Coverage for tests/test_butler.py: 12%
1157 statements
« prev ^ index » next coverage.py v6.5.0, created at 2023-02-08 10:28 +0000
« prev ^ index » next coverage.py v6.5.0, created at 2023-02-08 10:28 +0000
1# This file is part of daf_butler.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This 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"""
25import gc
26import logging
27import os
28import pathlib
29import pickle
30import posixpath
31import random
32import shutil
33import string
34import tempfile
35import unittest
37try:
38 import boto3
39 import botocore
40 from moto import mock_s3
41except ImportError:
42 boto3 = None
44 def mock_s3(cls):
45 """A no-op decorator in case moto mock_s3 can not be imported."""
46 return cls
49try:
50 # It's possible but silly to have testing.postgresql installed without
51 # having the postgresql server installed (because then nothing in
52 # testing.postgresql would work), so we use the presence of that module
53 # to test whether we can expect the server to be available.
54 import testing.postgresql
55except ImportError:
56 testing = None
58import astropy.time
59import sqlalchemy
60from lsst.daf.butler import (
61 Butler,
62 ButlerConfig,
63 CollectionType,
64 Config,
65 DatasetIdGenEnum,
66 DatasetRef,
67 DatasetType,
68 FileDataset,
69 FileTemplate,
70 FileTemplateValidationError,
71 StorageClassFactory,
72 ValidationError,
73 script,
74)
75from lsst.daf.butler.core.repoRelocation import BUTLER_ROOT_TAG
76from lsst.daf.butler.registry import (
77 CollectionError,
78 CollectionTypeError,
79 ConflictingDefinitionError,
80 DataIdValueError,
81 MissingCollectionError,
82)
83from lsst.daf.butler.tests import MetricsExample, MultiDetectorFormatter
84from lsst.daf.butler.tests.utils import makeTestTempDir, removeTestTempDir, safeTestTempDir
85from lsst.resources import ResourcePath
86from lsst.resources.s3utils import setAwsEnvCredentials, unsetAwsEnvCredentials
87from lsst.utils import doImport
88from lsst.utils.introspection import get_full_type_name
90TESTDIR = os.path.abspath(os.path.dirname(__file__))
93def makeExampleMetrics():
94 return MetricsExample(
95 {"AM1": 5.2, "AM2": 30.6},
96 {"a": [1, 2, 3], "b": {"blue": 5, "red": "green"}},
97 [563, 234, 456.7, 752, 8, 9, 27],
98 )
101class TransactionTestError(Exception):
102 """Specific error for testing transactions, to prevent misdiagnosing
103 that might otherwise occur when a standard exception is used.
104 """
106 pass
109class ButlerConfigTests(unittest.TestCase):
110 """Simple tests for ButlerConfig that are not tested in any other test
111 cases."""
113 def testSearchPath(self):
114 configFile = os.path.join(TESTDIR, "config", "basic", "butler.yaml")
115 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
116 config1 = ButlerConfig(configFile)
117 self.assertNotIn("testConfigs", "\n".join(cm.output))
119 overrideDirectory = os.path.join(TESTDIR, "config", "testConfigs")
120 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
121 config2 = ButlerConfig(configFile, searchPaths=[overrideDirectory])
122 self.assertIn("testConfigs", "\n".join(cm.output))
124 key = ("datastore", "records", "table")
125 self.assertNotEqual(config1[key], config2[key])
126 self.assertEqual(config2[key], "override_record")
129class ButlerPutGetTests:
130 """Helper method for running a suite of put/get tests from different
131 butler configurations."""
133 root = None
134 default_run = "ingésτ😺"
136 @staticmethod
137 def addDatasetType(datasetTypeName, dimensions, storageClass, registry):
138 """Create a DatasetType and register it"""
139 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
140 registry.registerDatasetType(datasetType)
141 return datasetType
143 @classmethod
144 def setUpClass(cls):
145 cls.storageClassFactory = StorageClassFactory()
146 cls.storageClassFactory.addFromConfig(cls.configFile)
148 def assertGetComponents(self, butler, datasetRef, components, reference, collections=None):
149 datasetType = datasetRef.datasetType
150 dataId = datasetRef.dataId
151 deferred = butler.getDirectDeferred(datasetRef)
153 for component in components:
154 compTypeName = datasetType.componentTypeName(component)
155 result = butler.get(compTypeName, dataId, collections=collections)
156 self.assertEqual(result, getattr(reference, component))
157 result_deferred = deferred.get(component=component)
158 self.assertEqual(result_deferred, result)
160 def tearDown(self):
161 removeTestTempDir(self.root)
163 def create_butler(self, run, storageClass, datasetTypeName):
164 butler = Butler(self.tmpConfigFile, run=run)
166 collections = set(butler.registry.queryCollections())
167 self.assertEqual(collections, set([run]))
169 # Create and register a DatasetType
170 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
172 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
174 # Add needed Dimensions
175 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
176 butler.registry.insertDimensionData(
177 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
178 )
179 butler.registry.insertDimensionData(
180 "visit_system", {"instrument": "DummyCamComp", "id": 1, "name": "default"}
181 )
182 visit_start = astropy.time.Time("2020-01-01 08:00:00.123456789", scale="tai")
183 visit_end = astropy.time.Time("2020-01-01 08:00:36.66", scale="tai")
184 butler.registry.insertDimensionData(
185 "visit",
186 {
187 "instrument": "DummyCamComp",
188 "id": 423,
189 "name": "fourtwentythree",
190 "physical_filter": "d-r",
191 "visit_system": 1,
192 "datetime_begin": visit_start,
193 "datetime_end": visit_end,
194 },
195 )
197 # Add more visits for some later tests
198 for visit_id in (424, 425):
199 butler.registry.insertDimensionData(
200 "visit",
201 {
202 "instrument": "DummyCamComp",
203 "id": visit_id,
204 "name": f"fourtwentyfour_{visit_id}",
205 "physical_filter": "d-r",
206 "visit_system": 1,
207 },
208 )
209 return butler, datasetType
211 def runPutGetTest(self, storageClass, datasetTypeName):
212 # New datasets will be added to run and tag, but we will only look in
213 # tag when looking up datasets.
214 run = self.default_run
215 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
217 # Create and store a dataset
218 metric = makeExampleMetrics()
219 dataId = {"instrument": "DummyCamComp", "visit": 423}
221 # Create a DatasetRef for put
222 refIn = DatasetRef(datasetType, dataId, id=None)
224 # Put with a preexisting id should fail
225 with self.assertRaises(ValueError):
226 butler.put(metric, DatasetRef(datasetType, dataId, id=100))
228 # Put and remove the dataset once as a DatasetRef, once as a dataId,
229 # and once with a DatasetType
231 # Keep track of any collections we add and do not clean up
232 expected_collections = {run}
234 counter = 0
235 for args in ((refIn,), (datasetTypeName, dataId), (datasetType, dataId)):
236 # Since we are using subTest we can get cascading failures
237 # here with the first attempt failing and the others failing
238 # immediately because the dataset already exists. Work around
239 # this by using a distinct run collection each time
240 counter += 1
241 this_run = f"put_run_{counter}"
242 butler.registry.registerCollection(this_run, type=CollectionType.RUN)
243 expected_collections.update({this_run})
245 with self.subTest(args=args):
246 ref = butler.put(metric, *args, run=this_run)
247 self.assertIsInstance(ref, DatasetRef)
249 # Test getDirect
250 metricOut = butler.getDirect(ref)
251 self.assertEqual(metric, metricOut)
252 # Test get
253 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
254 self.assertEqual(metric, metricOut)
255 # Test get with a datasetRef
256 metricOut = butler.get(ref, collections=this_run)
257 self.assertEqual(metric, metricOut)
258 # Test getDeferred with dataId
259 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
260 self.assertEqual(metric, metricOut)
261 # Test getDeferred with a datasetRef
262 metricOut = butler.getDeferred(ref, collections=this_run).get()
263 self.assertEqual(metric, metricOut)
264 # and deferred direct with ref
265 metricOut = butler.getDirectDeferred(ref).get()
266 self.assertEqual(metric, metricOut)
268 # Check we can get components
269 if storageClass.isComposite():
270 self.assertGetComponents(
271 butler, ref, ("summary", "data", "output"), metric, collections=this_run
272 )
274 # Can the artifacts themselves be retrieved?
275 if not butler.datastore.isEphemeral:
276 root_uri = ResourcePath(self.root)
278 for preserve_path in (True, False):
279 destination = root_uri.join(f"artifacts/{preserve_path}_{counter}/")
280 # Use copy so that we can test that overwrite
281 # protection works (using "auto" for File URIs would
282 # use hard links and subsequent transfer would work
283 # because it knows they are the same file).
