Coverage for tests/test_butler.py: 13%

1065 statements  

« prev     ^ index     » next       coverage.py v6.5.0, created at 2023-02-14 02:05 -0800

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/>. 

21 

22"""Tests for Butler. 

23""" 

24 

25import gc 

26import logging 

27import os 

28import pathlib 

29import pickle 

30import posixpath 

31import random 

32import shutil 

33import string 

34import tempfile 

35import unittest 

36 

37try: 

38 import boto3 

39 import botocore 

40 from moto import mock_s3 

41except ImportError: 

42 boto3 = None 

43 

44 def mock_s3(cls): 

45 """A no-op decorator in case moto mock_s3 can not be imported.""" 

46 return cls 

47 

48 

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 

57 

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 

89 

90TESTDIR = os.path.abspath(os.path.dirname(__file__)) 

91 

92 

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 ) 

99 

100 

101class TransactionTestError(Exception): 

102 """Specific error for testing transactions, to prevent misdiagnosing 

103 that might otherwise occur when a standard exception is used. 

104 """ 

105 

106 pass 

107 

108 

109class ButlerConfigTests(unittest.TestCase): 

110 """Simple tests for ButlerConfig that are not tested in any other test 

111 cases.""" 

112 

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)) 

118 

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)) 

123 

124 key = ("datastore", "records", "table") 

125 self.assertNotEqual(config1[key], config2[key]) 

126 self.assertEqual(config2[key], "override_record") 

127 

128 

129class ButlerPutGetTests: 

130 """Helper method for running a suite of put/get tests from different 

131 butler configurations.""" 

132 

133 root = None 

134 default_run = "ingésτ😺" 

135 

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 

142 

143 @classmethod 

144 def setUpClass(cls): 

145 cls.storageClassFactory = StorageClassFactory() 

146 cls.storageClassFactory.addFromConfig(cls.configFile) 

147 

148 def assertGetComponents(self, butler, datasetRef, components, reference, collections=None): 

149 datasetType = datasetRef.datasetType 

150 dataId = datasetRef.dataId 

151 deferred = butler.getDirectDeferred(datasetRef) 

152 

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) 

159 

160 def tearDown(self): 

161 removeTestTempDir(self.root) 

162 

163 def create_butler(self, run, storageClass, datasetTypeName): 

164 butler = Butler(self.tmpConfigFile, run=run) 

165 

166 collections = set(butler.registry.queryCollections()) 

167 self.assertEqual(collections, set([run])) 

168 

169 # Create and register a DatasetType 

170 dimensions = butler.registry.dimensions.extract(["instrument", "visit"]) 

171 

172 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry) 

173 

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 ) 

196 

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 

210 

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) 

216 

217 # Create and store a dataset 

218 metric = makeExampleMetrics() 

219 dataId = {"instrument": "DummyCamComp", "visit": 423} 

220 

221 # Create a DatasetRef for put 

222 refIn = DatasetRef(datasetType, dataId, id=None) 

223 

224 # Put with a preexisting id should fail 

225 with self.assertRaises(ValueError): 

226 butler.put(metric, DatasetRef(datasetType, dataId, id=100)) 

227 

228 # Put and remove the dataset once as a DatasetRef, once as a dataId, 

229 # and once with a DatasetType 

230 

231 # Keep track of any collections we add and do not clean up 

232 expected_collections = {run} 

233 

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}) 

244 

245 with self.subTest(args=args): 

246 ref = butler.put(metric, *args, run=this_run) 

247 self.assertIsInstance(ref, DatasetRef) 

248 

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) 

267 

268 # Check we can get components 

269 if storageClass.isComposite(): 

270 self.assertGetComponents( 

271 butler, ref, ("summary", "data", "output"), metric, collections=this_run 

272 ) 

273 

274 # Can the artifacts themselves be retrieved? 

