file else Config constructor does not know the file
1720 # type.
1721 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1722 super().testConfigExistence()
1725class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1726 """Test that a config file created by makeRepo outside of repo works."""
1728 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1730 def setUp(self) -> None:
1731 self.root = makeTestTempDir(TESTDIR)
1732 self.root2 = makeTestTempDir(TESTDIR)
1734 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1735 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1738@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1739class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1740 """S3Datastore specialization of a butler; an S3 storage Datastore +
1741 a local in-memory SqlRegistry.
1742 """
1744 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1745 fullConfigKey = None
1746 validationCanFail = True
1748 bucketName = "anybucketname"
1749 """Name of the Bucket that will be used in the tests. The name is read from
1750 the config file used with the tests during set-up.
1751 """
1753 root = "butlerRoot/"
1754 """Root repository directory expected to be used in case useTempRoot=False.
1755 Otherwise the root is set to a 20 characters long randomly generated string
1756 during set-up.
1757 """
1759 datastoreStr = [f"datastore={root}"]
1760 """Contains all expected root locations in a format expected to be
1761 returned by Butler stringification.
1762 """
1764 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1765 """The expected format of the S3 Datastore string."""
1767 registryStr = "/gen3.sqlite3"
1768 """Expected format of the Registry string."""
1770 mock_s3 = mock_s3()
1771 """The mocked s3 interface from moto."""
1773 def genRoot(self) -> str:
1774 """Returns a random string of len 20 to serve as a root
1775 name for the temporary bucket repo.
1777 This is equivalent to tempfile.mkdtemp as this is what self.root
1778 becomes when useTempRoot is True.
1779 """
1780 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1781 return rndstr + "/"
1783 def setUp(self) -> None:
1784 config = Config(self.configFile)
1785 uri = ResourcePath(config[".datastore.datastore.root"])
1786 self.bucketName = uri.netloc
1788 # Enable S3 mocking of tests.
1789 self.mock_s3.start()
1791 # set up some fake credentials if they do not exist
1792 self.usingDummyCredentials = setAwsEnvCredentials()
1794 if self.useTempRoot:
1795 self.root = self.genRoot()
1796 rooturi = f"s3://{self.bucketName}/{self.root}"
1797 config.update({"datastore": {"datastore": {"root": rooturi}}})
1799 # need local folder to store registry database
1800 self.reg_dir = makeTestTempDir(TESTDIR)
1801 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
1803 # MOTO needs to know that we expect Bucket bucketname to exist
1804 # (this used to be the class attribute bucketName)
1805 s3 = boto3.resource("s3")
1806 s3.create_bucket(Bucket=self.bucketName)
1808 self.datastoreStr = [f"datastore='{rooturi}'"]
1809 self.datastoreName = [f"FileDatastore@{rooturi}"]
1810 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
1811 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
1813 def tearDown(self) -> None:
1814 s3 = boto3.resource("s3")
1815 bucket = s3.Bucket(self.bucketName)
1816 try:
1817 bucket.objects.all().delete()
1818 except botocore.exceptions.ClientError as e:
1819 if e.response["Error"]["Code"] == "404":
1820 # the key was not reachable - pass
1821 pass
1822 else:
1823 raise
1825 bucket = s3.Bucket(self.bucketName)
1826 bucket.delete()
1828 # Stop the S3 mock.
1829 self.mock_s3.stop()
1831 # unset any potentially set dummy credentials
1832 if self.usingDummyCredentials:
1833 unsetAwsEnvCredentials()
1835 if self.reg_dir is not None and os.path.exists(self.reg_dir):
1836 shutil.rmtree(self.reg_dir, ignore_errors=True)
1838 if self.useTempRoot and os.path.exists(self.root):
1839 shutil.rmtree(self.root, ignore_errors=True)
1841 super().tearDown()
1844class PosixDatastoreTransfers(unittest.TestCase):
1845 """Test data transfers between butlers.