284 transferred = butler.retrieveArtifacts(
285 [ref], destination, preserve_path=preserve_path, transfer="copy"
286 )
287 self.assertGreater(len(transferred), 0)
288 artifacts = list(ResourcePath.findFileResources([destination]))
289 self.assertEqual(set(transferred), set(artifacts))
291 for artifact in transferred:
292 path_in_destination = artifact.relative_to(destination)
293 self.assertIsNotNone(path_in_destination)
295 # when path is not preserved there should not be
296 # any path separators.
297 num_seps = path_in_destination.count("/")
298 if preserve_path:
299 self.assertGreater(num_seps, 0)
300 else:
301 self.assertEqual(num_seps, 0)
303 primary_uri, secondary_uris = butler.datastore.getURIs(ref)
304 n_uris = len(secondary_uris)
305 if primary_uri:
306 n_uris += 1
307 self.assertEqual(
308 len(artifacts),
309 n_uris,
310 "Comparing expected artifacts vs actual:"
311 f" {artifacts} vs {primary_uri} and {secondary_uris}",
312 )
314 if preserve_path:
315 # No need to run these twice
316 with self.assertRaises(ValueError):
317 butler.retrieveArtifacts([ref], destination, transfer="move")
319 with self.assertRaises(FileExistsError):
320 butler.retrieveArtifacts([ref], destination)
322 transferred_again = butler.retrieveArtifacts(
323 [ref], destination, preserve_path=preserve_path, overwrite=True
324 )
325 self.assertEqual(set(transferred_again), set(transferred))
327 # Now remove the dataset completely.
328 butler.pruneDatasets([ref], purge=True, unstore=True)
329 # Lookup with original args should still fail.
330 with self.assertRaises(LookupError):
331 butler.datasetExists(*args, collections=this_run)
332 # getDirect() should still fail.
333 with self.assertRaises(FileNotFoundError):
334 butler.getDirect(ref)
335 # Registry shouldn't be able to find it by dataset_id anymore.
336 self.assertIsNone(butler.registry.getDataset(ref.id))
338 # Do explicit registry removal since we know they are
339 # empty
340 butler.registry.removeCollection(this_run)
341 expected_collections.remove(this_run)
343 # Put the dataset again, since the last thing we did was remove it
344 # and we want to use the default collection.
345 ref = butler.put(metric, refIn)
347 # Get with parameters
348 stop = 4
349 sliced = butler.get(ref, parameters={"slice": slice(stop)})
350 self.assertNotEqual(metric, sliced)
351 self.assertEqual(metric.summary, sliced.summary)
352 self.assertEqual(metric.output, sliced.output)
353 self.assertEqual(metric.data[:stop], sliced.data)
354 # getDeferred with parameters
355 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
356 self.assertNotEqual(metric, sliced)
357 self.assertEqual(metric.summary, sliced.summary)
358 self.assertEqual(metric.output, sliced.output)
359 self.assertEqual(metric.data[:stop], sliced.data)
360 # getDeferred with deferred parameters
361 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
362 self.assertNotEqual(metric, sliced)
363 self.assertEqual(metric.summary, sliced.summary)
364 self.assertEqual(metric.output, sliced.output)
365 self.assertEqual(metric.data[:stop], sliced.data)
367 if storageClass.isComposite():
368 # Check that components can be retrieved
369 metricOut = butler.get(ref.datasetType.name, dataId)
370 compNameS = ref.datasetType.componentTypeName("summary")
371 compNameD = ref.datasetType.componentTypeName("data")
372 summary = butler.get(compNameS, dataId)
373 self.assertEqual(summary, metric.summary)
374 data = butler.get(compNameD, dataId)
375 self.assertEqual(data, metric.data)
377 if "counter" in storageClass.derivedComponents:
378 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
379 self.assertEqual(count, len(data))
381 count = butler.get(
382 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
383 )
384 self.assertEqual(count, stop)
386 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections)
387 summary = butler.getDirect(compRef)
388 self.assertEqual(summary, metric.summary)
390 # Create a Dataset type that has the same name but is inconsistent.
391 inconsistentDatasetType = DatasetType(
392 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
393 )
395 # Getting with a dataset type that does not match registry fails
396 with self.assertRaises(ValueError):
397 butler.get(inconsistentDatasetType, dataId)
399 # Combining a DatasetRef with a dataId should fail
400 with self.assertRaises(ValueError):
401 butler.get(ref, dataId)
402 # Getting with an explicit ref should fail if the id doesn't match
403 with self.assertRaises(ValueError):
404 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=101))
406 # Getting a dataset with unknown parameters should fail
407 with self.assertRaises(KeyError):
408 butler.get(ref, parameters={"unsupported": True})
410 # Check we have a collection
411 collections = set(butler.registry.queryCollections())
412 self.assertEqual(collections, expected_collections)
414 # Clean up to check that we can remove something that may have
415 # already had a component removed
416 butler.pruneDatasets([ref], unstore=True, purge=True)
418 # Check that we can configure a butler to accept a put even
419 # if it already has the dataset in registry.
420 ref = butler.put(metric, refIn)
422 # Repeat put will fail.
423 with self.assertRaises(ConflictingDefinitionError):
424 butler.put(metric, refIn)
426 # Remove the datastore entry.
427 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
429 # Put will still fail
430 with self.assertRaises(ConflictingDefinitionError):
431 butler.put(metric, refIn)
433 # Allow the put to succeed
434 butler._allow_put_of_predefined_dataset = True
435 ref2 = butler.put(metric, refIn)
436 self.assertEqual(ref2.id, ref.id)
438 # A second put will still fail but with a different exception
439 # than before.
440 with self.assertRaises(ConflictingDefinitionError):
441 butler.put(metric, refIn)
443 # Reset the flag to avoid confusion
444 butler._allow_put_of_predefined_dataset = False
446 # Leave the dataset in place since some downstream tests require
447 # something to be present
449 return butler
451 def testDeferredCollectionPassing(self):
452 # Construct a butler with no run or collection, but make it writeable.
453 butler = Butler(self.tmpConfigFile, writeable=True)
454 # Create and register a DatasetType
455 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
456 datasetType = self.addDatasetType(
457 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
458 )
459 # Add needed Dimensions
460 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
461 butler.registry.insertDimensionData(
462 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
463 )
464 butler.registry.insertDimensionData(
465 "visit",
466 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
467 )
468 dataId = {"instrument": "DummyCamComp", "visit": 423}
469 # Create dataset.
470 metric = makeExampleMetrics()
471 # Register a new run and put dataset.
472 run = "deferred"
473 self.assertTrue(butler.registry.registerRun(run))
474 # Second time it will be allowed but indicate no-op
475 self.assertFalse(butler.registry.registerRun(run))
476 ref = butler.put(metric, datasetType, dataId, run=run)
477 # Putting with no run should fail with TypeError.
478 with self.assertRaises(CollectionError):
479 butler.put(metric, datasetType, dataId)
480 # Dataset should exist.
481 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
482 # We should be able to get the dataset back, but with and without
483 # a deferred dataset handle.
484 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
485 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
486 # Trying to find the dataset without any collection is a TypeError.
487 with self.assertRaises(CollectionError):
488 butler.datasetExists(datasetType, dataId)
489 with self.assertRaises(CollectionError):
490 butler.get(datasetType, dataId)
491 # Associate the dataset with a different collection.
492 butler.registry.registerCollection("tagged")
493 butler.registry.associate("tagged", [ref])
494 # Deleting the dataset from the new collection should make it findable
495 # in the original collection.
496 butler.pruneDatasets([ref], tags=["tagged"])
497 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
500class ButlerTests(ButlerPutGetTests):
501 """Tests for Butler."""
503 useTempRoot = True
505 def setUp(self):
506 """Create a new butler root for each test."""
507 self.root = makeTestTempDir(TESTDIR)
508 Butler.makeRepo(self.root, config=Config(self.configFile))
509 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
511 def testConstructor(self):
512 """Independent test of constructor."""
513 butler = Butler(self.tmpConfigFile, run=self.default_run)
514 self.assertIsInstance(butler, Butler)
516 # Check that butler.yaml is added automatically.
517 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
518 config_dir = self.tmpConfigFile[: -len(end)]
519 butler = Butler(config_dir, run=self.default_run)
520 self.assertIsInstance(butler, Butler)
522 # Even with a ResourcePath.
523 butler = Butler(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
524 self.assertIsInstance(butler, Butler)
526 collections = set(butler.registry.queryCollections())
527 self.assertEqual(collections, {self.default_run})
529 # Check that some special characters can be included in run name.
530 special_run = "u@b.c-A"
531 butler_special = Butler(butler=butler, run=special_run)
532 collections = set(butler_special.registry.queryCollections("*@*"))
533 self.assertEqual(collections, {special_run})
535 butler2 = Butler(butler=butler, collections=["other"])
536 self.assertEqual(butler2.collections, ("other",))
537 self.assertIsNone(butler2.run)
538 self.assertIs(butler.datastore, butler2.datastore)
540 # Test that we can use an environment variable to find this
541 # repository.