275 if not butler.datastore.isEphemeral: 

276 root_uri = ResourcePath(self.root) 

277 

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)) 

290 

291 for artifact in transferred: 

292 path_in_destination = artifact.relative_to(destination) 

293 self.assertIsNotNone(path_in_destination) 

294 

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) 

302 

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 ) 

313 

314 if preserve_path: 

315 # No need to run these twice 

316 with self.assertRaises(ValueError): 

317 butler.retrieveArtifacts([ref], destination, transfer="move") 

318 

319 with self.assertRaises(FileExistsError): 

320 butler.retrieveArtifacts([ref], destination) 

321 

322 transferred_again = butler.retrieveArtifacts( 

323 [ref], destination, preserve_path=preserve_path, overwrite=True 

324 ) 

325 self.assertEqual(set(transferred_again), set(transferred)) 

326 

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)) 

337 

338 # Do explicit registry removal since we know they are 

339 # empty 

340 butler.registry.removeCollection(this_run) 

341 expected_collections.remove(this_run) 

342 

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) 

346 

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) 

366 

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) 

376 

377 if "counter" in storageClass.derivedComponents: 

378 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId) 

379 self.assertEqual(count, len(data)) 

380 

381 count = butler.get( 

382 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)} 

383 ) 

384 self.assertEqual(count, stop) 

385 

386 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections) 

387 summary = butler.getDirect(compRef) 

388 self.assertEqual(summary, metric.summary) 

389 

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 ) 

394 

395 # Getting with a dataset type that does not match registry fails 

396 with self.assertRaises(ValueError): 

397 butler.get(inconsistentDatasetType, dataId) 

398 

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)) 

405 

406 # Getting a dataset with unknown parameters should fail 

407 with self.assertRaises(KeyError): 

408 butler.get(ref, parameters={"unsupported": True}) 

409 

410 # Check we have a collection 

411 collections = set(butler.registry.queryCollections()) 

412 self.assertEqual(collections, expected_collections) 

413 

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) 

417 

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) 

421 

422 # Repeat put will fail. 

423 with self.assertRaises(ConflictingDefinitionError): 

424 butler.put(metric, refIn) 

425 

426 # Remove the datastore entry. 

427 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False) 

428 

429 # Put will still fail 

430 with self.assertRaises(ConflictingDefinitionError): 

431 butler.put(metric, refIn) 

432 

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) 

437 

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) 

442 

443 # Reset the flag to avoid confusion 

444 butler._allow_put_of_predefined_dataset = False 

445 

446 # Leave the dataset in place since some downstream tests require 

447 # something to be present 

448 

449 return butler 

450 

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])) 

498 

499 

500class ButlerTests(ButlerPutGetTests): 

501 """Tests for Butler.""" 

502 

503 useTempRoot = True 

504 

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") 

510 

511 def testConstructor(self): 

512 """Independent test of constructor.""" 

513 butler = Butler(self.tmpConfigFile, run=self.default_run) 

514 self.assertIsInstance(butler, Butler) 

515 

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) 

521 

522 # Even with a ResourcePath. 

523 butler = Butler(ResourcePath(config_dir, forceDirectory=True), run=self.default_run) 

524 self.assertIsInstance(butler, Butler) 

525 

526 collections = set(butler.registry.queryCollections()) 

527 self.assertEqual(collections, {self.default_run}) 

528 

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}) 

534 

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) 

539 

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()) 

577 

578 def testBasicPutGet(self): 

579 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

580 self.runPutGetTest(storageClass, "test_metric") 

581 

582 def testCompositePutGetConcrete(self): 

583 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly") 

584 butler = self.runPutGetTest(storageClass, "test_metric") 

585 

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}") 

594 

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}") 

602 

603 def testCompositePutGetVirtual(self): 

604 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp") 

605 butler = self.runPutGetTest(storageClass, "test_metric_comp") 

606 

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]) 

611 

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}") 

625 

626 # Predicted dataset 

627 dataId = {"instrument": "DummyCamComp", "visit": 424} 

628 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True) 

629 

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}") 

643 

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 

649 

650 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName) 

651 

652 # Create and store a dataset. 