1847 Test for different managers. UUID to UUID and integer to integer are
1848 tested. UUID to integer is not supported since we do not currently
1849 want to allow that. Integer to UUID is supported with the caveat
1850 that UUID4 will be generated and this will be incorrect for raw
1851 dataset types. The test ignores that.
1852 """
1854 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1855 storageClassFactory: StorageClassFactory
1857 @classmethod
1858 def setUpClass(cls) -> None:
1859 cls.storageClassFactory = StorageClassFactory()
1860 cls.storageClassFactory.addFromConfig(cls.configFile)
1862 def setUp(self) -> None:
1863 self.root = makeTestTempDir(TESTDIR)
1864 self.config = Config(self.configFile)
1866 def tearDown(self) -> None:
1867 removeTestTempDir(self.root)
1869 def create_butler(self, manager: str, label: str) -> Butler:
1870 config = Config(self.configFile)
1871 config["registry", "managers", "datasets"] = manager
1872 return Butler(Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True)
1874 def create_butlers(self, manager1: str | None = None, manager2: str | None = None) -> None:
1875 default = "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID"
1876 if manager1 is None:
1877 manager1 = default
1878 if manager2 is None:
1879 manager2 = default
1880 self.source_butler = self.create_butler(manager1, "1")
1881 self.target_butler = self.create_butler(manager2, "2")
1883 def testTransferUuidToUuid(self) -> None:
1884 self.create_butlers()
1885 self.assertButlerTransfers()
1887 def _enable_trust(self, datastore: Datastore) -> None:
1888 if hasattr(datastore, "trustGetRequest"):
1889 datastore.trustGetRequest = True
1890 elif hasattr(datastore, "datastores"):
1891 for datastore in datastore.datastores:
1892 if hasattr(datastore, "trustGetRequest"):
1893 datastore.trustGetRequest = True
1895 def testTransferMissing(self) -> None:
1896 """Test transfers where datastore records are missing.
1898 This is how execution butler works.
1899 """
1900 self.create_butlers()
1902 # Configure the source butler to allow trust.
1903 self._enable_trust(self.source_butler.datastore)
1905 self.assertButlerTransfers(purge=True)
1907 def testTransferMissingDisassembly(self) -> None:
1908 """Test transfers where datastore records are missing.
1910 This is how execution butler works.
1911 """
1912 self.create_butlers()
1914 # Configure the source butler to allow trust.
1915 self._enable_trust(self.source_butler.datastore)
1917 # Test disassembly.
1918 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
1920 def testAbsoluteURITransferDirect(self) -> None:
1921 """Test transfer using an absolute URI."""
1922 self._absolute_transfer("auto")
1924 def testAbsoluteURITransferCopy(self) -> None:
1925 """Test transfer using an absolute URI."""
1926 self._absolute_transfer("copy")
1928 def _absolute_transfer(self, transfer: str) -> None:
1929 self.create_butlers()
1931 storageClassName = "StructuredData"
1932 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
1933 datasetTypeName = "random_data"
1934 run = "run1"
1935 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
1937 dimensions = self.source_butler.registry.dimensions.extract(())
1938 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1939 self.source_butler.registry.registerDatasetType(datasetType)
1941 metrics = makeExampleMetrics()
1942 with ResourcePath.temporary_uri(suffix=".json") as temp:
1943 dataId = DataCoordinate.makeEmpty(self.source_butler.dimensions)
1944 source_refs = [DatasetRef(datasetType, dataId, run=run)]
1945 temp.write(json.dumps(metrics.exportAsDict()).encode())
1946 dataset = FileDataset(path=temp, refs=source_refs)
1947 self.source_butler.ingest(dataset, transfer="direct")
1949 self.target_butler.transfer_from(
1950 self.source_butler, dataset.refs, register_dataset_types=True, transfer=transfer
1951 )
1953 uri = self.target_butler.getURI(dataset.refs[0])
1954 if transfer == "auto":
1955 self.assertEqual(uri, temp)
1956 else:
1957 self.assertNotEqual(uri, temp)
1959 def assertButlerTransfers(self, purge: bool = False, storageClassName: str = "StructuredData") -> None:
1960 """Test that a run can be transferred to another butler."""