542 butler_index = Config()
543 butler_index["label"] = self.tmpConfigFile
544 for suffix in (".yaml", ".json"):
545 # Ensure that the content differs so that we know that
546 # we aren't reusing the cache.
547 bad_label = f"s3://bucket/not_real{suffix}"
548 butler_index["bad_label"] = bad_label
549 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
550 butler_index.dumpToUri(temp_file)
551 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
552 self.assertEqual(Butler.get_known_repos(), set(("label", "bad_label")))
553 uri = Butler.get_repo_uri("bad_label")
554 self.assertEqual(uri, ResourcePath(bad_label))
555 uri = Butler.get_repo_uri("label")
556 butler = Butler(uri, writeable=False)
557 self.assertIsInstance(butler, Butler)
558 butler = Butler("label", writeable=False)
559 self.assertIsInstance(butler, Butler)
560 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
561 Butler("not_there", writeable=False)
562 with self.assertRaises(KeyError) as cm:
563 Butler.get_repo_uri("missing")
564 self.assertIn("not known to", str(cm.exception))
565 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
566 with self.assertRaises(FileNotFoundError):
567 Butler.get_repo_uri("label")
568 self.assertEqual(Butler.get_known_repos(), set())
569 with self.assertRaises(KeyError) as cm:
570 # No environment variable set.
571 Butler.get_repo_uri("label")
572 self.assertIn("No repository index defined", str(cm.exception))
573 with self.assertRaisesRegex(FileNotFoundError, "no known aliases"):
574 # No aliases registered.
575 Butler("not_there")
576 self.assertEqual(Butler.get_known_repos(), set())
578 def testBasicPutGet(self):
579 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
580 self.runPutGetTest(storageClass, "test_metric")
582 def testCompositePutGetConcrete(self):
583 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
584 butler = self.runPutGetTest(storageClass, "test_metric")
586 # Should *not* be disassembled
587 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
588 self.assertEqual(len(datasets), 1)
589 uri, components = butler.getURIs(datasets[0])
590 self.assertIsInstance(uri, ResourcePath)
591 self.assertFalse(components)
592 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
593 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
595 # Predicted dataset
596 dataId = {"instrument": "DummyCamComp", "visit": 424}
597 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
598 self.assertFalse(components)
599 self.assertIsInstance(uri, ResourcePath)
600 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
601 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
603 def testCompositePutGetVirtual(self):
604 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
605 butler = self.runPutGetTest(storageClass, "test_metric_comp")
607 # Should be disassembled
608 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
609 self.assertEqual(len(datasets), 1)
610 uri, components = butler.getURIs(datasets[0])
612 if butler.datastore.isEphemeral:
613 # Never disassemble in-memory datastore
614 self.assertIsInstance(uri, ResourcePath)
615 self.assertFalse(components)
616 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
617 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
618 else:
619 self.assertIsNone(uri)
620 self.assertEqual(set(components), set(storageClass.components))
621 for compuri in components.values():
622 self.assertIsInstance(compuri, ResourcePath)
623 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
624 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
626 # Predicted dataset
627 dataId = {"instrument": "DummyCamComp", "visit": 424}
628 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
630 if butler.datastore.isEphemeral:
631 # Never disassembled
632 self.assertIsInstance(uri, ResourcePath)
633 self.assertFalse(components)
634 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
635 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
636 else:
637 self.assertIsNone(uri)
638 self.assertEqual(set(components), set(storageClass.components))
639 for compuri in components.values():
640 self.assertIsInstance(compuri, ResourcePath)
641 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
642 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
644 def testStorageClassOverrideGet(self):
645 """Test storage class conversion on get with override."""
646 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
647 datasetTypeName = "anything"
648 run = self.default_run
650 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
652 # Create and store a dataset.
653 metric = makeExampleMetrics()
654 dataId = {"instrument": "DummyCamComp", "visit": 423}
656 ref = butler.put(metric, datasetType, dataId)
658 # Return native type.
659 retrieved = butler.get(ref)
660 self.assertEqual(retrieved, metric)
662 # Specify an override.
663 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
664 model = butler.getDirect(ref, storageClass=new_sc)
665 self.assertNotEqual(type(model), type(retrieved))
666 self.assertIs(type(model), new_sc.pytype)
667 self.assertEqual(retrieved, model)
669 # Defer but override later.
670 deferred = butler.getDirectDeferred(ref)
671 model = deferred.get(storageClass=new_sc)
672 self.assertIs(type(model), new_sc.pytype)
673 self.assertEqual(retrieved, model)
675 # Defer but override up front.
676 deferred = butler.getDirectDeferred(ref, storageClass=new_sc)
677 model = deferred.get()
678 self.assertIs(type(model), new_sc.pytype)
679 self.assertEqual(retrieved, model)
681 # Retrieve a component. Should be a tuple.
682 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
683 self.assertIs(type(data), tuple)
684 self.assertEqual(data, tuple(retrieved.data))
686 # Parameter on the write storage class should work regardless
687 # of read storage class.
688 data = butler.get(
689 "anything.data",
690 dataId,
691 storageClass="StructuredDataDataTestTuple",
692 parameters={"slice": slice(2, 4)},
693 )
694 self.assertEqual(len(data), 2)
696 # Try a parameter that is known to the read storage class but not
697 # the write storage class.
698 with self.assertRaises(KeyError):
699 butler.get(
700 "anything.data",
701 dataId,
702 storageClass="StructuredDataDataTestTuple",
703 parameters={"xslice": slice(2, 4)},
704 )
706 def testPytypePutCoercion(self):
707 """Test python type coercion on Butler.get and put."""
709 # Store some data with the normal example storage class.
710 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
711 datasetTypeName = "test_metric"
712 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
714 dataId = {"instrument": "DummyCamComp", "visit": 423}
716 # Put a dict and this should coerce to a MetricsExample
717 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
718 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
719 test_metric = butler.getDirect(metric_ref)
720 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
721 self.assertEqual(test_metric.summary, test_dict["summary"])
722 self.assertEqual(test_metric.output, test_dict["output"])
724 # Check that the put still works if a DatasetType is given with
725 # a definition matching this python type.
726 registry_type = butler.registry.getDatasetType(datasetTypeName)
727 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
728 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
729 self.assertEqual(metric2_ref.datasetType, registry_type)
731 # The get will return the type expected by registry.
732 test_metric2 = butler.getDirect(metric2_ref)
733 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
735 # Make a new DatasetRef with the compatible but different DatasetType.
736 # This should now return a dict.
737 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
738 test_dict2 = butler.getDirect(new_ref)
739 self.assertEqual(get_full_type_name(test_dict2), "dict")
741 # Get it again with the wrong dataset type definition using get()
742 # rather than getDirect(). This should be consistent with getDirect()
743 # behavior and return the type of the DatasetType.
744 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
745 self.assertEqual(get_full_type_name(test_dict3), "dict")
747 def testIngest(self):
748 butler = Butler(self.tmpConfigFile, run=self.default_run)
750 # Create and register a DatasetType
751 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "detector"])
753 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
754 datasetTypeName = "metric"
756 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
758 # Add needed Dimensions
759 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
760 butler.registry.insertDimensionData(
761 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
762 )
763 for detector in (1, 2):
764 butler.registry.insertDimensionData(
765 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
766 )
768 butler.registry.insertDimensionData(
769 "visit",
770 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
771 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
772 )
774 formatter = doImport("lsst.daf.butler.formatters.yaml.YamlFormatter")
775 dataRoot = os.path.join(TESTDIR, "data", "basic")
776 datasets = []
777 for detector in (1, 2):
778 detector_name = f"detector_{detector}"
779 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
780 dataId = {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
781 # Create a DatasetRef for ingest
782 refIn = DatasetRef(datasetType, dataId, id=None)
784 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
786 butler.ingest(*datasets, transfer="copy")
788 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
789 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
791 metrics1 = butler.get(datasetTypeName, dataId1)
792 metrics2 = butler.get(datasetTypeName, dataId2)
793 self.assertNotEqual(metrics1, metrics2)
795 # Compare URIs
796 uri1 = butler.getURI(datasetTypeName, dataId1)
797 uri2 = butler.getURI(datasetTypeName, dataId2)
798 self.assertNotEqual(uri1, uri2)
800 # Now do a multi-dataset but single file ingest
801 metricFile = os.path.join(dataRoot, "detectors.yaml")
802 refs = []
803 for detector in (1, 2):
804 detector_name = f"detector_{detector}"
805 dataId = {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
806 # Create a DatasetRef for ingest
807 refs.append(DatasetRef(datasetType, dataId, id=None))
809 # Test "move" transfer to ensure that the files themselves
810 # have disappeared following ingest.
