Coverage for tests/test_butler.py: 12%
1232 statements
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-08 05:05 -0700
« prev ^ index » next coverage.py v7.2.7, created at 2023-06-08 05:05 -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 DatasetExistence,
72 DatasetRef,
73 DatasetType,
74 FileDataset,
75 FileTemplate,
76 FileTemplateValidationError,
77 StorageClassFactory,
78 ValidationError,
79 script,
80)
81from lsst.daf.butler.core.repoRelocation import BUTLER_ROOT_TAG
82from lsst.daf.butler.datastores.fileDatastore import FileDatastore
83from lsst.daf.butler.registries.sql import SqlRegistry
84from lsst.daf.butler.registry import (
85 CollectionError,
86 CollectionTypeError,
87 ConflictingDefinitionError,
88 DataIdValueError,
89 MissingCollectionError,
90 OrphanedRecordError,
91)
92from lsst.daf.butler.tests import MetricsExample, MultiDetectorFormatter
93from lsst.daf.butler.tests.utils import TestCaseMixin, makeTestTempDir, removeTestTempDir, safeTestTempDir
94from lsst.resources import ResourcePath
95from lsst.resources.s3utils import setAwsEnvCredentials, unsetAwsEnvCredentials
96from lsst.utils import doImportType
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 kwargs: dict[str, Any] = {}
262 if not isinstance(args[0], DatasetRef): # type: ignore
263 kwargs["run"] = this_run
264 ref = butler.put(metric, *args, **kwargs)
265 self.assertIsInstance(ref, DatasetRef)
267 # Test getDirect
268 metricOut = butler.get(ref)
269 self.assertEqual(metric, metricOut)
270 # Test get
271 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
272 self.assertEqual(metric, metricOut)
273 # Test get with a datasetRef
274 metricOut = butler.get(ref)
275 self.assertEqual(metric, metricOut)
276 # Test getDeferred with dataId
277 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
278 self.assertEqual(metric, metricOut)
279 # Test getDeferred with a 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 self.assertFalse(butler.exists(*args, collections=this_run))
347 # get() should still fail.
348 with self.assertRaises(FileNotFoundError):
349 butler.get(ref)
350 # Registry shouldn't be able to find it by dataset_id anymore.
351 self.assertIsNone(butler.registry.getDataset(ref.id))
353 # Do explicit registry removal since we know they are
354 # empty
355 butler.registry.removeCollection(this_run)
356 expected_collections.remove(this_run)
358 # Create DatasetRef for put using default run.
359 refIn = DatasetRef(datasetType, dataId, id=uuid.UUID(int=1), run=butler.run)
361 # Put the dataset again, since the last thing we did was remove it
362 # and we want to use the default collection.
363 ref = butler.put(metric, refIn)
365 # Get with parameters
366 stop = 4
367 sliced = butler.get(ref, parameters={"slice": slice(stop)})
368 self.assertNotEqual(metric, sliced)
369 self.assertEqual(metric.summary, sliced.summary)
370 self.assertEqual(metric.output, sliced.output)
371 self.assertEqual(metric.data[:stop], sliced.data)
372 # getDeferred with parameters
373 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
374 self.assertNotEqual(metric, sliced)
375 self.assertEqual(metric.summary, sliced.summary)
376 self.assertEqual(metric.output, sliced.output)
377 self.assertEqual(metric.data[:stop], sliced.data)
378 # getDeferred with deferred parameters
379 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
380 self.assertNotEqual(metric, sliced)
381 self.assertEqual(metric.summary, sliced.summary)
382 self.assertEqual(metric.output, sliced.output)
383 self.assertEqual(metric.data[:stop], sliced.data)
385 if storageClass.isComposite():
386 # Check that components can be retrieved
387 metricOut = butler.get(ref.datasetType.name, dataId)
388 compNameS = ref.datasetType.componentTypeName("summary")
389 compNameD = ref.datasetType.componentTypeName("data")
390 summary = butler.get(compNameS, dataId)
391 self.assertEqual(summary, metric.summary)
392 data = butler.get(compNameD, dataId)
393 self.assertEqual(data, metric.data)
395 if "counter" in storageClass.derivedComponents:
396 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
397 self.assertEqual(count, len(data))
399 count = butler.get(
400 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
401 )
402 self.assertEqual(count, stop)
404 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections)
405 assert compRef is not None
406 summary = butler.get(compRef)
407 self.assertEqual(summary, metric.summary)
409 # Create a Dataset type that has the same name but is inconsistent.
410 inconsistentDatasetType = DatasetType(
411 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
412 )
414 # Getting with a dataset type that does not match registry fails
415 with self.assertRaisesRegex(ValueError, "Supplied dataset type .* inconsistent with registry"):
416 butler.get(inconsistentDatasetType, dataId)
418 # Combining a DatasetRef with a dataId should fail
419 with self.assertRaisesRegex(ValueError, "DatasetRef given, cannot use dataId as well"):
420 butler.get(ref, dataId)
421 # Getting with an explicit ref should fail if the id doesn't match.
422 with self.assertRaises(FileNotFoundError):
423 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=uuid.UUID(int=101), run=butler.run))
425 # Getting a dataset with unknown parameters should fail
426 with self.assertRaisesRegex(KeyError, "Parameter 'unsupported' not understood"):
427 butler.get(ref, parameters={"unsupported": True})
429 # Check we have a collection
430 collections = set(butler.registry.queryCollections())
431 self.assertEqual(collections, expected_collections)
433 # Clean up to check that we can remove something that may have
434 # already had a component removed
435 butler.pruneDatasets([ref], unstore=True, purge=True)
437 # Add the same ref again, so we can check that duplicate put fails.
438 ref = butler.put(metric, datasetType, dataId)
440 # Repeat put will fail.
441 with self.assertRaisesRegex(
442 ConflictingDefinitionError, "A database constraint failure was triggered"
443 ):
444 butler.put(metric, datasetType, dataId)
446 # Remove the datastore entry.
447 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
449 # Put will still fail
450 with self.assertRaisesRegex(
451 ConflictingDefinitionError, "A database constraint failure was triggered"
452 ):
453 butler.put(metric, datasetType, dataId)
455 # Repeat the same sequence with resolved ref.
456 butler.pruneDatasets([ref], unstore=True, purge=True)
457 ref = butler.put(metric, refIn)
459 # Repeat put will fail.
460 with self.assertRaisesRegex(ConflictingDefinitionError, "Datastore already contains dataset"):
461 butler.put(metric, refIn)
463 # Remove the datastore entry.
464 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
466 # In case of resolved ref this write will succeed.
467 ref = butler.put(metric, refIn)
469 # Leave the dataset in place since some downstream tests require
470 # something to be present
472 return butler
474 def testDeferredCollectionPassing(self) -> None:
475 # Construct a butler with no run or collection, but make it writeable.
476 butler = Butler(self.tmpConfigFile, writeable=True)
477 # Create and register a DatasetType
478 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
479 datasetType = self.addDatasetType(
480 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
481 )
482 # Add needed Dimensions
483 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
484 butler.registry.insertDimensionData(
485 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
486 )
487 butler.registry.insertDimensionData(
488 "visit",
489 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
490 )
491 dataId = {"instrument": "DummyCamComp", "visit": 423}
492 # Create dataset.
493 metric = makeExampleMetrics()
494 # Register a new run and put dataset.
495 run = "deferred"
496 self.assertTrue(butler.registry.registerRun(run))
497 # Second time it will be allowed but indicate no-op
498 self.assertFalse(butler.registry.registerRun(run))
499 ref = butler.put(metric, datasetType, dataId, run=run)
500 # Putting with no run should fail with TypeError.
501 with self.assertRaises(CollectionError):
502 butler.put(metric, datasetType, dataId)
503 # Dataset should exist.
504 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
505 # We should be able to get the dataset back, but with and without
506 # a deferred dataset handle.
507 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
508 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
509 # Trying to find the dataset without any collection is a TypeError.
510 self.assertFalse(butler.exists(datasetType, dataId))
511 with self.assertRaises(CollectionError):
512 butler.get(datasetType, dataId)
513 # Associate the dataset with a different collection.
514 butler.registry.registerCollection("tagged")
515 butler.registry.associate("tagged", [ref])
516 # Deleting the dataset from the new collection should make it findable
517 # in the original collection.
518 butler.pruneDatasets([ref], tags=["tagged"])
519 self.assertTrue(butler.exists(datasetType, dataId, collections=[run]))
522class ButlerTests(ButlerPutGetTests):
523 """Tests for Butler."""