653 metric = makeExampleMetrics() 

654 dataId = {"instrument": "DummyCamComp", "visit": 423} 

655 

656 ref = butler.put(metric, datasetType, dataId) 

657 

658 # Return native type. 

659 retrieved = butler.get(ref) 

660 self.assertEqual(retrieved, metric) 

661 

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) 

668 

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) 

674 

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) 

680 

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)) 

685 

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) 

695 

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 ) 

705 

706 def testPytypePutCoercion(self): 

707 """Test python type coercion on Butler.get and put.""" 

708 

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) 

713 

714 dataId = {"instrument": "DummyCamComp", "visit": 423} 

715 

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"]) 

723 

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) 

730 

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") 

734 

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") 

740 

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") 

746 

747 def testIngest(self): 

748 butler = Butler(self.tmpConfigFile, run=self.default_run) 

749 

750 # Create and register a DatasetType 

751 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "detector"]) 

752 

753 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml") 

754 datasetTypeName = "metric" 

755 

756 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry) 

757 

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 ) 

767 

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 ) 

773 

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) 

783 

784 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter)) 

785 

786 butler.ingest(*datasets, transfer="copy") 

787 

788 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423} 

789 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423} 

790 

791 metrics1 = butler.get(datasetTypeName, dataId1) 

792 metrics2 = butler.get(datasetTypeName, dataId2) 

793 self.assertNotEqual(metrics1, metrics2) 

794 

795 # Compare URIs 

796 uri1 = butler.getURI(datasetTypeName, dataId1) 

797 uri2 = butler.getURI(datasetTypeName, dataId2) 

798 self.assertNotEqual(uri1, uri2) 

799 

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)) 

808 

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") 

813 

814 datasets = [] 

815 datasets.append(FileDataset(path=tempFile, refs=refs, formatter=MultiDetectorFormatter)) 

816 

817 butler.ingest(*datasets, transfer="move", record_validation_info=False) 

818 self.assertFalse(tempFile.exists()) 

819 

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 

827 

828 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424} 

829 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424} 

830 

831 multi1 = butler.get(datasetTypeName, dataId1) 

832 multi2 = butler.get(datasetTypeName, dataId2) 

833 

834 self.assertEqual(multi1, metrics1) 

835 self.assertEqual(multi2, metrics2) 

836 

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}") 

841 

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) 

851 

852 def testPickle(self): 

853 """Test pickle support.""" 

854 butler = Butler(self.tmpConfigFile, run=self.default_run) 

855 butlerOut = pickle.loads(pickle.dumps(butler)) 

856 self.assertIsInstance(butlerOut, Butler) 

857 self.assertEqual(butlerOut._config, butler._config) 

858 self.assertEqual(butlerOut.collections, butler.collections) 

859 self.assertEqual(butlerOut.run, butler.run) 

860 

861 def testGetDatasetTypes(self): 

862 butler = Butler(self.tmpConfigFile, run=self.default_run) 

863 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "physical_filter"]) 

864 dimensionEntries = [ 

865 ( 

866 "instrument", 

867 {"instrument": "DummyCam"}, 

868 {"instrument": "DummyHSC"}, 

869 {"instrument": "DummyCamComp"}, 

870 ), 

871 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}), 

872 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}), 

873 ] 

874 storageClass = self.storageClassFactory.getStorageClass("StructuredData") 

875 # Add needed Dimensions 

876 for args in dimensionEntries: 

877 butler.registry.insertDimensionData(*args) 

878 

879 # When a DatasetType is added to the registry entries are not created 

880 # for components but querying them can return the components. 