811 with ResourcePath.temporary_uri(suffix=".yaml") as tempFile:
812 tempFile.transfer_from(ResourcePath(metricFile), transfer="copy")
814 datasets = []
815 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter))
817 butler.ingest(*datasets, transfer="move", record_validation_info=False)
818 self.assertFalse(tempFile.exists())
820 # Check that the datastore recorded no file size.
821 # Not all datastores can support this.
822 try:
823 infos = butler.datastore.getStoredItemsInfo(datasets[0].refs[0])
824 self.assertEqual(infos[0].file_size, -1)
825 except AttributeError:
826 pass
828 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
829 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
831 multi1 = butler.get(datasetTypeName, dataId1)
832 multi2 = butler.get(datasetTypeName, dataId2)
834 self.assertEqual(multi1, metrics1)
835 self.assertEqual(multi2, metrics2)
837 # Compare URIs
838 uri1 = butler.getURI(datasetTypeName, dataId1)
839 uri2 = butler.getURI(datasetTypeName, dataId2)
840 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
842 # Test that removing one does not break the second
843 # This line will issue a warning log message for a ChainedDatastore
844 # that uses an InMemoryDatastore since in-memory can not ingest
845 # files.
846 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
847 self.assertFalse(butler.datasetExists(datasetTypeName, dataId1))
848 self.assertTrue(butler.datasetExists(datasetTypeName, dataId2))
849 multi2b = butler.get(datasetTypeName, dataId2)
850 self.assertEqual(multi2, multi2b)
852 def testPruneCollections(self):
853 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
854 butler = Butler(self.tmpConfigFile, writeable=True)
855 # Load registry data with dimensions to hang datasets off of.
856 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
857 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
858 # Add some RUN-type collections.
859 run1 = "run1"
860 butler.registry.registerRun(run1)
861 run2 = "run2"
862 butler.registry.registerRun(run2)
863 # put some datasets. ref1 and ref2 have the same data ID, and are in
864 # different runs. ref3 has a different data ID.
865 metric = makeExampleMetrics()
866 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
867 datasetType = self.addDatasetType(
868 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
869 )
870 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
871 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
872 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
874 # Try to delete a RUN collection without purge, or with purge and not
875 # unstore.
876 with self.assertRaises(TypeError):
877 butler.pruneCollection(run1)
878 with self.assertRaises(TypeError):
879 butler.pruneCollection(run2, purge=True)
880 # Add a TAGGED collection and associate ref3 only into it.
881 tag1 = "tag1"
882 registered = butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
883 self.assertTrue(registered)
884 # Registering a second time should be allowed.
885 registered = butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
886 self.assertFalse(registered)
887 butler.registry.associate(tag1, [ref3])
888 # Add a CHAINED collection that searches run1 and then run2. It
889 # logically contains only ref1, because ref2 is shadowed due to them
890 # having the same data ID and dataset type.
891 chain1 = "chain1"
892 butler.registry.registerCollection(chain1, type=CollectionType.CHAINED)
893 butler.registry.setCollectionChain(chain1, [run1, run2])
894 # Try to delete RUN collections, which should fail with complete
895 # rollback because they're still referenced by the CHAINED
896 # collection.
897 with self.assertRaises(sqlalchemy.exc.IntegrityError):
898 butler.pruneCollection(run1, purge=True, unstore=True)
899 with self.assertRaises(sqlalchemy.exc.IntegrityError):
900 butler.pruneCollection(run2, purge=True, unstore=True)
901 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
902 existence = butler.datastore.mexists([ref1, ref2, ref3])
903 self.assertTrue(existence[ref1])
904 self.assertTrue(existence[ref2])
905 self.assertTrue(existence[ref3])
906 # Try to delete CHAINED and TAGGED collections with purge; should not
907 # work.
908 with self.assertRaises(TypeError):
909 butler.pruneCollection(tag1, purge=True, unstore=True)
910 with self.assertRaises(TypeError):
911 butler.pruneCollection(chain1, purge=True, unstore=True)
912 # Remove the tagged collection with unstore=False. This should not
913 # affect the datasets.
914 butler.pruneCollection(tag1)
915 with self.assertRaises(MissingCollectionError):
916 butler.registry.getCollectionType(tag1)
917 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
918 existence = butler.datastore.mexists([ref1, ref2, ref3])
919 self.assertTrue(existence[ref1])
920 self.assertTrue(existence[ref2])
921 self.assertTrue(existence[ref3])
922 # Add the tagged collection back in, and remove it with unstore=True.
923 # This should remove ref3 only from the datastore.
924 butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
925 butler.registry.associate(tag1, [ref3])
926 butler.pruneCollection(tag1, unstore=True)
927 with self.assertRaises(MissingCollectionError):
928 butler.registry.getCollectionType(tag1)
929 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
930 existence = butler.datastore.mexists([ref1, ref2, ref3])
931 self.assertTrue(existence[ref1])
932 self.assertTrue(existence[ref2])
933 self.assertFalse(existence[ref3])
934 # Delete the chain with unstore=False. The datasets should not be
935 # affected at all.
936 butler.pruneCollection(chain1)
937 with self.assertRaises(MissingCollectionError):
938 butler.registry.getCollectionType(chain1)
939 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
940 existence = butler.datastore.mexists([ref1, ref2, ref3])
941 self.assertTrue(existence[ref1])
942 self.assertTrue(existence[ref2])
943 self.assertFalse(existence[ref3])
944 existence = butler.datastore.knows_these([ref1, ref2, ref3])
945 self.assertTrue(existence[ref1])
946 self.assertTrue(existence[ref2])
947 self.assertFalse(existence[ref3])
948 # Redefine and then delete the chain with unstore=True. Only ref1
949 # should be unstored (ref3 has already been unstored, but otherwise
950 # would be now).
951 butler.registry.registerCollection(chain1, type=CollectionType.CHAINED)
952 butler.registry.setCollectionChain(chain1, [run1, run2])
953 butler.pruneCollection(chain1, unstore=True)
954 with self.assertRaises(MissingCollectionError):
955 butler.registry.getCollectionType(chain1)
956 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
957 existence = butler.datastore.mexists([ref1, ref2, ref3])
958 self.assertFalse(existence[ref1])
959 self.assertTrue(existence[ref2])
960 self.assertFalse(existence[ref3])
961 # Remove run1. This removes ref1 and ref3 from the registry (they're
962 # already gone from the datastore, which is fine).
963 butler.pruneCollection(run1, purge=True, unstore=True)
964 with self.assertRaises(MissingCollectionError):
965 butler.registry.getCollectionType(run1)
966 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref2])
967 self.assertTrue(butler.datastore.exists(ref2))
968 self.assertTrue(butler.datastore.knows(ref2))
969 # Remove run2. This removes ref2 from the registry and the datastore.
970 butler.pruneCollection(run2, purge=True, unstore=True)
971 with self.assertRaises(MissingCollectionError):
972 butler.registry.getCollectionType(run2)
973 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [])
975 # Now that the collections have been pruned we can remove the
976 # dataset type
977 butler.registry.removeDatasetType(datasetType.name)
979 def testPickle(self):
980 """Test pickle support."""
981 butler = Butler(self.tmpConfigFile, run=self.default_run)
982 butlerOut = pickle.loads(pickle.dumps(butler))
983 self.assertIsInstance(butlerOut, Butler)
984 self.assertEqual(butlerOut._config, butler._config)
985 self.assertEqual(butlerOut.collections, butler.collections)
986 self.assertEqual(butlerOut.run, butler.run)
988 def testGetDatasetTypes(self):
989 butler = Butler(self.tmpConfigFile, run=self.default_run)
990 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "physical_filter"])
991 dimensionEntries = [
992 (
993 "instrument",
994 {"instrument": "DummyCam"},
995 {"instrument": "DummyHSC"},
996 {"instrument": "DummyCamComp"},
997 ),
998 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
999 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1000 ]
1001 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1002 # Add needed Dimensions
1003 for args in dimensionEntries:
1004 butler.registry.insertDimensionData(*args)
1006 # When a DatasetType is added to the registry entries are not created
1007 # for components but querying them can return the components.