525 useTempRoot = True
526 validationCanFail: bool
527 fullConfigKey: str | None
528 registryStr: str | None
529 datastoreName: list[str] | None
530 datastoreStr: list[str]
532 def setUp(self) -> None:
533 """Create a new butler root for each test."""
534 self.root = makeTestTempDir(TESTDIR)
535 Butler.makeRepo(self.root, config=Config(self.configFile))
536 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
538 def testConstructor(self) -> None:
539 """Independent test of constructor."""
540 butler = Butler(self.tmpConfigFile, run=self.default_run)
541 self.assertIsInstance(butler, Butler)
543 # Check that butler.yaml is added automatically.
544 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
545 config_dir = self.tmpConfigFile[: -len(end)]
546 butler = Butler(config_dir, run=self.default_run)
547 self.assertIsInstance(butler, Butler)
549 # Even with a ResourcePath.
550 butler = Butler(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
551 self.assertIsInstance(butler, Butler)
553 collections = set(butler.registry.queryCollections())
554 self.assertEqual(collections, {self.default_run})
556 # Check that some special characters can be included in run name.
557 special_run = "u@b.c-A"
558 butler_special = Butler(butler=butler, run=special_run)
559 collections = set(butler_special.registry.queryCollections("*@*"))
560 self.assertEqual(collections, {special_run})
562 butler2 = Butler(butler=butler, collections=["other"])
563 self.assertEqual(butler2.collections, ("other",))
564 self.assertIsNone(butler2.run)
565 self.assertIs(butler.datastore, butler2.datastore)
567 # Test that we can use an environment variable to find this
568 # repository.
569 butler_index = Config()
570 butler_index["label"] = self.tmpConfigFile
571 for suffix in (".yaml", ".json"):
572 # Ensure that the content differs so that we know that
573 # we aren't reusing the cache.
574 bad_label = f"s3://bucket/not_real{suffix}"
575 butler_index["bad_label"] = bad_label
576 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
577 butler_index.dumpToUri(temp_file)
578 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
579 self.assertEqual(Butler.get_known_repos(), set(("label", "bad_label")))
580 uri = Butler.get_repo_uri("bad_label")
581 self.assertEqual(uri, ResourcePath(bad_label))
582 uri = Butler.get_repo_uri("label")
583 butler = Butler(uri, writeable=False)
584 self.assertIsInstance(butler, Butler)
585 butler = Butler("label", writeable=False)
586 self.assertIsInstance(butler, Butler)
587 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
588 Butler("not_there", writeable=False)
589 with self.assertRaises(KeyError) as cm:
590 Butler.get_repo_uri("missing")
591 self.assertEqual(Butler.get_repo_uri("missing", True), ResourcePath("missing"))
592 self.assertIn("not known to", str(cm.exception))
593 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
594 with self.assertRaises(FileNotFoundError):
595 Butler.get_repo_uri("label")
596 self.assertEqual(Butler.get_known_repos(), set())
597 with self.assertRaises(KeyError) as cm:
598 # No environment variable set.
599 Butler.get_repo_uri("label")
600 self.assertEqual(Butler.get_repo_uri("label", True), ResourcePath("label"))
601 self.assertIn("No repository index defined", str(cm.exception))
602 with self.assertRaisesRegex(FileNotFoundError, "no known aliases"):
603 # No aliases registered.
604 Butler("not_there")
605 self.assertEqual(Butler.get_known_repos(), set())
607 def testBasicPutGet(self) -> None:
608 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
609 self.runPutGetTest(storageClass, "test_metric")
611 def testCompositePutGetConcrete(self) -> None:
612 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
613 butler = self.runPutGetTest(storageClass, "test_metric")
615 # Should *not* be disassembled
616 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
617 self.assertEqual(len(datasets), 1)
618 uri, components = butler.getURIs(datasets[0])
619 self.assertIsInstance(uri, ResourcePath)
620 self.assertFalse(components)
621 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
622 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
624 # Predicted dataset
625 dataId = {"instrument": "DummyCamComp", "visit": 424}
626 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
627 self.assertFalse(components)
628 self.assertIsInstance(uri, ResourcePath)
629 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
630 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
632 def testCompositePutGetVirtual(self) -> None:
633 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
634 butler = self.runPutGetTest(storageClass, "test_metric_comp")
636 # Should be disassembled
637 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
638 self.assertEqual(len(datasets), 1)
639 uri, components = butler.getURIs(datasets[0])
641 if butler.datastore.isEphemeral:
642 # Never disassemble in-memory datastore
643 self.assertIsInstance(uri, ResourcePath)
644 self.assertFalse(components)
645 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
646 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
647 else:
648 self.assertIsNone(uri)
649 self.assertEqual(set(components), set(storageClass.components))
650 for compuri in components.values():
651 self.assertIsInstance(compuri, ResourcePath)
652 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
653 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
655 # Predicted dataset
656 dataId = {"instrument": "DummyCamComp", "visit": 424}
657 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
659 if butler.datastore.isEphemeral:
660 # Never disassembled
661 self.assertIsInstance(uri, ResourcePath)
662 self.assertFalse(components)
663 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
664 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
665 else:
666 self.assertIsNone(uri)
667 self.assertEqual(set(components), set(storageClass.components))
668 for compuri in components.values():
669 self.assertIsInstance(compuri, ResourcePath)
670 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
671 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
673 def testStorageClassOverrideGet(self) -> None:
674 """Test storage class conversion on get with override."""
675 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
676 datasetTypeName = "anything"
677 run = self.default_run
679 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
681 # Create and store a dataset.
682 metric = makeExampleMetrics()
683 dataId = {"instrument": "DummyCamComp", "visit": 423}
685 ref = butler.put(metric, datasetType, dataId)
687 # Return native type.
688 retrieved = butler.get(ref)
689 self.assertEqual(retrieved, metric)
691 # Specify an override.
692 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
693 model = butler.get(ref, storageClass=new_sc)
694 self.assertNotEqual(type(model), type(retrieved))
695 self.assertIs(type(model), new_sc.pytype)
696 self.assertEqual(retrieved, model)
698 # Defer but override later.
699 deferred = butler.getDeferred(ref)
700 model = deferred.get(storageClass=new_sc)
701 self.assertIs(type(model), new_sc.pytype)
702 self.assertEqual(retrieved, model)
704 # Defer but override up front.
705 deferred = butler.getDeferred(ref, storageClass=new_sc)
706 model = deferred.get()
707 self.assertIs(type(model), new_sc.pytype)
708 self.assertEqual(retrieved, model)
710 # Retrieve a component. Should be a tuple.
711 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
712 self.assertIs(type(data), tuple)
713 self.assertEqual(data, tuple(retrieved.data))
715 # Parameter on the write storage class should work regardless
716 # of read storage class.
717 data = butler.get(
718 "anything.data",
719 dataId,
720 storageClass="StructuredDataDataTestTuple",
721 parameters={"slice": slice(2, 4)},
722 )
723 self.assertEqual(len(data), 2)
725 # Try a parameter that is known to the read storage class but not
726 # the write storage class.
727 with self.assertRaises(KeyError):
728 butler.get(
729 "anything.data",
730 dataId,
731 storageClass="StructuredDataDataTestTuple",
732 parameters={"xslice": slice(2, 4)},
733 )
735 def testPytypePutCoercion(self) -> None:
736 """Test python type coercion on Butler.get and put."""
738 # Store some data with the normal example storage class.
739 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
740 datasetTypeName = "test_metric"
741 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
743 dataId = {"instrument": "DummyCamComp", "visit": 423}
745 # Put a dict and this should coerce to a MetricsExample
746 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
747 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
748 test_metric = butler.get(metric_ref)
749 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
750 self.assertEqual(test_metric.summary, test_dict["summary"])
751 self.assertEqual(test_metric.output, test_dict["output"])
753 # Check that the put still works if a DatasetType is given with
754 # a definition matching this python type.
755 registry_type = butler.registry.getDatasetType(datasetTypeName)
756 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
757 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
758 self.assertEqual(metric2_ref.datasetType, registry_type)
760 # The get will return the type expected by registry.
761 test_metric2 = butler.get(metric2_ref)
762 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
764 # Make a new DatasetRef with the compatible but different DatasetType.
765 # This should now return a dict.
766 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
767 test_dict2 = butler.get(new_ref)
768 self.assertEqual(get_full_type_name(test_dict2), "dict")
770 # Get it again with the wrong dataset type definition using get()
771 # rather than get(). This should be consistent with get()
772 # behavior and return the type of the DatasetType.