881 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"} 

882 components = set() 

883 for datasetTypeName in datasetTypeNames: 

884 # Create and register a DatasetType 

885 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry) 

886 

887 for componentName in storageClass.components: 

888 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName)) 

889 

890 fromRegistry: set[DatasetType] = set() 

891 for parent_dataset_type in butler.registry.queryDatasetTypes(): 

892 fromRegistry.add(parent_dataset_type) 

893 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes()) 

894 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components) 

895 

896 # Now that we have some dataset types registered, validate them 

897 butler.validateConfiguration( 

898 ignore=[ 

899 "test_metric_comp", 

900 "metric3", 

901 "metric5", 

902 "calexp", 

903 "DummySC", 

904 "datasetType.component", 

905 "random_data", 

906 "random_data_2", 

907 ] 

908 ) 

909 

910 # Add a new datasetType that will fail template validation 

911 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry) 

912 if self.validationCanFail: 

913 with self.assertRaises(ValidationError): 

914 butler.validateConfiguration() 

915 

916 # Rerun validation but with a subset of dataset type names 

917 butler.validateConfiguration(datasetTypeNames=["metric4"]) 

918 

919 # Rerun validation but ignore the bad datasetType 

920 butler.validateConfiguration( 

921 ignore=[ 

922 "test_metric_comp", 

923 "metric3", 

924 "metric5", 

925 "calexp", 

926 "DummySC", 

927 "datasetType.component", 

928 "random_data", 

929 "random_data_2", 

930 ] 

931 ) 

932 

933 def testTransaction(self): 

934 butler = Butler(self.tmpConfigFile, run=self.default_run) 

935 datasetTypeName = "test_metric" 

936 dimensions = butler.registry.dimensions.extract(["instrument", "visit"]) 

937 dimensionEntries = ( 

938 ("instrument", {"instrument": "DummyCam"}), 

939 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}), 

940 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}), 

941 ) 

942 storageClass = self.storageClassFactory.getStorageClass("StructuredData") 

943 metric = makeExampleMetrics() 

944 dataId = {"instrument": "DummyCam", "visit": 42} 

945 # Create and register a DatasetType 

946 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry) 

947 with self.assertRaises(TransactionTestError): 

948 with butler.transaction(): 

949 # Add needed Dimensions 

950 for args in dimensionEntries: 

951 butler.registry.insertDimensionData(*args) 

952 # Store a dataset 

953 ref = butler.put(metric, datasetTypeName, dataId) 

954 self.assertIsInstance(ref, DatasetRef) 

955 # Test getDirect 

956 metricOut = butler.getDirect(ref) 

957 self.assertEqual(metric, metricOut) 

958 # Test get 

959 metricOut = butler.get(datasetTypeName, dataId) 

960 self.assertEqual(metric, metricOut) 

961 # Check we can get components 

962 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric) 

963 raise TransactionTestError("This should roll back the entire transaction") 

964 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"): 

965 butler.registry.expandDataId(dataId) 

966 # Should raise LookupError for missing data ID value 

967 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"): 

968 butler.get(datasetTypeName, dataId) 

969 # Also check explicitly if Dataset entry is missing 

970 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections)) 

971 # Direct retrieval should not find the file in the Datastore 

972 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"): 

973 butler.getDirect(ref) 

974 

975 def testMakeRepo(self): 

976 """Test that we can write butler configuration to a new repository via 

977 the Butler.makeRepo interface and then instantiate a butler from the 

978 repo root. 

979 """ 

980 # Do not run the test if we know this datastore configuration does 

981 # not support a file system root 

982 if self.fullConfigKey is None: 

983 return 

984 

985 # create two separate directories 

986 root1 = tempfile.mkdtemp(dir=self.root) 

987 root2 = tempfile.mkdtemp(dir=self.root) 

988 

989 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile)) 

990 limited = Config(self.configFile) 

991 butler1 = Butler(butlerConfig) 

992 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile)) 

993 full = Config(self.tmpConfigFile) 

994 butler2 = Butler(butlerConfig) 

995 # Butlers should have the same configuration regardless of whether 

996 # defaults were expanded. 

997 self.assertEqual(butler1._config, butler2._config) 

998 # Config files loaded directly should not be the same. 

999 self.assertNotEqual(limited, full) 

1000 # Make sure "limited" doesn't have a few keys we know it should be 

1001 # inheriting from defaults. 