1962 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
1963 datasetTypeName = "random_data"
1965 # Test will create 3 collections and we will want to transfer
1966 # two of those three.
1967 runs = ["run1", "run2", "other"]
1969 # Also want to use two different dataset types to ensure that
1970 # grouping works.
1971 datasetTypeNames = ["random_data", "random_data_2"]
1973 # Create the run collections in the source butler.
1974 for run in runs:
1975 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
1977 # Create dimensions in source butler.
1978 n_exposures = 30
1979 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1980 self.source_butler.registry.insertDimensionData(
1981 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1982 )
1983 self.source_butler.registry.insertDimensionData(
1984 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1985 )
1987 for i in range(n_exposures):
1988 self.source_butler.registry.insertDimensionData(
1989 "exposure",
1990 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
1991 )
1993 # Create dataset types in the source butler.
1994 dimensions = self.source_butler.registry.dimensions.extract(["instrument", "exposure"])
1995 for datasetTypeName in datasetTypeNames:
1996 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1997 self.source_butler.registry.registerDatasetType(datasetType)
1999 # Write a dataset to an unrelated run -- this will ensure that
2000 # we are rewriting integer dataset ids in the target if necessary.
2001 # Will not be relevant for UUID.
2002 run = "distraction"
2003 butler = Butler(butler=self.source_butler, run=run)
2004 butler.put(
2005 makeExampleMetrics(),
2006 datasetTypeName,
2007 exposure=1,
2008 instrument="DummyCamComp",
2009 physical_filter="d-r",
2010 )
2012 # Write some example metrics to the source
2013 butler = Butler(butler=self.source_butler)
2015 # Set of DatasetRefs that should be in the list of refs to transfer
2016 # but which will not be transferred.
2017 deleted: set[DatasetRef] = set()
2019 n_expected = 20 # Number of datasets expected to be transferred
2020 source_refs = []
2021 for i in range(n_exposures):
2022 # Put a third of datasets into each collection, only retain
2023 # two thirds.
2024 index = i % 3
2025 run = runs[index]
2026 datasetTypeName = datasetTypeNames[i % 2]
2028 metric = MetricsExample(
2029 summary={"counter": i}, output={"text": "metric"}, data=[2 * x for x in range(i)]
2030 )
2031 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2032 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2034 # Remove the datastore record using low-level API
2035 if purge:
2036 # Remove records for a fraction.
2037 if index == 1:
2038 # For one of these delete the file as well.
2039 # This allows the "missing" code to filter the
2040 # file out.
2041 # Access the individual datastores.
2042 datastores = []
2043 if hasattr(butler.datastore, "datastores"):
2044 datastores.extend(butler.datastore.datastores)
2045 else:
2046 datastores.append(butler.datastore)
2048 if not deleted:
2049 # For a chained datastore we need to remove
2050 # files in each chain.
2051 for datastore in datastores:
2052 # The file might not be known to the datastore
2053 # if constraints are used.
2054 try:
2055 primary, uris = datastore.getURIs(ref)
2056 except FileNotFoundError:
2057 continue
2058 if primary:
2059 if primary.scheme != "mem":
2060 primary.remove()
2061 for uri in uris.values():
2062 if uri.scheme != "mem":
2063 uri.remove()
2064 n_expected -= 1
2065 deleted.add(ref)
2067 # Remove the datastore record.
2068 for datastore in datastores:
2069 if hasattr(datastore, "removeStoredItemInfo"):
2070 datastore.removeStoredItemInfo(ref)
2072 if index < 2:
2073 source_refs.append(ref)
2074 if ref not in deleted:
2075 new_metric = butler.get(ref)
2076 self.assertEqual(new_metric, metric)
2078 # Create some bad dataset types to ensure we check for inconsistent
2079 # definitions.