773 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
774 self.assertEqual(get_full_type_name(test_dict3), "dict")
776 def testIngest(self) -> None:
777 butler = Butler(self.tmpConfigFile, run=self.default_run)
779 # Create and register a DatasetType
780 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "detector"])
782 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
783 datasetTypeName = "metric"
785 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
787 # Add needed Dimensions
788 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
789 butler.registry.insertDimensionData(
790 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
791 )
792 for detector in (1, 2):
793 butler.registry.insertDimensionData(
794 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
795 )
797 butler.registry.insertDimensionData(
798 "visit",
799 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
800 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
801 )
803 formatter = doImportType("lsst.daf.butler.formatters.yaml.YamlFormatter")
804 dataRoot = os.path.join(TESTDIR, "data", "basic")
805 datasets = []
806 for detector in (1, 2):
807 detector_name = f"detector_{detector}"
808 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
809 dataId = butler.registry.expandDataId(
810 {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
811 )
812 # Create a DatasetRef for ingest
813 refIn = DatasetRef(datasetType, dataId, run=self.default_run)
815 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
817 butler.ingest(*datasets, transfer="copy")
819 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
820 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
822 metrics1 = butler.get(datasetTypeName, dataId1)
823 metrics2 = butler.get(datasetTypeName, dataId2)
824 self.assertNotEqual(metrics1, metrics2)
826 # Compare URIs
827 uri1 = butler.getURI(datasetTypeName, dataId1)
828 uri2 = butler.getURI(datasetTypeName, dataId2)
829 self.assertNotEqual(uri1, uri2)
831 # Now do a multi-dataset but single file ingest
832 metricFile = os.path.join(dataRoot, "detectors.yaml")
833 refs = []
834 for detector in (1, 2):
835 detector_name = f"detector_{detector}"
836 dataId = butler.registry.expandDataId(
837 {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
838 )
839 # Create a DatasetRef for ingest
840 refs.append(DatasetRef(datasetType, dataId, run=self.default_run))
842 # Test "move" transfer to ensure that the files themselves
843 # have disappeared following ingest.
844 with ResourcePath.temporary_uri(suffix=".yaml") as tempFile:
845 tempFile.transfer_from(ResourcePath(metricFile), transfer="copy")
847 datasets = []
848 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter))
850 # For first ingest use copy.
851 butler.ingest(*datasets, transfer="copy", record_validation_info=False)
853 # Now try to ingest again in "execution butler" mode where
854 # the registry entries exist but the datastore does not have
855 # the files. We also need to strip the dimension records to ensure
856 # that they will be re-added by the ingest.
857 ref = datasets[0].refs[0]
858 datasets[0].refs = [
859 cast(
860 DatasetRef,
861 butler.registry.findDataset(ref.datasetType, dataId=ref.dataId, collections=ref.run),
862 )
863 for ref in datasets[0].refs
864 ]
865 all_refs = []
866 for dataset in datasets:
867 refs = []
868 for ref in dataset.refs:
869 # Create a dict from the dataId to drop the records.
870 new_data_id = {str(k): v for k, v in ref.dataId.items()}
871 new_ref = butler.registry.findDataset(ref.datasetType, new_data_id, collections=ref.run)
872 assert new_ref is not None
873 self.assertFalse(new_ref.dataId.hasRecords())
874 refs.append(new_ref)
875 dataset.refs = refs
876 all_refs.extend(dataset.refs)
877 butler.pruneDatasets(all_refs, disassociate=False, unstore=True, purge=False)
879 # Use move mode to test that the file is deleted. Also
880 # disable recording of file size.
881 butler.ingest(*datasets, transfer="move", record_validation_info=False)
883 # Check that every ref now has records.
884 for dataset in datasets:
885 for ref in dataset.refs:
886 self.assertTrue(ref.dataId.hasRecords())
888 # Ensure that the file has disappeared.
889 self.assertFalse(tempFile.exists())
891 # Check that the datastore recorded no file size.
892 # Not all datastores can support this.
893 try:
894 infos = butler.datastore.getStoredItemsInfo(datasets[0].refs[0]) # type: ignore[attr-defined]
895 self.assertEqual(infos[0].file_size, -1)
896 except AttributeError:
897 pass
899 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
900 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
902 multi1 = butler.get(datasetTypeName, dataId1)
903 multi2 = butler.get(datasetTypeName, dataId2)
905 self.assertEqual(multi1, metrics1)
906 self.assertEqual(multi2, metrics2)
908 # Compare URIs
909 uri1 = butler.getURI(datasetTypeName, dataId1)
910 uri2 = butler.getURI(datasetTypeName, dataId2)
911 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
913 # Test that removing one does not break the second
914 # This line will issue a warning log message for a ChainedDatastore
915 # that uses an InMemoryDatastore since in-memory can not ingest
916 # files.
917 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
918 self.assertFalse(butler.exists(datasetTypeName, dataId1))
919 self.assertTrue(butler.exists(datasetTypeName, dataId2))
920 multi2b = butler.get(datasetTypeName, dataId2)
921 self.assertEqual(multi2, multi2b)
923 # Ensure we can ingest 0 datasets
924 datasets = []
925 butler.ingest(*datasets)
927 def testPickle(self) -> None:
928 """Test pickle support."""
929 butler = Butler(self.tmpConfigFile, run=self.default_run)
930 butlerOut = pickle.loads(pickle.dumps(butler))
931 self.assertIsInstance(butlerOut, Butler)
932 self.assertEqual(butlerOut._config, butler._config)
933 self.assertEqual(butlerOut.collections, butler.collections)
934 self.assertEqual(butlerOut.run, butler.run)
936 def testGetDatasetTypes(self) -> None:
937 butler = Butler(self.tmpConfigFile, run=self.default_run)
938 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "physical_filter"])
939 dimensionEntries: list[tuple[str, list[Mapping[str, Any]]]] = [
940 (
941 "instrument",
942 [
943 {"instrument": "DummyCam"},
944 {"instrument": "DummyHSC"},
945 {"instrument": "DummyCamComp"},
946 ],
947 ),
948 ("physical_filter", [{"instrument": "DummyCam", "name": "d-r", "band": "R"}]),
949 ("visit", [{"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}]),
950 ]
951 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
952 # Add needed Dimensions
953 for element, data in dimensionEntries:
954 butler.registry.insertDimensionData(element, *data)
956 # When a DatasetType is added to the registry entries are not created
957 # for components but querying them can return the components.
958 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
959 components = set()
960 for datasetTypeName in datasetTypeNames:
961 # Create and register a DatasetType
962 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
964 for componentName in storageClass.components:
965 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
967 fromRegistry: set[DatasetType] = set()
968 for parent_dataset_type in butler.registry.queryDatasetTypes():
969 fromRegistry.add(parent_dataset_type)
970 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
971 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
973 # Now that we have some dataset types registered, validate them
974 butler.validateConfiguration(
975 ignore=[
976 "test_metric_comp",
977 "metric3",
978 "metric5",
979 "calexp",
980 "DummySC",
981 "datasetType.component",
982 "random_data",
983 "random_data_2",
984 ]
985 )
987 # Add a new datasetType that will fail template validation
988 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
989 if self.validationCanFail:
990 with self.assertRaises(ValidationError):
991 butler.validateConfiguration()
993 # Rerun validation but with a subset of dataset type names
994 butler.validateConfiguration(datasetTypeNames=["metric4"])
996 # Rerun validation but ignore the bad datasetType
997 butler.validateConfiguration(
998 ignore=[
999 "test_metric_comp",
1000 "metric3",
1001 "metric5",
1002 "calexp",
1003 "DummySC",
1004 "datasetType.component",
1005 "random_data",
1006 "random_data_2",
1007 ]
1008 )
1010 def testTransaction(self) -> None:
1011 butler = Butler(self.tmpConfigFile, run=self.default_run)
1012 datasetTypeName = "test_metric"
1013 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1014 dimensionEntries: tuple[tuple[str, Mapping[str, Any]], ...] = (
1015 ("instrument", {"instrument": "DummyCam"}),
1016 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1017 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1018 )
1019 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1020 metric = makeExampleMetrics()
1021 dataId = {"instrument": "DummyCam", "visit": 42}
1022 # Create and register a DatasetType
1023 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1024 with self.assertRaises(TransactionTestError):
1025 with butler.transaction():
1026 # Add needed Dimensions
1027 for args in dimensionEntries:
1028 butler.registry.insertDimensionData(*args)
1029 # Store a dataset
1030 ref = butler.put(metric, datasetTypeName, dataId)
1031 self.assertIsInstance(ref, DatasetRef)
1032 # Test getDirect
1033 metricOut = butler.get(ref)
1034 self.assertEqual(metric, metricOut)
1035 # Test get
1036 metricOut = butler.get(datasetTypeName, dataId)
1037 self.assertEqual(metric, metricOut)
1038 # Check we can get components
1039 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1040 raise TransactionTestError("This should roll back the entire transaction")
1041 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1042 butler.registry.expandDataId(dataId)
1043 # Should raise LookupError for missing data ID value
1044 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1045 butler.get(datasetTypeName, dataId)
1046 # Also check explicitly if Dataset entry is missing
1047 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections))
1048 # Direct retrieval should not find the file in the Datastore
1049 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1050 butler.get(ref)
1052 def testMakeRepo(self) -> None:
1053 """Test that we can write butler configuration to a new repository via
1054 the Butler.makeRepo interface and then instantiate a butler from the
1055 repo root.