1002 self.assertIn(self.fullConfigKey, full) 

1003 self.assertNotIn(self.fullConfigKey, limited) 

1004 

1005 # Collections don't appear until something is put in them 

1006 collections1 = set(butler1.registry.queryCollections()) 

1007 self.assertEqual(collections1, set()) 

1008 self.assertEqual(set(butler2.registry.queryCollections()), collections1) 

1009 

1010 # Check that a config with no associated file name will not 

1011 # work properly with relocatable Butler repo 

1012 butlerConfig.configFile = None 

1013 with self.assertRaises(ValueError): 

1014 Butler(butlerConfig) 

1015 

1016 with self.assertRaises(FileExistsError): 

1017 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False) 

1018 

1019 def testStringification(self): 

1020 butler = Butler(self.tmpConfigFile, run=self.default_run) 

1021 butlerStr = str(butler) 

1022 

1023 if self.datastoreStr is not None: 

1024 for testStr in self.datastoreStr: 

1025 self.assertIn(testStr, butlerStr) 

1026 if self.registryStr is not None: 

1027 self.assertIn(self.registryStr, butlerStr) 

1028 

1029 datastoreName = butler.datastore.name 

1030 if self.datastoreName is not None: 

1031 for testStr in self.datastoreName: 

1032 self.assertIn(testStr, datastoreName) 

1033 

1034 def testButlerRewriteDataId(self): 

1035 """Test that dataIds can be rewritten based on dimension records.""" 

1036 

1037 butler = Butler(self.tmpConfigFile, run=self.default_run) 

1038 

1039 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict") 

1040 datasetTypeName = "random_data" 

1041 

1042 # Create dimension records. 

1043 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"}) 

1044 butler.registry.insertDimensionData( 

1045 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"} 

1046 ) 

1047 butler.registry.insertDimensionData( 

1048 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"} 

1049 ) 

1050 

1051 dimensions = butler.registry.dimensions.extract(["instrument", "exposure"]) 

1052 datasetType = DatasetType(datasetTypeName, dimensions, storageClass) 

1053 butler.registry.registerDatasetType(datasetType) 

1054 

1055 n_exposures = 5 

1056 dayobs = 20210530 

1057 

1058 for i in range(n_exposures): 

1059 butler.registry.insertDimensionData( 

1060 "exposure", 

1061 { 

1062 "instrument": "DummyCamComp", 

1063 "id": i, 

1064 "obs_id": f"exp{i}", 

1065 "seq_num": i, 

1066 "day_obs": dayobs, 

1067 "physical_filter": "d-r", 

1068 }, 

1069 ) 

1070 

1071 # Write some data. 

1072 for i in range(n_exposures): 

1073 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]} 

1074 

1075 # Use the seq_num for the put to test rewriting. 

1076 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"} 

1077 ref = butler.put(metric, datasetTypeName, dataId=dataId) 

1078 

1079 # Check that the exposure is correct in the dataId 

1080 self.assertEqual(ref.dataId["exposure"], i) 

1081 

1082 # and check that we can get the dataset back with the same dataId 

1083 new_metric = butler.get(datasetTypeName, dataId=dataId) 

1084 self.assertEqual(new_metric, metric) 

1085 

1086 

1087class FileDatastoreButlerTests(ButlerTests): 

1088 """Common tests and specialization of ButlerTests for butlers backed 

1089 by datastores that inherit from FileDatastore. 

1090 """ 

1091 

1092 def checkFileExists(self, root, relpath): 

1093 """Checks if file exists at a given path (relative to root). 

1094 

1095 Test testPutTemplates verifies actual physical existance of the files 

1096 in the requested location. 