1008 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
1009 components = set()
1010 for datasetTypeName in datasetTypeNames:
1011 # Create and register a DatasetType
1012 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1014 for componentName in storageClass.components:
1015 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
1017 fromRegistry: set[DatasetType] = set()
1018 for parent_dataset_type in butler.registry.queryDatasetTypes():
1019 fromRegistry.add(parent_dataset_type)
1020 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
1021 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
1023 # Now that we have some dataset types registered, validate them
1024 butler.validateConfiguration(
1025 ignore=[
1026 "test_metric_comp",
1027 "metric3",
1028 "metric5",
1029 "calexp",
1030 "DummySC",
1031 "datasetType.component",
1032 "random_data",
1033 "random_data_2",
1034 ]
1035 )
1037 # Add a new datasetType that will fail template validation
1038 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
1039 if self.validationCanFail:
1040 with self.assertRaises(ValidationError):
1041 butler.validateConfiguration()
1043 # Rerun validation but with a subset of dataset type names
1044 butler.validateConfiguration(datasetTypeNames=["metric4"])
1046 # Rerun validation but ignore the bad datasetType
1047 butler.validateConfiguration(
1048 ignore=[
1049 "test_metric_comp",
1050 "metric3",
1051 "metric5",
1052 "calexp",
1053 "DummySC",
1054 "datasetType.component",
1055 "random_data",
1056 "random_data_2",
1057 ]
1058 )
1060 def testTransaction(self):
1061 butler = Butler(self.tmpConfigFile, run=self.default_run)
1062 datasetTypeName = "test_metric"
1063 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1064 dimensionEntries = (
1065 ("instrument", {"instrument": "DummyCam"}),
1066 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1067 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1068 )
1069 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1070 metric = makeExampleMetrics()
1071 dataId = {"instrument": "DummyCam", "visit": 42}
1072 # Create and register a DatasetType
1073 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1074 with self.assertRaises(TransactionTestError):
1075 with butler.transaction():
1076 # Add needed Dimensions
1077 for args in dimensionEntries:
1078 butler.registry.insertDimensionData(*args)
1079 # Store a dataset
1080 ref = butler.put(metric, datasetTypeName, dataId)
1081 self.assertIsInstance(ref, DatasetRef)
1082 # Test getDirect
1083 metricOut = butler.getDirect(ref)
1084 self.assertEqual(metric, metricOut)
1085 # Test get
1086 metricOut = butler.get(datasetTypeName, dataId)
1087 self.assertEqual(metric, metricOut)
1088 # Check we can get components
1089 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1090 raise TransactionTestError("This should roll back the entire transaction")
1091 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1092 butler.registry.expandDataId(dataId)
1093 # Should raise LookupError for missing data ID value
1094 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1095 butler.get(datasetTypeName, dataId)
1096 # Also check explicitly if Dataset entry is missing
1097 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections))
1098 # Direct retrieval should not find the file in the Datastore
1099 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1100 butler.getDirect(ref)
1102 def testMakeRepo(self):
1103 """Test that we can write butler configuration to a new repository via
1104 the Butler.makeRepo interface and then instantiate a butler from the
1105 repo root.
1106 """
1107 # Do not run the test if we know this datastore configuration does
1108 # not support a file system root
1109 if self.fullConfigKey is None:
1110 return
1112 # create two separate directories
1113 root1 = tempfile.mkdtemp(dir=self.root)
1114 root2 = tempfile.mkdtemp(dir=self.root)
1116 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1117 limited = Config(self.configFile)
1118 butler1 = Butler(butlerConfig)
1119 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1120 full = Config(self.tmpConfigFile)
1121 butler2 = Butler(butlerConfig)
1122 # Butlers should have the same configuration regardless of whether
1123 # defaults were expanded.
1124 self.assertEqual(butler1._config, butler2._config)
1125 # Config files loaded directly should not be the same.
1126 self.assertNotEqual(limited, full)
1127 # Make sure "limited" doesn't have a few keys we know it should be
1128 # inheriting from defaults.
1129 self.assertIn(self.fullConfigKey, full)
1130 self.assertNotIn(self.fullConfigKey, limited)
1132 # Collections don't appear until something is put in them
1133 collections1 = set(butler1.registry.queryCollections())
1134 self.assertEqual(collections1, set())
1135 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1137 # Check that a config with no associated file name will not
1138 # work properly with relocatable Butler repo
1139 butlerConfig.configFile = None
1140 with self.assertRaises(ValueError):
1141 Butler(butlerConfig)
1143 with self.assertRaises(FileExistsError):
1144 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1146 def testStringification(self):
1147 butler = Butler(self.tmpConfigFile, run=self.default_run)
1148 butlerStr = str(butler)
1150 if self.datastoreStr is not None:
1151 for testStr in self.datastoreStr:
1152 self.assertIn(testStr, butlerStr)
1153 if self.registryStr is not None:
1154 self.assertIn(self.registryStr, butlerStr)
1156 datastoreName = butler.datastore.name
1157 if self.datastoreName is not None:
1158 for testStr in self.datastoreName:
1159 self.assertIn(testStr, datastoreName)
1161 def testButlerRewriteDataId(self):
1162 """Test that dataIds can be rewritten based on dimension records."""
1164 butler = Butler(self.tmpConfigFile, run=self.default_run)
1166 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1167 datasetTypeName = "random_data"
1169 # Create dimension records.
1170 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1171 butler.registry.insertDimensionData(
1172 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1173 )
1174 butler.registry.insertDimensionData(
1175 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1176 )
1178 dimensions = butler.registry.dimensions.extract(["instrument", "exposure"])
1179 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1180 butler.registry.registerDatasetType(datasetType)
1182 n_exposures = 5
1183 dayobs = 20210530
1185 for i in range(n_exposures):
1186 butler.registry.insertDimensionData(
1187 "exposure",
1188 {
1189 "instrument": "DummyCamComp",
1190 "id": i,
1191 "obs_id": f"exp{i}",
1192 "seq_num": i,
1193 "day_obs": dayobs,
1194 "physical_filter": "d-r",
1195 },
1196 )
1198 # Write some data.
1199 for i in range(n_exposures):
1200 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1202 # Use the seq_num for the put to test rewriting.
1203 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1204 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1206 # Check that the exposure is correct in the dataId
1207 self.assertEqual(ref.dataId["exposure"], i)
1209 # and check that we can get the dataset back with the same dataId
1210 new_metric = butler.get(datasetTypeName, dataId=dataId)
1211 self.assertEqual(new_metric, metric)
1214class FileDatastoreButlerTests(ButlerTests):
1215 """Common tests and specialization of ButlerTests for butlers backed
1216 by datastores that inherit from FileDatastore.
1217 """
1219 def checkFileExists(self, root, relpath):
1220 """Checks if file exists at a given path (relative to root).
1222 Test testPutTemplates verifies actual physical existance of the files
1223 in the requested location.
1224 """
1225 uri = ResourcePath(root, forceDirectory=True)
1226 return uri.join(relpath).exists()
1228 def testPutTemplates(self):
1229 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1230 butler = Butler(self.tmpConfigFile, run=self.default_run)
1232 # Add needed Dimensions
1233 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1234 butler.registry.insertDimensionData(
1235 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1236 )
1237 butler.registry.insertDimensionData(
1238 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1239 )
1240 butler.registry.insertDimensionData(
1241 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1242 )
1244 # Create and store a dataset
1245 metric = makeExampleMetrics()
1247 # Create two almost-identical DatasetTypes (both will use default
1248 # template)
1249 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1250 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1251 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1252 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1254 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1255 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1257 # Put with exactly the data ID keys needed
1258 ref = butler.put(metric, "metric1", dataId1)
1259 uri = butler.getURI(ref)
1260 self.assertTrue(uri.exists())
1261 self.assertTrue(
1262 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle")
1263 )
1265 # Check the template based on dimensions
1266 if hasattr(butler.datastore, "templates"):
1267 butler.datastore.templates.validateTemplates([ref])
1269 # Put with extra data ID keys (physical_filter is an optional
1270 # dependency); should not change template (at least the way we're
1271 # defining them to behave now; the important thing is that they
1272 # must be consistent).
1273 ref = butler.put(metric, "metric2", dataId2)
1274 uri = butler.getURI(ref)
1275 self.assertTrue(uri.exists())
1276 self.assertTrue(
1277 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle")
1278 )
1280 # Check the template based on dimensions
1281 if hasattr(butler.datastore, "templates"):
1282 butler.datastore.templates.validateTemplates([ref])
1284 # Use a template that has a typo in dimension record metadata.
1285 # Easier to test with a butler that has a ref with records attached.
1286 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1287 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1288 path = template.format(ref)
1289 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1291 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1292 with self.assertRaises(KeyError):
1293 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1294 template.format(ref)
1296 # Now use a file template that will not result in unique filenames
1297 with self.assertRaises(FileTemplateValidationError):
1298 butler.put(metric, "metric3", dataId1)
1300 def testImportExport(self):
1301 # Run put/get tests just to create and populate a repo.
1302 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1303 self.runImportExportTest(storageClass)
1305 @unittest.expectedFailure
1306 def testImportExportVirtualComposite(self):
1307 # Run put/get tests just to create and populate a repo.
1308 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1309 self.runImportExportTest(storageClass)
1311 def runImportExportTest(self, storageClass):
1312 """This test does an export to a temp directory and an import back
1313 into a new temp directory repo. It does not assume a posix datastore"""
1314 exportButler = self.runPutGetTest(storageClass, "test_metric")
1315 # Test that the repo actually has at least one dataset.
1316 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1317 self.assertGreater(len(datasets), 0)
1318 # Add a DimensionRecord that's unused by those datasets.
1319 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1320 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1321 # Export and then import datasets.