1056 """
1057 # Do not run the test if we know this datastore configuration does
1058 # not support a file system root
1059 if self.fullConfigKey is None:
1060 return
1062 # create two separate directories
1063 root1 = tempfile.mkdtemp(dir=self.root)
1064 root2 = tempfile.mkdtemp(dir=self.root)
1066 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1067 limited = Config(self.configFile)
1068 butler1 = Butler(butlerConfig)
1069 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1070 full = Config(self.tmpConfigFile)
1071 butler2 = Butler(butlerConfig)
1072 # Butlers should have the same configuration regardless of whether
1073 # defaults were expanded.
1074 self.assertEqual(butler1._config, butler2._config)
1075 # Config files loaded directly should not be the same.
1076 self.assertNotEqual(limited, full)
1077 # Make sure "limited" doesn't have a few keys we know it should be
1078 # inheriting from defaults.
1079 self.assertIn(self.fullConfigKey, full)
1080 self.assertNotIn(self.fullConfigKey, limited)
1082 # Collections don't appear until something is put in them
1083 collections1 = set(butler1.registry.queryCollections())
1084 self.assertEqual(collections1, set())
1085 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1087 # Check that a config with no associated file name will not
1088 # work properly with relocatable Butler repo
1089 butlerConfig.configFile = None
1090 with self.assertRaises(ValueError):
1091 Butler(butlerConfig)
1093 with self.assertRaises(FileExistsError):
1094 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1096 def testStringification(self) -> None:
1097 butler = Butler(self.tmpConfigFile, run=self.default_run)
1098 butlerStr = str(butler)
1100 if self.datastoreStr is not None:
1101 for testStr in self.datastoreStr:
1102 self.assertIn(testStr, butlerStr)
1103 if self.registryStr is not None:
1104 self.assertIn(self.registryStr, butlerStr)
1106 datastoreName = butler.datastore.name
1107 if self.datastoreName is not None:
1108 for testStr in self.datastoreName:
1109 self.assertIn(testStr, datastoreName)
1111 def testButlerRewriteDataId(self) -> None:
1112 """Test that dataIds can be rewritten based on dimension records."""
1114 butler = Butler(self.tmpConfigFile, run=self.default_run)
1116 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1117 datasetTypeName = "random_data"
1119 # Create dimension records.
1120 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1121 butler.registry.insertDimensionData(
1122 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1123 )
1124 butler.registry.insertDimensionData(
1125 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1126 )
1128 dimensions = butler.registry.dimensions.extract(["instrument", "exposure"])
1129 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1130 butler.registry.registerDatasetType(datasetType)
1132 n_exposures = 5
1133 dayobs = 20210530
1135 for i in range(n_exposures):
1136 butler.registry.insertDimensionData(
1137 "exposure",
1138 {
1139 "instrument": "DummyCamComp",
1140 "id": i,
1141 "obs_id": f"exp{i}",
1142 "seq_num": i,
1143 "day_obs": dayobs,
1144 "physical_filter": "d-r",
1145 },
1146 )
1148 # Write some data.
1149 for i in range(n_exposures):
1150 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1152 # Use the seq_num for the put to test rewriting.
1153 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1154 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1156 # Check that the exposure is correct in the dataId
1157 self.assertEqual(ref.dataId["exposure"], i)
1159 # and check that we can get the dataset back with the same dataId
1160 new_metric = butler.get(datasetTypeName, dataId=dataId)
1161 self.assertEqual(new_metric, metric)
1164class FileDatastoreButlerTests(ButlerTests):
1165 """Common tests and specialization of ButlerTests for butlers backed
1166 by datastores that inherit from FileDatastore.
1167 """
1169 def checkFileExists(self, root: str | ResourcePath, relpath: str | ResourcePath) -> bool:
1170 """Checks if file exists at a given path (relative to root).
1172 Test testPutTemplates verifies actual physical existance of the files
1173 in the requested location.
1174 """
1175 uri = ResourcePath(root, forceDirectory=True)
1176 return uri.join(relpath).exists()
1178 def testPutTemplates(self) -> None:
1179 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1180 butler = Butler(self.tmpConfigFile, run=self.default_run)
1182 # Add needed Dimensions
1183 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1184 butler.registry.insertDimensionData(
1185 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1186 )
1187 butler.registry.insertDimensionData(
1188 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1189 )
1190 butler.registry.insertDimensionData(
1191 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1192 )
1194 # Create and store a dataset
1195 metric = makeExampleMetrics()
1197 # Create two almost-identical DatasetTypes (both will use default
1198 # template)
1199 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1200 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1201 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1202 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1204 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1205 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1207 # Put with exactly the data ID keys needed
1208 ref = butler.put(metric, "metric1", dataId1)
1209 uri = butler.getURI(ref)
1210 self.assertTrue(uri.exists())
1211 self.assertTrue(
1212 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle")
1213 )
1215 # Check the template based on dimensions
1216 if hasattr(butler.datastore, "templates"):
1217 butler.datastore.templates.validateTemplates([ref])
1219 # Put with extra data ID keys (physical_filter is an optional
1220 # dependency); should not change template (at least the way we're
1221 # defining them to behave now; the important thing is that they
1222 # must be consistent).
1223 ref = butler.put(metric, "metric2", dataId2)
1224 uri = butler.getURI(ref)
1225 self.assertTrue(uri.exists())
1226 self.assertTrue(
1227 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle")
1228 )
1230 # Check the template based on dimensions
1231 if hasattr(butler.datastore, "templates"):
1232 butler.datastore.templates.validateTemplates([ref])
1234 # Use a template that has a typo in dimension record metadata.
1235 # Easier to test with a butler that has a ref with records attached.
1236 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1237 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1238 path = template.format(ref)
1239 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1241 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1242 with self.assertRaises(KeyError):
1243 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1244 template.format(ref)
1246 # Now use a file template that will not result in unique filenames
1247 with self.assertRaises(FileTemplateValidationError):
1248 butler.put(metric, "metric3", dataId1)
1250 def testImportExport(self) -> None:
1251 # Run put/get tests just to create and populate a repo.
1252 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1253 self.runImportExportTest(storageClass)
1255 @unittest.expectedFailure
1256 def testImportExportVirtualComposite(self) -> None:
1257 # Run put/get tests just to create and populate a repo.
1258 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1259 self.runImportExportTest(storageClass)
1261 def runImportExportTest(self, storageClass: StorageClass) -> None:
1262 """This test does an export to a temp directory and an import back
1263 into a new temp directory repo. It does not assume a posix datastore"""
1264 exportButler = self.runPutGetTest(storageClass, "test_metric")
1266 # Test that we must have a file extension.
1267 with self.assertRaises(ValueError):
1268 with exportButler.export(filename="dump", directory=".") as export:
1269 pass
1271 # Test that unknown format is not allowed.
1272 with self.assertRaises(ValueError):
1273 with exportButler.export(filename="dump.fits", directory=".") as export:
1274 pass
1276 # Test that the repo actually has at least one dataset.
1277 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1278 self.assertGreater(len(datasets), 0)
1279 # Add a DimensionRecord that's unused by those datasets.
1280 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1281 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1282 # Export and then import datasets.
1283 with safeTestTempDir(TESTDIR) as exportDir:
1284 exportFile = os.path.join(exportDir, "exports.yaml")
1285 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1286 export.saveDatasets(datasets)
1287 # Export the same datasets again. This should quietly do
1288 # nothing because of internal deduplication, and it shouldn't
1289 # complain about being asked to export the "htm7" elements even
1290 # though there aren't any in these datasets or in the database.
1291 export.saveDatasets(datasets, elements=["htm7"])
1292 # Save one of the data IDs again; this should be harmless
1293 # because of internal deduplication.