2080 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2081 for datasetTypeName in datasetTypeNames:
2082 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2083 self.target_butler.registry.registerDatasetType(datasetType)
2084 with self.assertRaises(ConflictingDefinitionError) as cm:
2085 self.target_butler.transfer_from(self.source_butler, source_refs)
2086 self.assertIn("dataset type differs", str(cm.exception))
2088 # And remove the bad definitions.
2089 for datasetTypeName in datasetTypeNames:
2090 self.target_butler.registry.removeDatasetType(datasetTypeName)
2092 # Transfer without creating dataset types should fail.
2093 with self.assertRaises(KeyError):
2094 self.target_butler.transfer_from(self.source_butler, source_refs)
2096 # Transfer without creating dimensions should fail.
2097 with self.assertRaises(ConflictingDefinitionError) as cm:
2098 self.target_butler.transfer_from(self.source_butler, source_refs, register_dataset_types=True)
2099 self.assertIn("dimension", str(cm.exception))
2101 # The failed transfer above leaves registry in an inconsistent
2102 # state because the run is created but then rolled back without
2103 # the collection cache being cleared. For now force a refresh.
2104 # Can remove with DM-35498.
2105 self.target_butler.registry.refresh()
2107 # Now transfer them to the second butler, including dimensions.
2108 with self.assertLogs(level=logging.DEBUG) as log_cm:
2109 transferred = self.target_butler.transfer_from(
2110 self.source_butler,
2111 source_refs,
2112 register_dataset_types=True,
2113 transfer_dimensions=True,
2114 )
2115 self.assertEqual(len(transferred), n_expected)
2116 log_output = ";".join(log_cm.output)
2118 # A ChainedDatastore will use the in-memory datastore for mexists
2119 # so we can not rely on the mexists log message.
2120 self.assertIn("Number of datastore records found in source", log_output)
2121 self.assertIn("Creating output run", log_output)
2123 # Do the transfer twice to ensure that it will do nothing extra.
2124 # Only do this if purge=True because it does not work for int
2125 # dataset_id.
2126 if purge:
2127 # This should not need to register dataset types.
2128 transferred = self.target_butler.transfer_from(self.source_butler, source_refs)
2129 self.assertEqual(len(transferred), n_expected)
2131 # Also do an explicit low-level transfer to trigger some
2132 # edge cases.
2133 with self.assertLogs(level=logging.DEBUG) as log_cm:
2134 self.target_butler.datastore.transfer_from(self.source_butler.datastore, source_refs)
2135 log_output = ";".join(log_cm.output)
2136 self.assertIn("no file artifacts exist", log_output)
2138 with self.assertRaises((TypeError, AttributeError)):
2139 self.target_butler.datastore.transfer_from(self.source_butler, source_refs) # type: ignore
2141 with self.assertRaises(ValueError):
2142 self.target_butler.datastore.transfer_from(
2143 self.source_butler.datastore, source_refs, transfer="split"
2144 )
2146 # Now try to get the same refs from the new butler.
2147 for ref in source_refs:
2148 if ref not in deleted:
2149 new_metric = self.target_butler.get(ref)
2150 old_metric = self.source_butler.get(ref)
2151 self.assertEqual(new_metric, old_metric)
2153 # Now prune run2 collection and create instead a CHAINED collection.
2154 # This should block the transfer.
2155 self.target_butler.removeRuns(["run2"], unstore=True)
2156 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2157 with self.assertRaises(CollectionTypeError):
2158 # Re-importing the run1 datasets can be problematic if they
2159 # use integer IDs so filter those out.
2160 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2161 self.target_butler.transfer_from(self.source_butler, to_transfer)
2164class ChainedDatastoreTransfers(PosixDatastoreTransfers):
2165 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
2168if __name__ == "__main__":
2169 unittest.main()