1322 with safeTestTempDir(TESTDIR) as exportDir:
1323 exportFile = os.path.join(exportDir, "exports.yaml")
1324 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1325 export.saveDatasets(datasets)
1326 # Export the same datasets again. This should quietly do
1327 # nothing because of internal deduplication, and it shouldn't
1328 # complain about being asked to export the "htm7" elements even
1329 # though there aren't any in these datasets or in the database.
1330 export.saveDatasets(datasets, elements=["htm7"])
1331 # Save one of the data IDs again; this should be harmless
1332 # because of internal deduplication.
1333 export.saveDataIds([datasets[0].dataId])
1334 # Save some dimension records directly.
1335 export.saveDimensionData("skymap", [skymapRecord])
1336 self.assertTrue(os.path.exists(exportFile))
1337 with safeTestTempDir(TESTDIR) as importDir:
1338 # We always want this to be a local posix butler
1339 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1340 # Calling script.butlerImport tests the implementation of the
1341 # butler command line interface "import" subcommand. Functions
1342 # in the script folder are generally considered protected and
1343 # should not be used as public api.
1344 with open(exportFile, "r") as f:
1345 script.butlerImport(
1346 importDir,
1347 export_file=f,
1348 directory=exportDir,
1349 transfer="auto",
1350 skip_dimensions=None,
1351 reuse_ids=False,
1352 )
1353 importButler = Butler(importDir, run=self.default_run)
1354 for ref in datasets:
1355 with self.subTest(ref=ref):
1356 # Test for existence by passing in the DatasetType and
1357 # data ID separately, to avoid lookup by dataset_id.
1358 self.assertTrue(importButler.datasetExists(ref.datasetType, ref.dataId))
1359 self.assertEqual(
1360 list(importButler.registry.queryDimensionRecords("skymap")),
1361 [importButler.registry.dimensions["skymap"].RecordClass(**skymapRecord)],
1362 )
1364 def testRemoveRuns(self):
1365 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1366 butler = Butler(self.tmpConfigFile, writeable=True)
1367 # Load registry data with dimensions to hang datasets off of.
1368 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1369 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1370 # Add some RUN-type collection.
1371 run1 = "run1"
1372 butler.registry.registerRun(run1)
1373 run2 = "run2"
1374 butler.registry.registerRun(run2)
1375 # put a dataset in each
1376 metric = makeExampleMetrics()
1377 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1378 datasetType = self.addDatasetType(
1379 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1380 )
1381 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1382 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1383 uri1 = butler.getURI(ref1, collections=[run1])
1384 uri2 = butler.getURI(ref2, collections=[run2])
1385 # Remove from both runs with different values for unstore.
1386 butler.removeRuns([run1], unstore=True)
1387 butler.removeRuns([run2], unstore=False)
1388 # Should be nothing in registry for either one, and datastore should
1389 # not think either exists.
1390 with self.assertRaises(MissingCollectionError):
1391 butler.registry.getCollectionType(run1)
1392 with self.assertRaises(MissingCollectionError):
1393 butler.registry.getCollectionType(run2)
1394 self.assertFalse(butler.datastore.exists(ref1))
1395 self.assertFalse(butler.datastore.exists(ref2))
1396 # The ref we unstored should be gone according to the URI, but the
1397 # one we forgot should still be around.
1398 self.assertFalse(uri1.exists())
1399 self.assertTrue(uri2.exists())
1402class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1403 """PosixDatastore specialization of a butler"""
1405 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1406 fullConfigKey = ".datastore.formatters"
1407 validationCanFail = True
1408 datastoreStr = ["/tmp"]
1409 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1410 registryStr = "/gen3.sqlite3"
1412 def testPathConstructor(self):
1413 """Independent test of constructor using PathLike."""
1414 butler = Butler(self.tmpConfigFile, run=self.default_run)
1415 self.assertIsInstance(butler, Butler)
1417 # And again with a Path object with the butler yaml
1418 path = pathlib.Path(self.tmpConfigFile)
1419 butler = Butler(path, writeable=False)
1420 self.assertIsInstance(butler, Butler)
1422 # And again with a Path object without the butler yaml
1423 # (making sure we skip it if the tmp config doesn't end
1424 # in butler.yaml -- which is the case for a subclass)
1425 if self.tmpConfigFile.endswith("butler.yaml"):
1426 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1427 butler = Butler(path, writeable=False)
1428 self.assertIsInstance(butler, Butler)
1430 def testExportTransferCopy(self):
1431 """Test local export using all transfer modes"""
1432 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1433 exportButler = self.runPutGetTest(storageClass, "test_metric")
1434 # Test that the repo actually has at least one dataset.
1435 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1436 self.assertGreater(len(datasets), 0)
1437 uris = [exportButler.getURI(d) for d in datasets]
1438 datastoreRoot = exportButler.datastore.root
1440 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1442 for path in pathsInStore:
1443 # Assume local file system
1444 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1446 for transfer in ("copy", "link", "symlink", "relsymlink"):
1447 with safeTestTempDir(TESTDIR) as exportDir:
1448 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1449 export.saveDatasets(datasets)
1450 for path in pathsInStore:
1451 self.assertTrue(
1452 self.checkFileExists(exportDir, path),
1453 f"Check that mode {transfer} exported files",
1454 )
1456 def testPruneDatasets(self):
1457 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1458 butler = Butler(self.tmpConfigFile, writeable=True)
1459 # Load registry data with dimensions to hang datasets off of.
1460 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1461 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1462 # Add some RUN-type collections.
1463 run1 = "run1"
1464 butler.registry.registerRun(run1)
1465 run2 = "run2"
1466 butler.registry.registerRun(run2)
1467 # put some datasets. ref1 and ref2 have the same data ID, and are in
1468 # different runs. ref3 has a different data ID.
1469 metric = makeExampleMetrics()
1470 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1471 datasetType = self.addDatasetType(
1472 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1473 )
1474 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1475 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1476 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1478 # Simple prune.
1479 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1480 with self.assertRaises(LookupError):
1481 butler.datasetExists(ref1.datasetType, ref1.dataId, collections=run1)
1483 # Put data back.
1484 ref1 = butler.put(metric, ref1.unresolved(), run=run1)
1485 ref2 = butler.put(metric, ref2.unresolved(), run=run2)
1486 ref3 = butler.put(metric, ref3.unresolved(), run=run1)
1488 # Check that in normal mode, deleting the record will lead to
1489 # trash not touching the file.
1490 uri1 = butler.datastore.getURI(ref1)
1491 butler.datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1492 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref1.id})
1493 butler.datastore.trash(ref1)
1494 butler.datastore.emptyTrash()
1495 self.assertTrue(uri1.exists())
1496 uri1.remove() # Clean it up.
1498 # Simulate execution butler setup by deleting the datastore
1499 # record but keeping the file around and trusting.
1500 butler.datastore.trustGetRequest = True
1501 uri2 = butler.datastore.getURI(ref2)
1502 uri3 = butler.datastore.getURI(ref3)
1503 self.assertTrue(uri2.exists())
1504 self.assertTrue(uri3.exists())
1506 # Remove the datastore record.
1507 butler.datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1508 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref2.id})
1509 self.assertTrue(uri2.exists())
1510 butler.datastore.trash([ref2, ref3])
1511 # Immediate removal for ref2 file
1512 self.assertFalse(uri2.exists())
1513 # But ref3 has to wait for the empty.
1514 self.assertTrue(uri3.exists())
1515 butler.datastore.emptyTrash()
1516 self.assertFalse(uri3.exists())
1518 # Clear out the datasets from registry.
1519 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1521 def testPytypeCoercion(self):
1522 """Test python type coercion on Butler.get and put."""
1524 # Store some data with the normal example storage class.
1525 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1526 datasetTypeName = "test_metric"
1527 butler = self.runPutGetTest(storageClass, datasetTypeName)
1529 dataId = {"instrument": "DummyCamComp", "visit": 423}
1530 metric = butler.get(datasetTypeName, dataId=dataId)
1531 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1533 datasetType_ori = butler.registry.getDatasetType(datasetTypeName)
1534 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1536 # Now need to hack the registry dataset type definition.
1537 # There is no API for this.
1538 manager = butler.registry._managers.datasets
1539 manager._db.update(
1540 manager._static.dataset_type,
1541 {"name": datasetTypeName},
1542 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1543 )
1545 # Force reset of dataset type cache
1546 butler.registry.refresh()
1548 datasetType_new = butler.registry.getDatasetType(datasetTypeName)
1549 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1550 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1552 metric_model = butler.get(datasetTypeName, dataId=dataId)
1553 self.assertNotEqual(type(metric_model), type(metric))
1554 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1556 # Put the model and read it back to show that everything now
1557 # works as normal.
1558 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1559 metric_model_new = butler.get(metric_ref)
1560 self.assertEqual(metric_model_new, metric_model)
1562 # Hack the storage class again to something that will fail on the
1563 # get with no conversion class.