1294 export.saveDataIds([datasets[0].dataId])
1295 # Save some dimension records directly.
1296 export.saveDimensionData("skymap", [skymapRecord])
1297 self.assertTrue(os.path.exists(exportFile))
1298 with safeTestTempDir(TESTDIR) as importDir:
1299 # We always want this to be a local posix butler
1300 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1301 # Calling script.butlerImport tests the implementation of the
1302 # butler command line interface "import" subcommand. Functions
1303 # in the script folder are generally considered protected and
1304 # should not be used as public api.
1305 with open(exportFile, "r") as f:
1306 script.butlerImport(
1307 importDir,
1308 export_file=f,
1309 directory=exportDir,
1310 transfer="auto",
1311 skip_dimensions=None,
1312 )
1313 importButler = Butler(importDir, run=self.default_run)
1314 for ref in datasets:
1315 with self.subTest(ref=ref):
1316 # Test for existence by passing in the DatasetType and
1317 # data ID separately, to avoid lookup by dataset_id.
1318 self.assertTrue(importButler.exists(ref.datasetType, ref.dataId))
1319 self.assertEqual(
1320 list(importButler.registry.queryDimensionRecords("skymap")),
1321 [importButler.registry.dimensions["skymap"].RecordClass(**skymapRecord)],
1322 )
1324 def testRemoveRuns(self) -> None:
1325 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1326 butler = Butler(self.tmpConfigFile, writeable=True)
1327 # Load registry data with dimensions to hang datasets off of.
1328 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1329 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1330 # Add some RUN-type collection.
1331 run1 = "run1"
1332 butler.registry.registerRun(run1)
1333 run2 = "run2"
1334 butler.registry.registerRun(run2)
1335 # put a dataset in each
1336 metric = makeExampleMetrics()
1337 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1338 datasetType = self.addDatasetType(
1339 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1340 )
1341 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1342 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1343 uri1 = butler.getURI(ref1)
1344 uri2 = butler.getURI(ref2)
1346 with self.assertRaises(OrphanedRecordError):
1347 butler.registry.removeDatasetType(datasetType.name)
1349 # Remove from both runs with different values for unstore.
1350 butler.removeRuns([run1], unstore=True)
1351 butler.removeRuns([run2], unstore=False)
1352 # Should be nothing in registry for either one, and datastore should
1353 # not think either exists.
1354 with self.assertRaises(MissingCollectionError):
1355 butler.registry.getCollectionType(run1)
1356 with self.assertRaises(MissingCollectionError):
1357 butler.registry.getCollectionType(run2)
1358 self.assertFalse(butler.datastore.exists(ref1))
1359 self.assertFalse(butler.datastore.exists(ref2))
1360 # The ref we unstored should be gone according to the URI, but the
1361 # one we forgot should still be around.
1362 self.assertFalse(uri1.exists())
1363 self.assertTrue(uri2.exists())
1365 # Now that the collections have been pruned we can remove the
1366 # dataset type
1367 butler.registry.removeDatasetType(datasetType.name)
1369 with self.assertLogs("lsst.daf.butler.registries", "INFO") as cm:
1370 butler.registry.removeDatasetType(tuple(["test*", "test*"]))
1371 self.assertIn("not defined", "\n".join(cm.output))
1374class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1375 """PosixDatastore specialization of a butler"""
1377 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1378 fullConfigKey: str | None = ".datastore.formatters"
1379 validationCanFail = True
1380 datastoreStr = ["/tmp"]
1381 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1382 registryStr = "/gen3.sqlite3"
1384 def testPathConstructor(self) -> None:
1385 """Independent test of constructor using PathLike."""
1386 butler = Butler(self.tmpConfigFile, run=self.default_run)
1387 self.assertIsInstance(butler, Butler)
1389 # And again with a Path object with the butler yaml
1390 path = pathlib.Path(self.tmpConfigFile)
1391 butler = Butler(path, writeable=False)
1392 self.assertIsInstance(butler, Butler)
1394 # And again with a Path object without the butler yaml
1395 # (making sure we skip it if the tmp config doesn't end
1396 # in butler.yaml -- which is the case for a subclass)
1397 if self.tmpConfigFile.endswith("butler.yaml"):
1398 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1399 butler = Butler(path, writeable=False)
1400 self.assertIsInstance(butler, Butler)
1402 def testExportTransferCopy(self) -> None:
1403 """Test local export using all transfer modes"""
1404 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1405 exportButler = self.runPutGetTest(storageClass, "test_metric")
1406 # Test that the repo actually has at least one dataset.
1407 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1408 self.assertGreater(len(datasets), 0)
1409 uris = [exportButler.getURI(d) for d in datasets]
1410 assert isinstance(exportButler.datastore, FileDatastore)
1411 datastoreRoot = exportButler.datastore.root
1413 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1415 for path in pathsInStore:
1416 # Assume local file system
1417 assert path is not None
1418 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1420 for transfer in ("copy", "link", "symlink", "relsymlink"):
1421 with safeTestTempDir(TESTDIR) as exportDir:
1422 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1423 export.saveDatasets(datasets)
1424 for path in pathsInStore:
1425 assert path is not None
1426 self.assertTrue(
1427 self.checkFileExists(exportDir, path),
1428 f"Check that mode {transfer} exported files",
1429 )
1431 def testPruneDatasets(self) -> None:
1432 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1433 butler = Butler(self.tmpConfigFile, writeable=True)
1434 assert isinstance(butler.datastore, FileDatastore)
1435 # Load registry data with dimensions to hang datasets off of.
1436 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1437 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1438 # Add some RUN-type collections.
1439 run1 = "run1"
1440 butler.registry.registerRun(run1)
1441 run2 = "run2"
1442 butler.registry.registerRun(run2)
1443 # put some datasets. ref1 and ref2 have the same data ID, and are in
1444 # different runs. ref3 has a different data ID.
1445 metric = makeExampleMetrics()
1446 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1447 datasetType = self.addDatasetType(
1448 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1449 )
1450 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1451 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1452 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1454 many_stored = butler.stored_many([ref1, ref2, ref3])
1455 for ref, stored in many_stored.items():
1456 self.assertTrue(stored, f"Ref {ref} should be stored")
1458 many_exists = butler._exists_many([ref1, ref2, ref3])
1459 for ref, exists in many_exists.items():
1460 self.assertTrue(exists, f"Checking ref {ref} exists.")
1461 self.assertEqual(exists, DatasetExistence.VERIFIED, f"Ref {ref} should be stored")
1463 # Simple prune.
1464 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1465 self.assertFalse(butler.exists(ref1.datasetType, ref1.dataId, collections=run1))
1467 many_stored = butler.stored_many([ref1, ref2, ref3])
1468 for ref, stored in many_stored.items():
1469 self.assertFalse(stored, f"Ref {ref} should not be stored")
1471 many_exists = butler._exists_many([ref1, ref2, ref3])
1472 for ref, exists in many_exists.items():
1473 self.assertEqual(exists, DatasetExistence.UNRECOGNIZED, f"Ref {ref} should not be stored")
1475 # Put data back.
1476 ref1_new = butler.put(metric, ref1)
1477 self.assertEqual(ref1_new, ref1) # Reuses original ID.
1478 ref2 = butler.put(metric, ref2)
1480 many_stored = butler.stored_many([ref1, ref2, ref3])
1481 self.assertTrue(many_stored[ref1])
1482 self.assertTrue(many_stored[ref2])
1483 self.assertFalse(many_stored[ref3])
1485 ref3 = butler.put(metric, ref3)
1487 many_exists = butler._exists_many([ref1, ref2, ref3])
1488 for ref, exists in many_exists.items():
1489 self.assertTrue(exists, f"Ref {ref} should not be stored")
1491 # Clear out the datasets from registry and start again.
1492 refs = [ref1, ref2, ref3]
1493 butler.pruneDatasets(refs, purge=True, unstore=True)
1494 for ref in refs:
1495 butler.put(metric, ref)
1497 # Test different forms of file availability.
1498 # Need to be in a state where:
1499 # - one ref just has registry record.
1500 # - one ref has a missing file but a datastore record.
1501 # - one ref has a missing datastore record but file is there.
1502 # - one ref does not exist anywhere.
1503 # Do not need to test a ref that has everything since that is tested
1504 # above.
1505 ref0 = DatasetRef(
1506 datasetType,
1507 DataCoordinate.standardize(
1508 {"instrument": "Cam1", "physical_filter": "Cam1-G"}, universe=butler.dimensions
1509 ),
1510 run=run1,
1511 )
1513 # Delete from datastore and retain in Registry.