1097 """ 

1098 uri = ResourcePath(root, forceDirectory=True) 

1099 return uri.join(relpath).exists() 

1100 

1101 def testPutTemplates(self): 

1102 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1103 butler = Butler(self.tmpConfigFile, run=self.default_run) 

1104 

1105 # Add needed Dimensions 

1106 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"}) 

1107 butler.registry.insertDimensionData( 

1108 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"} 

1109 ) 

1110 butler.registry.insertDimensionData( 

1111 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"} 

1112 ) 

1113 butler.registry.insertDimensionData( 

1114 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"} 

1115 ) 

1116 

1117 # Create and store a dataset 

1118 metric = makeExampleMetrics() 

1119 

1120 # Create two almost-identical DatasetTypes (both will use default 

1121 # template) 

1122 dimensions = butler.registry.dimensions.extract(["instrument", "visit"]) 

1123 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass)) 

1124 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass)) 

1125 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass)) 

1126 

1127 dataId1 = {"instrument": "DummyCamComp", "visit": 423} 

1128 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"} 

1129 

1130 # Put with exactly the data ID keys needed 

1131 ref = butler.put(metric, "metric1", dataId1) 

1132 uri = butler.getURI(ref) 

1133 self.assertTrue(uri.exists()) 

1134 self.assertTrue( 

1135 uri.unquoted_path.endswith(f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle") 

1136 ) 

1137 

1138 # Check the template based on dimensions 

1139 if hasattr(butler.datastore, "templates"): 

1140 butler.datastore.templates.validateTemplates([ref]) 

1141 

1142 # Put with extra data ID keys (physical_filter is an optional 

1143 # dependency); should not change template (at least the way we're 

1144 # defining them to behave now; the important thing is that they 

1145 # must be consistent). 

1146 ref = butler.put(metric, "metric2", dataId2) 

1147 uri = butler.getURI(ref) 

1148 self.assertTrue(uri.exists()) 

1149 self.assertTrue( 

1150 uri.unquoted_path.endswith(f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle") 

1151 ) 

1152 

1153 # Check the template based on dimensions 

1154 if hasattr(butler.datastore, "templates"): 

1155 butler.datastore.templates.validateTemplates([ref]) 

1156 

1157 # Use a template that has a typo in dimension record metadata. 

1158 # Easier to test with a butler that has a ref with records attached. 

1159 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits") 

1160 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"): 

1161 path = template.format(ref) 

1162 self.assertEqual(path, f"a/v423/{ref.id}_fits") 

1163 

1164 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits") 

1165 with self.assertRaises(KeyError): 

1166 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"): 

1167 template.format(ref) 

1168 

1169 # Now use a file template that will not result in unique filenames 

1170 with self.assertRaises(FileTemplateValidationError): 

1171 butler.put(metric, "metric3", dataId1) 

1172 

1173 def testImportExport(self): 

1174 # Run put/get tests just to create and populate a repo. 

1175 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1176 self.runImportExportTest(storageClass) 

1177 

1178 @unittest.expectedFailure 

1179 def testImportExportVirtualComposite(self): 

1180 # Run put/get tests just to create and populate a repo. 

1181 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite") 

1182 self.runImportExportTest(storageClass) 

1183 

1184 def runImportExportTest(self, storageClass): 

1185 """This test does an export to a temp directory and an import back 

1186 into a new temp directory repo. It does not assume a posix datastore""" 

1187 exportButler = self.runPutGetTest(storageClass, "test_metric") 

1188 # Test that the repo actually has at least one dataset. 

1189 datasets = list(exportButler.registry.queryDatasets(..., collections=...)) 

1190 self.assertGreater(len(datasets), 0) 

1191 # Add a DimensionRecord that's unused by those datasets. 

1192 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")} 

1193 exportButler.registry.insertDimensionData("skymap", skymapRecord) 

1194 # Export and then import datasets. 

1195 with safeTestTempDir(TESTDIR) as exportDir: 

1196 exportFile = os.path.join(exportDir, "exports.yaml") 

1197 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export: 

1198 export.saveDatasets(datasets) 

1199 # Export the same datasets again. This should quietly do 

1200 # nothing because of internal deduplication, and it shouldn't 

1201 # complain about being asked to export the "htm7" elements even 

1202 # though there aren't any in these datasets or in the database. 