1564 manager._db.update(
1565 manager._static.dataset_type,
1566 {"name": datasetTypeName},
1567 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1568 )
1569 butler.registry.refresh()
1571 with self.assertRaises(ValueError):
1572 butler.get(datasetTypeName, dataId=dataId)
1575@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1576class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1577 """PosixDatastore specialization of a butler using Postgres"""
1579 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1580 fullConfigKey = ".datastore.formatters"
1581 validationCanFail = True
1582 datastoreStr = ["/tmp"]
1583 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1584 registryStr = "PostgreSQL@test"
1586 @staticmethod
1587 def _handler(postgresql):
1588 engine = sqlalchemy.engine.create_engine(postgresql.url())
1589 with engine.begin() as connection:
1590 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1592 @classmethod
1593 def setUpClass(cls):
1594 # Create the postgres test server.
1595 cls.postgresql = testing.postgresql.PostgresqlFactory(
1596 cache_initialized_db=True, on_initialized=cls._handler
1597 )
1598 super().setUpClass()
1600 @classmethod
1601 def tearDownClass(cls):
1602 # Clean up any lingering SQLAlchemy engines/connections
1603 # so they're closed before we shut down the server.
1604 gc.collect()
1605 cls.postgresql.clear_cache()
1606 super().tearDownClass()
1608 def setUp(self):
1609 self.server = self.postgresql()
1611 # Need to add a registry section to the config.
1612 self._temp_config = False
1613 config = Config(self.configFile)
1614 config["registry", "db"] = self.server.url()
1615 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1616 config.dump(fh)
1617 self.configFile = fh.name
1618 self._temp_config = True
1619 super().setUp()
1621 def tearDown(self):
1622 self.server.stop()
1623 if self._temp_config and os.path.exists(self.configFile):
1624 os.remove(self.configFile)
1625 super().tearDown()
1627 def testMakeRepo(self):
1628 # The base class test assumes that it's using sqlite and assumes
1629 # the config file is acceptable to sqlite.
1630 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1633class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1634 """InMemoryDatastore specialization of a butler"""
1636 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1637 fullConfigKey = None
1638 useTempRoot = False
1639 validationCanFail = False
1640 datastoreStr = ["datastore='InMemory"]
1641 datastoreName = ["InMemoryDatastore@"]
1642 registryStr = "/gen3.sqlite3"
1644 def testIngest(self):
1645 pass
1648class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1649 """PosixDatastore specialization"""
1651 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1652 fullConfigKey = ".datastore.datastores.1.formatters"
1653 validationCanFail = True
1654 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1655 datastoreName = [
1656 "InMemoryDatastore@",
1657 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1658 "SecondDatastore",
1659 ]
1660 registryStr = "/gen3.sqlite3"
1663class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1664 """Test that a yaml file in one location can refer to a root in another."""
1666 datastoreStr = ["dir1"]
1667 # Disable the makeRepo test since we are deliberately not using
1668 # butler.yaml as the config name.
1669 fullConfigKey = None
1671 def setUp(self):
1672 self.root = makeTestTempDir(TESTDIR)
1674 # Make a new repository in one place
1675 self.dir1 = os.path.join(self.root, "dir1")
1676 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1678 # Move the yaml file to a different place and add a "root"
1679 self.dir2 = os.path.join(self.root, "dir2")
1680 os.makedirs(self.dir2, exist_ok=True)
1681 configFile1 = os.path.join(self.dir1, "butler.yaml")
1682 config = Config(configFile1)
1683 config["root"] = self.dir1
1684 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1685 config.dumpToUri(configFile2)
1686 os.remove(configFile1)
1687 self.tmpConfigFile = configFile2
1689 def testFileLocations(self):
1690 self.assertNotEqual(self.dir1, self.dir2)
1691 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1692 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1693 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1696class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1697 """Test that a config file created by makeRepo outside of repo works."""
1699 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1701 def setUp(self):
1702 self.root = makeTestTempDir(TESTDIR)
1703 self.root2 = makeTestTempDir(TESTDIR)
1705 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1706 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1708 def tearDown(self):
1709 if os.path.exists(self.root2):
1710 shutil.rmtree(self.root2, ignore_errors=True)
1711 super().tearDown()
1713 def testConfigExistence(self):
1714 c = Config(self.tmpConfigFile)
1715 uri_config = ResourcePath(c["root"])
1716 uri_expected = ResourcePath(self.root, forceDirectory=True)
1717 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1718 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1720 def testPutGet(self):
1721 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1722 self.runPutGetTest(storageClass, "test_metric")
1725class ButlerMakeRepoOutfileDirTestCase(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):
1731 self.root = makeTestTempDir(TESTDIR)
1732 self.root2 = makeTestTempDir(TESTDIR)
1734 self.tmpConfigFile = self.root2
1735 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1737 def testConfigExistence(self):
1738 # Append the yaml file else Config constructor does not know the file
1739 # type.
1740 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1741 super().testConfigExistence()
1744class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1745 """Test that a config file created by makeRepo outside of repo works."""
1747 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1749 def setUp(self):
1750 self.root = makeTestTempDir(TESTDIR)
1751 self.root2 = makeTestTempDir(TESTDIR)
1753 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1754 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1757@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1758class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1759 """S3Datastore specialization of a butler; an S3 storage Datastore +
1760 a local in-memory SqlRegistry.
1761 """
1763 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1764 fullConfigKey = None
1765 validationCanFail = True
1767 bucketName = "anybucketname"
1768 """Name of the Bucket that will be used in the tests. The name is read from
1769 the config file used with the tests during set-up.
1770 """
1772 root = "butlerRoot/"
1773 """Root repository directory expected to be used in case useTempRoot=False.
1774 Otherwise the root is set to a 20 characters long randomly generated string
1775 during set-up.
1776 """
1778 datastoreStr = [f"datastore={root}"]
1779 """Contains all expected root locations in a format expected to be
1780 returned by Butler stringification.
1781 """
1783 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1784 """The expected format of the S3 Datastore string."""
1786 registryStr = "/gen3.sqlite3"
1787 """Expected format of the Registry string."""
1789 mock_s3 = mock_s3()
1790 """The mocked s3 interface from moto."""
1792 def genRoot(self):
1793 """Returns a random string of len 20 to serve as a root
1794 name for the temporary bucket repo.
1796 This is equivalent to tempfile.mkdtemp as this is what self.root
1797 becomes when useTempRoot is True.
1798 """
1799 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1800 return rndstr + "/"
1802 def setUp(self):
1803 config = Config(self.configFile)
1804 uri = ResourcePath(config[".datastore.datastore.root"])
1805 self.bucketName = uri.netloc
1807 # Enable S3 mocking of tests.
1808 self.mock_s3.start()
1810 # set up some fake credentials if they do not exist
1811 self.usingDummyCredentials = setAwsEnvCredentials()
1813 if self.useTempRoot:
1814 self.root = self.genRoot()
1815 rooturi = f"s3://{self.bucketName}/{self.root}"
1816 config.update({"datastore": {"datastore": {"root": rooturi}}})
1818 # need local folder to store registry database
1819 self.reg_dir = makeTestTempDir(TESTDIR)
1820 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
1822 # MOTO needs to know that we expect Bucket bucketname to exist
1823 # (this used to be the class attribute bucketName)
1824 s3 = boto3.resource("s3")
1825 s3.create_bucket(Bucket=self.bucketName)
1827 self.datastoreStr = f"datastore={self.root}"
1828 self.datastoreName = [f"FileDatastore@{rooturi}"]
1829 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
1830 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
1832 def tearDown(self):
1833 s3 = boto3.resource("s3")
1834 bucket = s3.Bucket(self.bucketName)
1835 try:
1836 bucket.objects.all().delete()
1837 except botocore.exceptions.ClientError as e:
1838 if e.response["Error"]["Code"] == "404":
1839 # the key was not reachable - pass
1840 pass
1841 else:
1842 raise
1844 bucket = s3.Bucket(self.bucketName)
1845 bucket.delete()
1847 # Stop the S3 mock.
1848 self.mock_s3.stop()
1850 # unset any potentially set dummy credentials
1851 if self.usingDummyCredentials:
1852 unsetAwsEnvCredentials()
1854 if self.reg_dir is not None and os.path.exists(self.reg_dir):
1855 shutil.rmtree(self.reg_dir, ignore_errors=True)
1857 if self.useTempRoot and os.path.exists(self.root):
1858 shutil.rmtree(self.root, ignore_errors=True)
1860 super().tearDown()
1863class PosixDatastoreTransfers(unittest.TestCase):
1864 """Test data transfers between butlers.
1866 Test for different managers. UUID to UUID and integer to integer are
1867 tested. UUID to integer is not supported since we do not currently
1868 want to allow that. Integer to UUID is supported with the caveat
1869 that UUID4 will be generated and this will be incorrect for raw
1870 dataset types. The test ignores that.