1514 butler.pruneDatasets([ref1], purge=False, unstore=True, disassociate=False)
1516 # File has been removed.
1517 uri2 = butler.datastore.getURI(ref2)
1518 uri2.remove()
1520 # Datastore has lost track.
1521 butler.datastore.forget([ref3])
1523 # First test with a standard butler.
1524 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1525 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1526 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1527 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1528 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED)
1530 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=False)
1531 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1532 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1533 self.assertEqual(exists_many[ref2], DatasetExistence.KNOWN)
1534 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ASSUMED)
1535 self.assertTrue(exists_many[ref2])
1537 # Check that per-ref query gives the same answer as many query.
1538 for ref, exists in exists_many.items():
1539 self.assertEqual(butler.exists(ref, full_check=False), exists)
1541 # Test again with a trusting butler.
1542 butler.datastore.trustGetRequest = True
1543 exists_many = butler._exists_many([ref0, ref1, ref2, ref3], full_check=True)
1544 self.assertEqual(exists_many[ref0], DatasetExistence.UNRECOGNIZED)
1545 self.assertEqual(exists_many[ref1], DatasetExistence.RECORDED)
1546 self.assertEqual(exists_many[ref2], DatasetExistence.RECORDED | DatasetExistence.DATASTORE)
1547 self.assertEqual(exists_many[ref3], DatasetExistence.RECORDED | DatasetExistence._ARTIFACT)
1549 # Check that per-ref query gives the same answer as many query.
1550 for ref, exists in exists_many.items():
1551 self.assertEqual(butler.exists(ref, full_check=True), exists)
1553 # Create a ref that surprisingly has the UUID of an existing ref
1554 # but is not the same.
1555 ref_bad = DatasetRef(datasetType, dataId=ref3.dataId, run=ref3.run, id=ref2.id)
1556 with self.assertRaises(ValueError):
1557 butler.exists(ref_bad)
1559 # Create a ref that has a compatible storage class.
1560 ref_compat = ref2.overrideStorageClass("StructuredDataDict")
1561 exists = butler.exists(ref_compat)
1562 self.assertEqual(exists, exists_many[ref2])
1564 # Remove everything and start from scratch.
1565 butler.datastore.trustGetRequest = False
1566 butler.pruneDatasets(refs, purge=True, unstore=True)
1567 for ref in refs:
1568 butler.put(metric, ref)
1570 # These tests mess directly with the trash table and can leave the
1571 # datastore in an odd state. Do them at the end.
1572 # Check that in normal mode, deleting the record will lead to
1573 # trash not touching the file.
1574 uri1 = butler.datastore.getURI(ref1)
1575 butler.datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1576 butler.datastore.forget([ref1])
1577 butler.datastore.trash(ref1)
1578 butler.datastore.emptyTrash()
1579 self.assertTrue(uri1.exists())
1580 uri1.remove() # Clean it up.
1582 # Simulate execution butler setup by deleting the datastore
1583 # record but keeping the file around and trusting.
1584 butler.datastore.trustGetRequest = True
1585 uri2 = butler.datastore.getURI(ref2)
1586 uri3 = butler.datastore.getURI(ref3)
1587 self.assertTrue(uri2.exists())
1588 self.assertTrue(uri3.exists())
1590 # Remove the datastore record.
1591 butler.datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1592 butler.datastore.forget([ref2])
1593 self.assertTrue(uri2.exists())
1594 butler.datastore.trash([ref2, ref3])
1595 # Immediate removal for ref2 file
1596 self.assertFalse(uri2.exists())
1597 # But ref3 has to wait for the empty.
1598 self.assertTrue(uri3.exists())
1599 butler.datastore.emptyTrash()
1600 self.assertFalse(uri3.exists())
1602 # Clear out the datasets from registry.
1603 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1605 def testPytypeCoercion(self) -> None:
1606 """Test python type coercion on Butler.get and put."""
1608 # Store some data with the normal example storage class.
1609 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1610 datasetTypeName = "test_metric"
1611 butler = self.runPutGetTest(storageClass, datasetTypeName)
1613 dataId = {"instrument": "DummyCamComp", "visit": 423}
1614 metric = butler.get(datasetTypeName, dataId=dataId)
1615 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1617 datasetType_ori = butler.registry.getDatasetType(datasetTypeName)
1618 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1620 # Now need to hack the registry dataset type definition.
1621 # There is no API for this.
1622 assert isinstance(butler.registry, SqlRegistry)
1623 manager = butler.registry._managers.datasets
1624 assert hasattr(manager, "_db") and hasattr(manager, "_static")
1625 manager._db.update(
1626 manager._static.dataset_type,
1627 {"name": datasetTypeName},
1628 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1629 )
1631 # Force reset of dataset type cache
1632 butler.registry.refresh()
1634 datasetType_new = butler.registry.getDatasetType(datasetTypeName)
1635 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1636 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1638 metric_model = butler.get(datasetTypeName, dataId=dataId)
1639 self.assertNotEqual(type(metric_model), type(metric))
1640 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1642 # Put the model and read it back to show that everything now
1643 # works as normal.
1644 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1645 metric_model_new = butler.get(metric_ref)
1646 self.assertEqual(metric_model_new, metric_model)
1648 # Hack the storage class again to something that will fail on the
1649 # get with no conversion class.
1650 manager._db.update(
1651 manager._static.dataset_type,
1652 {"name": datasetTypeName},
1653 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1654 )
1655 butler.registry.refresh()
1657 with self.assertRaises(ValueError):
1658 butler.get(datasetTypeName, dataId=dataId)
1661@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1662class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1663 """PosixDatastore specialization of a butler using Postgres"""
1665 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1666 fullConfigKey = ".datastore.formatters"
1667 validationCanFail = True
1668 datastoreStr = ["/tmp"]
1669 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1670 registryStr = "PostgreSQL@test"
1671 postgresql: Any
1673 @staticmethod
1674 def _handler(postgresql: Any) -> None:
1675 engine = sqlalchemy.engine.create_engine(postgresql.url())
1676 with engine.begin() as connection:
1677 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1679 @classmethod
1680 def setUpClass(cls) -> None:
1681 # Create the postgres test server.
1682 cls.postgresql = testing.postgresql.PostgresqlFactory(
1683 cache_initialized_db=True, on_initialized=cls._handler
1684 )
1685 super().setUpClass()
1687 @classmethod
1688 def tearDownClass(cls) -> None:
1689 # Clean up any lingering SQLAlchemy engines/connections
1690 # so they're closed before we shut down the server.
1691 gc.collect()
1692 cls.postgresql.clear_cache()
1693 super().tearDownClass()
1695 def setUp(self) -> None:
1696 self.server = self.postgresql()
1698 # Need to add a registry section to the config.
1699 self._temp_config = False
1700 config = Config(self.configFile)
1701 config["registry", "db"] = self.server.url()
1702 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1703 config.dump(fh)
1704 self.configFile = fh.name
1705 self._temp_config = True
1706 super().setUp()
1708 def tearDown(self) -> None:
1709 self.server.stop()
1710 if self._temp_config and os.path.exists(self.configFile):
1711 os.remove(self.configFile)
1712 super().tearDown()
1714 def testMakeRepo(self) -> None:
1715 # The base class test assumes that it's using sqlite and assumes
1716 # the config file is acceptable to sqlite.
1717 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1720class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1721 """InMemoryDatastore specialization of a butler"""
1723 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1724 fullConfigKey = None
1725 useTempRoot = False
1726 validationCanFail = False
1727 datastoreStr = ["datastore='InMemory"]
1728 datastoreName = ["InMemoryDatastore@"]
1729 registryStr = "/gen3.sqlite3"
1731 def testIngest(self) -> None:
1732 pass
1735class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1736 """PosixDatastore specialization"""
1738 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1739 fullConfigKey = ".datastore.datastores.1.formatters"
1740 validationCanFail = True
1741 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1742 datastoreName = [
1743 "InMemoryDatastore@",
1744 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1745 "SecondDatastore",
1746 ]
1747 registryStr = "/gen3.sqlite3"
1750class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1751 """Test that a yaml file in one location can refer to a root in another."""
1753 datastoreStr = ["dir1"]
1754 # Disable the makeRepo test since we are deliberately not using
1755 # butler.yaml as the config name.
1756 fullConfigKey = None
1758 def setUp(self) -> None:
1759 self.root = makeTestTempDir(TESTDIR)
1761 # Make a new repository in one place
1762 self.dir1 = os.path.join(self.root, "dir1")
1763 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1765 # Move the yaml file to a different place and add a "root"
1766 self.dir2 = os.path.join(self.root, "dir2")
1767 os.makedirs(self.dir2, exist_ok=True)
1768 configFile1 = os.path.join(self.dir1, "butler.yaml")
1769 config = Config(configFile1)
1770 config["root"] = self.dir1
1771 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1772 config.dumpToUri(configFile2)
1773 os.remove(configFile1)
1774 self.tmpConfigFile = configFile2
1776 def testFileLocations(self) -> None:
1777 self.assertNotEqual(self.dir1, self.dir2)
1778 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1779 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1780 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1783class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1784 """Test that a config file created by makeRepo outside of repo works."""
1786 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1788 def setUp(self) -> None:
1789 self.root = makeTestTempDir(TESTDIR)
1790 self.root2 = makeTestTempDir(TESTDIR)
1792 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1793 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1795 def tearDown(self) -> None:
1796 if os.path.exists(self.root2):
1797 shutil.rmtree(self.root2, ignore_errors=True)
1798 super().tearDown()
1800 def testConfigExistence(self) -> None:
1801 c = Config(self.tmpConfigFile)
1802 uri_config = ResourcePath(c["root"])
1803 uri_expected = ResourcePath(self.root, forceDirectory=True)
1804 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1805 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1807 def testPutGet(self) -> None:
1808 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1809 self.runPutGetTest(storageClass, "test_metric")
1812class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase):
1813 """Test that a config file created by makeRepo outside of repo works."""
1815 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1817 def setUp(self) -> None:
1818 self.root = makeTestTempDir(TESTDIR)
1819 self.root2 = makeTestTempDir(TESTDIR)
1821 self.tmpConfigFile = self.root2
1822 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1824 def testConfigExistence(self) -> None:
1825 # Append the yaml file else Config constructor does not know the file
1826 # type.
1827 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1828 super().testConfigExistence()
1831class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1832 """Test that a config file created by makeRepo outside of repo works."""
1834 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1836 def setUp(self) -> None:
1837 self.root = makeTestTempDir(TESTDIR)
1838 self.root2 = makeTestTempDir(TESTDIR)
1840 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1841 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1844@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1845class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1846 """S3Datastore specialization of a butler; an S3 storage Datastore +
1847 a local in-memory SqlRegistry.
1848 """
1850 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1851 fullConfigKey = None
1852 validationCanFail = True
1854 bucketName = "anybucketname"
1855 """Name of the Bucket that will be used in the tests. The name is read from
1856 the config file used with the tests during set-up.
1857 """
1859 root = "butlerRoot/"
1860 """Root repository directory expected to be used in case useTempRoot=False.
1861 Otherwise the root is set to a 20 characters long randomly generated string
1862 during set-up.
1863 """
1865 datastoreStr = [f"datastore={root}"]
1866 """Contains all expected root locations in a format expected to be
1867 returned by Butler stringification.
1868 """
1870 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1871 """The expected format of the S3 Datastore string."""
1873 registryStr = "/gen3.sqlite3"
1874 """Expected format of the Registry string."""
1876 mock_s3 = mock_s3()
1877 """The mocked s3 interface from moto."""
1879 def genRoot(self) -> str:
1880 """Returns a random string of len 20 to serve as a root
1881 name for the temporary bucket repo.
1883 This is equivalent to tempfile.mkdtemp as this is what self.root
1884 becomes when useTempRoot is True.
1885 """
1886 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1887 return rndstr + "/"
1889 def setUp(self) -> None:
1890 config = Config(self.configFile)
1891 uri = ResourcePath(config[".datastore.datastore.root"])
1892 self.bucketName = uri.netloc
1894 # Enable S3 mocking of tests.
1895 self.mock_s3.start()
1897 # set up some fake credentials if they do not exist
1898 self.usingDummyCredentials = setAwsEnvCredentials()
1900 if self.useTempRoot:
1901 self.root = self.genRoot()
1902 rooturi = f"s3://{self.bucketName}/{self.root}"
1903 config.update({"datastore": {"datastore": {"root": rooturi}}})
1905 # need local folder to store registry database
1906 self.reg_dir = makeTestTempDir(TESTDIR)
1907 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
1909 # MOTO needs to know that we expect Bucket bucketname to exist
1910 # (this used to be the class attribute bucketName)
1911 s3 = boto3.resource("s3")
1912 s3.create_bucket(Bucket=self.bucketName)
1914 self.datastoreStr = [f"datastore='{rooturi}'"]
1915 self.datastoreName = [f"FileDatastore@{rooturi}"]
1916 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
1917 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
1919 def tearDown(self) -> None:
1920 s3 = boto3.resource("s3")
1921 bucket = s3.Bucket(self.bucketName)
1922 try:
1923 bucket.objects.all().delete()
1924 except botocore.exceptions.ClientError as e:
1925 if e.response["Error"]["Code"] == "404":
1926 # the key was not reachable - pass
1927 pass
1928 else:
1929 raise
1931 bucket = s3.Bucket(self.bucketName)
1932 bucket.delete()
1934 # Stop the S3 mock.
1935 self.mock_s3.stop()
1937 # unset any potentially set dummy credentials
1938 if self.usingDummyCredentials:
1939 unsetAwsEnvCredentials()
1941 if self.reg_dir is not None and os.path.exists(self.reg_dir):
1942 shutil.rmtree(self.reg_dir, ignore_errors=True)
1944 if self.useTempRoot and os.path.exists(self.root):
1945 shutil.rmtree(self.root, ignore_errors=True)
1947 super().tearDown()
1950class PosixDatastoreTransfers(unittest.TestCase):
1951 """Test data transfers between butlers.
1953 Test for different managers. UUID to UUID and integer to integer are
1954 tested. UUID to integer is not supported since we do not currently
1955 want to allow that. Integer to UUID is supported with the caveat
1956 that UUID4 will be generated and this will be incorrect for raw
1957 dataset types. The test ignores that.
1958 """
1960 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1961 storageClassFactory: StorageClassFactory
1963 @classmethod
1964 def setUpClass(cls) -> None:
1965 cls.storageClassFactory = StorageClassFactory()
1966 cls.storageClassFactory.addFromConfig(cls.configFile)
1968 def setUp(self) -> None:
1969 self.root = makeTestTempDir(TESTDIR)
1970 self.config = Config(self.configFile)
1972 def tearDown(self) -> None:
1973 removeTestTempDir(self.root)
1975 def create_butler(self, manager: str, label: str) -> Butler:
1976 config = Config(self.configFile)
1977 config["registry", "managers", "datasets"] = manager
1978 return Butler(Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True)
1980 def create_butlers(self, manager1: str | None = None, manager2: str | None = None) -> None:
1981 default = "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID"
1982 if manager1 is None:
1983 manager1 = default
1984 if manager2 is None:
1985 manager2 = default
1986 self.source_butler = self.create_butler(manager1, "1")
1987 self.target_butler = self.create_butler(manager2, "2")
1989 def testTransferUuidToUuid(self) -> None:
1990 self.create_butlers()
1991 self.assertButlerTransfers()
1993 def _enable_trust(self, datastore: Datastore) -> None:
1994 if hasattr(datastore, "trustGetRequest"):
1995 datastore.trustGetRequest = True
1996 elif hasattr(datastore, "datastores"):
1997 for datastore in datastore.datastores:
1998 if hasattr(datastore, "trustGetRequest"):
1999 datastore.trustGetRequest = True
2001 def testTransferMissing(self) -> None:
2002 """Test transfers where datastore records are missing.
2004 This is how execution butler works.
2005 """
2006 self.create_butlers()
2008 # Configure the source butler to allow trust.
2009 self._enable_trust(self.source_butler.datastore)
2011 self.assertButlerTransfers(purge=True)
2013 def testTransferMissingDisassembly(self) -> None:
2014 """Test transfers where datastore records are missing.
2016 This is how execution butler works.
2017 """
2018 self.create_butlers()
2020 # Configure the source butler to allow trust.
2021 self._enable_trust(self.source_butler.datastore)
2023 # Test disassembly.
2024 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
2026 def testAbsoluteURITransferDirect(self) -> None:
2027 """Test transfer using an absolute URI."""
2028 self._absolute_transfer("auto")
2030 def testAbsoluteURITransferCopy(self) -> None:
2031 """Test transfer using an absolute URI."""