1871 """
1873 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1875 @classmethod
1876 def setUpClass(cls):
1877 cls.storageClassFactory = StorageClassFactory()
1878 cls.storageClassFactory.addFromConfig(cls.configFile)
1880 def setUp(self):
1881 self.root = makeTestTempDir(TESTDIR)
1882 self.config = Config(self.configFile)
1884 def tearDown(self):
1885 removeTestTempDir(self.root)
1887 def create_butler(self, manager, label):
1888 config = Config(self.configFile)
1889 config["registry", "managers", "datasets"] = manager
1890 return Butler(Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True)
1892 def create_butlers(self, manager1, manager2):
1893 self.source_butler = self.create_butler(manager1, "1")
1894 self.target_butler = self.create_butler(manager2, "2")
1896 def testTransferUuidToUuid(self):
1897 self.create_butlers(
1898 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1899 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1900 )
1901 # Setting id_gen_map should have no effect here
1902 self.assertButlerTransfers(id_gen_map={"random_data_2": DatasetIdGenEnum.DATAID_TYPE})
1904 def testTransferMissing(self):
1905 """Test transfers where datastore records are missing.
1907 This is how execution butler works.
1908 """
1909 self.create_butlers(
1910 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1911 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1912 )
1914 # Configure the source butler to allow trust.
1915 self.source_butler.datastore.trustGetRequest = True
1917 self.assertButlerTransfers(purge=True)
1919 def testTransferMissingDisassembly(self):
1920 """Test transfers where datastore records are missing.
1922 This is how execution butler works.
1923 """
1924 self.create_butlers(
1925 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1926 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
1927 )
1929 # Configure the source butler to allow trust.
1930 self.source_butler.datastore.trustGetRequest = True
1932 # Test disassembly.
1933 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
1935 def assertButlerTransfers(self, id_gen_map=None, purge=False, storageClassName="StructuredData"):
1936 """Test that a run can be transferred to another butler."""
1938 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
1939 datasetTypeName = "random_data"
1941 # Test will create 3 collections and we will want to transfer
1942 # two of those three.
1943 runs = ["run1", "run2", "other"]
1945 # Also want to use two different dataset types to ensure that
1946 # grouping works.
1947 datasetTypeNames = ["random_data", "random_data_2"]
1949 # Create the run collections in the source butler.
1950 for run in runs:
1951 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
1953 # Create dimensions in source butler.
1954 n_exposures = 30
1955 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1956 self.source_butler.registry.insertDimensionData(
1957 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1958 )
1959 self.source_butler.registry.insertDimensionData(
1960 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1961 )
1963 for i in range(n_exposures):
1964 self.source_butler.registry.insertDimensionData(
1965 "exposure",
1966 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
1967 )
1969 # Create dataset types in the source butler.
1970 dimensions = self.source_butler.registry.dimensions.extract(["instrument", "exposure"])
1971 for datasetTypeName in datasetTypeNames:
1972 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1973 self.source_butler.registry.registerDatasetType(datasetType)
1975 # Write a dataset to an unrelated run -- this will ensure that
1976 # we are rewriting integer dataset ids in the target if necessary.
1977 # Will not be relevant for UUID.
1978 run = "distraction"
1979 butler = Butler(butler=self.source_butler, run=run)
1980 butler.put(
1981 makeExampleMetrics(),
1982 datasetTypeName,
1983 exposure=1,
1984 instrument="DummyCamComp",
1985 physical_filter="d-r",
1986 )
1988 # Write some example metrics to the source
1989 butler = Butler(butler=self.source_butler)
1991 # Set of DatasetRefs that should be in the list of refs to transfer
1992 # but which will not be transferred.
1993 deleted = set()
1995 n_expected = 20 # Number of datasets expected to be transferred
1996 source_refs = []
1997 for i in range(n_exposures):
1998 # Put a third of datasets into each collection, only retain
1999 # two thirds.
2000 index = i % 3
2001 run = runs[index]
2002 datasetTypeName = datasetTypeNames[i % 2]
2004 metric_data = {
2005 "summary": {"counter": i},
2006 "output": {"text": "metric"},
2007 "data": [2 * x for x in range(i)],
2008 }
2009 metric = MetricsExample(**metric_data)
2010 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2011 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2013 # Remove the datastore record using low-level API
2014 if purge:
2015 # Remove records for a fraction.
2016 if index == 1:
2017 # For one of these delete the file as well.
2018 # This allows the "missing" code to filter the
2019 # file out.
2020 if not deleted:
2021 primary, uris = butler.datastore.getURIs(ref)
2022 if primary:
2023 primary.remove()
2024 for uri in uris.values():
2025 uri.remove()
2026 n_expected -= 1
2027 deleted.add(ref)
2029 # Remove the datastore record.
2030 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref.id})
2032 if index < 2:
2033 source_refs.append(ref)
2034 if ref not in deleted:
2035 new_metric = butler.get(ref.unresolved(), collections=run)
2036 self.assertEqual(new_metric, metric)
2038 # Create some bad dataset types to ensure we check for inconsistent
2039 # definitions.
2040 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2041 for datasetTypeName in datasetTypeNames:
2042 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2043 self.target_butler.registry.registerDatasetType(datasetType)
2044 with self.assertRaises(ConflictingDefinitionError) as cm:
2045 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map)
2046 self.assertIn("dataset type differs", str(cm.exception))
2048 # And remove the bad definitions.
2049 for datasetTypeName in datasetTypeNames:
2050 self.target_butler.registry.removeDatasetType(datasetTypeName)
2052 # Transfer without creating dataset types should fail.
2053 with self.assertRaises(KeyError):
2054 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map)
2056 # Transfer without creating dimensions should fail.
2057 with self.assertRaises(ConflictingDefinitionError) as cm:
2058 self.target_butler.transfer_from(
2059 self.source_butler, source_refs, id_gen_map=id_gen_map, register_dataset_types=True
2060 )
2061 self.assertIn("dimension", str(cm.exception))
2063 # The failed transfer above leaves registry in an inconsistent
2064 # state because the run is created but then rolled back without
2065 # the collection cache being cleared. For now force a refresh.
2066 # Can remove with DM-35498.
2067 self.target_butler.registry.refresh()
2069 # Now transfer them to the second butler, including dimensions.
2070 with self.assertLogs(level=logging.DEBUG) as cm:
2071 transferred = self.target_butler.transfer_from(
2072 self.source_butler,
2073 source_refs,
2074 id_gen_map=id_gen_map,
2075 register_dataset_types=True,
2076 transfer_dimensions=True,
2077 )
2078 self.assertEqual(len(transferred), n_expected)
2079 log_output = ";".join(cm.output)
2080 self.assertIn("found in datastore for chunk", log_output)
2081 self.assertIn("Creating output run", log_output)
2083 # Do the transfer twice to ensure that it will do nothing extra.
2084 # Only do this if purge=True because it does not work for int
2085 # dataset_id.
2086 if purge:
2087 # This should not need to register dataset types.
2088 transferred = self.target_butler.transfer_from(
2089 self.source_butler, source_refs, id_gen_map=id_gen_map
2090 )
2091 self.assertEqual(len(transferred), n_expected)
2093 # Also do an explicit low-level transfer to trigger some
2094 # edge cases.
2095 with self.assertLogs(level=logging.DEBUG) as cm:
2096 self.target_butler.datastore.transfer_from(self.source_butler.datastore, source_refs)
2097 log_output = ";".join(cm.output)
2098 self.assertIn("no file artifacts exist", log_output)
2100 with self.assertRaises(TypeError):
2101 self.target_butler.datastore.transfer_from(self.source_butler, source_refs)
2103 with self.assertRaises(ValueError):
2104 self.target_butler.datastore.transfer_from(
2105 self.source_butler.datastore, source_refs, transfer="split"
2106 )
2108 # Now try to get the same refs from the new butler.
2109 for ref in source_refs:
2110 if ref not in deleted:
2111 unresolved_ref = ref.unresolved()
2112 new_metric = self.target_butler.get(unresolved_ref, collections=ref.run)
2113 old_metric = self.source_butler.get(unresolved_ref, collections=ref.run)
2114 self.assertEqual(new_metric, old_metric)
2116 # Now prune run2 collection and create instead a CHAINED collection.
2117 # This should block the transfer.
2118 self.target_butler.pruneCollection("run2", purge=True, unstore=True)
2119 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2120 with self.assertRaises(CollectionTypeError):
2121 # Re-importing the run1 datasets can be problematic if they
2122 # use integer IDs so filter those out.
2123 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2124 self.target_butler.transfer_from(self.source_butler, to_transfer, id_gen_map=id_gen_map)
2127if __name__ == "__main__": 2127 ↛ 2128line 2127 didn't jump to line 2128, because the condition on line 2127 was never true
2128 unittest.main()