2032 self._absolute_transfer("copy")
2034 def _absolute_transfer(self, transfer: str) -> None:
2035 self.create_butlers()
2037 storageClassName = "StructuredData"
2038 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2039 datasetTypeName = "random_data"
2040 run = "run1"
2041 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2043 dimensions = self.source_butler.registry.dimensions.extract(())
2044 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2045 self.source_butler.registry.registerDatasetType(datasetType)
2047 metrics = makeExampleMetrics()
2048 with ResourcePath.temporary_uri(suffix=".json") as temp:
2049 dataId = DataCoordinate.makeEmpty(self.source_butler.dimensions)
2050 source_refs = [DatasetRef(datasetType, dataId, run=run)]
2051 temp.write(json.dumps(metrics.exportAsDict()).encode())
2052 dataset = FileDataset(path=temp, refs=source_refs)
2053 self.source_butler.ingest(dataset, transfer="direct")
2055 self.target_butler.transfer_from(
2056 self.source_butler, dataset.refs, register_dataset_types=True, transfer=transfer
2057 )
2059 uri = self.target_butler.getURI(dataset.refs[0])
2060 if transfer == "auto":
2061 self.assertEqual(uri, temp)
2062 else:
2063 self.assertNotEqual(uri, temp)
2065 def assertButlerTransfers(self, purge: bool = False, storageClassName: str = "StructuredData") -> None:
2066 """Test that a run can be transferred to another butler."""
2068 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2069 datasetTypeName = "random_data"
2071 # Test will create 3 collections and we will want to transfer
2072 # two of those three.
2073 runs = ["run1", "run2", "other"]
2075 # Also want to use two different dataset types to ensure that
2076 # grouping works.
2077 datasetTypeNames = ["random_data", "random_data_2"]
2079 # Create the run collections in the source butler.
2080 for run in runs:
2081 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2083 # Create dimensions in source butler.
2084 n_exposures = 30
2085 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
2086 self.source_butler.registry.insertDimensionData(
2087 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
2088 )
2089 self.source_butler.registry.insertDimensionData(
2090 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
2091 )
2093 for i in range(n_exposures):
2094 self.source_butler.registry.insertDimensionData(
2095 "exposure",
2096 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
2097 )
2099 # Create dataset types in the source butler.
2100 dimensions = self.source_butler.registry.dimensions.extract(["instrument", "exposure"])
2101 for datasetTypeName in datasetTypeNames:
2102 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2103 self.source_butler.registry.registerDatasetType(datasetType)
2105 # Write a dataset to an unrelated run -- this will ensure that
2106 # we are rewriting integer dataset ids in the target if necessary.
2107 # Will not be relevant for UUID.
2108 run = "distraction"
2109 butler = Butler(butler=self.source_butler, run=run)
2110 butler.put(
2111 makeExampleMetrics(),
2112 datasetTypeName,
2113 exposure=1,
2114 instrument="DummyCamComp",
2115 physical_filter="d-r",
2116 )
2118 # Write some example metrics to the source
2119 butler = Butler(butler=self.source_butler)
2121 # Set of DatasetRefs that should be in the list of refs to transfer
2122 # but which will not be transferred.
2123 deleted: set[DatasetRef] = set()
2125 n_expected = 20 # Number of datasets expected to be transferred
2126 source_refs = []
2127 for i in range(n_exposures):
2128 # Put a third of datasets into each collection, only retain
2129 # two thirds.
2130 index = i % 3
2131 run = runs[index]
2132 datasetTypeName = datasetTypeNames[i % 2]
2134 metric = MetricsExample(
2135 summary={"counter": i}, output={"text": "metric"}, data=[2 * x for x in range(i)]
2136 )
2137 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2138 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2140 # Remove the datastore record using low-level API
2141 if purge:
2142 # Remove records for a fraction.
2143 if index == 1:
2144 # For one of these delete the file as well.
2145 # This allows the "missing" code to filter the
2146 # file out.
2147 # Access the individual datastores.
2148 datastores = []
2149 if hasattr(butler.datastore, "datastores"):
2150 datastores.extend(butler.datastore.datastores)
2151 else:
2152 datastores.append(butler.datastore)
2154 if not deleted:
2155 # For a chained datastore we need to remove
2156 # files in each chain.
2157 for datastore in datastores:
2158 # The file might not be known to the datastore
2159 # if constraints are used.
2160 try:
2161 primary, uris = datastore.getURIs(ref)
2162 except FileNotFoundError:
2163 continue
2164 if primary:
2165 if primary.scheme != "mem":
2166 primary.remove()
2167 for uri in uris.values():
2168 if uri.scheme != "mem":
2169 uri.remove()
2170 n_expected -= 1
2171 deleted.add(ref)
2173 # Remove the datastore record.
2174 for datastore in datastores:
2175 if hasattr(datastore, "removeStoredItemInfo"):
2176 datastore.removeStoredItemInfo(ref)
2178 if index < 2:
2179 source_refs.append(ref)
2180 if ref not in deleted:
2181 new_metric = butler.get(ref)
2182 self.assertEqual(new_metric, metric)
2184 # Create some bad dataset types to ensure we check for inconsistent
2185 # definitions.
2186 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2187 for datasetTypeName in datasetTypeNames:
2188 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2189 self.target_butler.registry.registerDatasetType(datasetType)
2190 with self.assertRaises(ConflictingDefinitionError) as cm:
2191 self.target_butler.transfer_from(self.source_butler, source_refs)
2192 self.assertIn("dataset type differs", str(cm.exception))
2194 # And remove the bad definitions.
2195 for datasetTypeName in datasetTypeNames:
2196 self.target_butler.registry.removeDatasetType(datasetTypeName)
2198 # Transfer without creating dataset types should fail.
2199 with self.assertRaises(KeyError):
2200 self.target_butler.transfer_from(self.source_butler, source_refs)
2202 # Transfer without creating dimensions should fail.
2203 with self.assertRaises(ConflictingDefinitionError) as cm:
2204 self.target_butler.transfer_from(self.source_butler, source_refs, register_dataset_types=True)
2205 self.assertIn("dimension", str(cm.exception))
2207 # The failed transfer above leaves registry in an inconsistent
2208 # state because the run is created but then rolled back without
2209 # the collection cache being cleared. For now force a refresh.
2210 # Can remove with DM-35498.
2211 self.target_butler.registry.refresh()
2213 # Now transfer them to the second butler, including dimensions.
2214 with self.assertLogs(level=logging.DEBUG) as log_cm:
2215 transferred = self.target_butler.transfer_from(
2216 self.source_butler,
2217 source_refs,
2218 register_dataset_types=True,
2219 transfer_dimensions=True,
2220 )
2221 self.assertEqual(len(transferred), n_expected)
2222 log_output = ";".join(log_cm.output)
2224 # A ChainedDatastore will use the in-memory datastore for mexists
2225 # so we can not rely on the mexists log message.
2226 self.assertIn("Number of datastore records found in source", log_output)
2227 self.assertIn("Creating output run", log_output)
2229 # Do the transfer twice to ensure that it will do nothing extra.
2230 # Only do this if purge=True because it does not work for int
2231 # dataset_id.
2232 if purge:
2233 # This should not need to register dataset types.
2234 transferred = self.target_butler.transfer_from(self.source_butler, source_refs)
2235 self.assertEqual(len(transferred), n_expected)
2237 # Also do an explicit low-level transfer to trigger some
2238 # edge cases.
2239 with self.assertLogs(level=logging.DEBUG) as log_cm:
2240 self.target_butler.datastore.transfer_from(self.source_butler.datastore, source_refs)
2241 log_output = ";".join(log_cm.output)
2242 self.assertIn("no file artifacts exist", log_output)
2244 with self.assertRaises((TypeError, AttributeError)):
2245 self.target_butler.datastore.transfer_from(self.source_butler, source_refs) # type: ignore
2247 with self.assertRaises(ValueError):
2248 self.target_butler.datastore.transfer_from(
2249 self.source_butler.datastore, source_refs, transfer="split"
2250 )
2252 # Now try to get the same refs from the new butler.
2253 for ref in source_refs:
2254 if ref not in deleted:
2255 new_metric = self.target_butler.get(ref)
2256 old_metric = self.source_butler.get(ref)
2257 self.assertEqual(new_metric, old_metric)
2259 # Now prune run2 collection and create instead a CHAINED collection.
2260 # This should block the transfer.
2261 self.target_butler.removeRuns(["run2"], unstore=True)
2262 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2263 with self.assertRaises(CollectionTypeError):
2264 # Re-importing the run1 datasets can be problematic if they
2265 # use integer IDs so filter those out.
2266 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2267 self.target_butler.transfer_from(self.source_butler, to_transfer)
2270class ChainedDatastoreTransfers(PosixDatastoreTransfers):
2271 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
2274if __name__ == "__main__":
2275 unittest.main()