1203 export.saveDatasets(datasets, elements=["htm7"]) 

1204 # Save one of the data IDs again; this should be harmless 

1205 # because of internal deduplication. 

1206 export.saveDataIds([datasets[0].dataId]) 

1207 # Save some dimension records directly. 

1208 export.saveDimensionData("skymap", [skymapRecord]) 

1209 self.assertTrue(os.path.exists(exportFile)) 

1210 with safeTestTempDir(TESTDIR) as importDir: 

1211 # We always want this to be a local posix butler 

1212 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml"))) 

1213 # Calling script.butlerImport tests the implementation of the 

1214 # butler command line interface "import" subcommand. Functions 

1215 # in the script folder are generally considered protected and 

1216 # should not be used as public api. 

1217 with open(exportFile, "r") as f: 

1218 script.butlerImport( 

1219 importDir, 

1220 export_file=f, 

1221 directory=exportDir, 

1222 transfer="auto", 

1223 skip_dimensions=None, 

1224 reuse_ids=False, 

1225 ) 

1226 importButler = Butler(importDir, run=self.default_run) 

1227 for ref in datasets: 

1228 with self.subTest(ref=ref): 

1229 # Test for existence by passing in the DatasetType and 

1230 # data ID separately, to avoid lookup by dataset_id. 

1231 self.assertTrue(importButler.datasetExists(ref.datasetType, ref.dataId)) 

1232 self.assertEqual( 

1233 list(importButler.registry.queryDimensionRecords("skymap")), 

1234 [importButler.registry.dimensions["skymap"].RecordClass(**skymapRecord)], 

1235 ) 

1236 

1237 def testRemoveRuns(self): 

1238 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1239 butler = Butler(self.tmpConfigFile, writeable=True) 

1240 # Load registry data with dimensions to hang datasets off of. 

1241 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry")) 

1242 butler.import_(filename=os.path.join(registryDataDir, "base.yaml")) 

1243 # Add some RUN-type collection. 

1244 run1 = "run1" 

1245 butler.registry.registerRun(run1) 

1246 run2 = "run2" 

1247 butler.registry.registerRun(run2) 

1248 # put a dataset in each 

1249 metric = makeExampleMetrics() 

1250 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"]) 

1251 datasetType = self.addDatasetType( 

1252 "prune_collections_test_dataset", dimensions, storageClass, butler.registry 

1253 ) 

1254 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1) 

1255 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2) 

1256 uri1 = butler.getURI(ref1, collections=[run1]) 

1257 uri2 = butler.getURI(ref2, collections=[run2]) 

1258 # Remove from both runs with different values for unstore. 

1259 butler.removeRuns([run1], unstore=True) 

1260 butler.removeRuns([run2], unstore=False) 

1261 # Should be nothing in registry for either one, and datastore should 

1262 # not think either exists. 

1263 with self.assertRaises(MissingCollectionError): 

1264 butler.registry.getCollectionType(run1) 

1265 with self.assertRaises(MissingCollectionError): 

1266 butler.registry.getCollectionType(run2) 

1267 self.assertFalse(butler.datastore.exists(ref1)) 

1268 self.assertFalse(butler.datastore.exists(ref2)) 

1269 # The ref we unstored should be gone according to the URI, but the 

1270 # one we forgot should still be around. 

1271 self.assertFalse(uri1.exists()) 

1272 self.assertTrue(uri2.exists()) 

1273 

1274 

1275class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase): 

1276 """PosixDatastore specialization of a butler""" 

1277 

1278 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1279 fullConfigKey = ".datastore.formatters" 

1280 validationCanFail = True 

1281 datastoreStr = ["/tmp"] 

1282 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"] 

1283 registryStr = "/gen3.sqlite3" 

1284 

1285 def testPathConstructor(self): 

1286 """Independent test of constructor using PathLike.""" 

1287 butler = Butler(self.tmpConfigFile, run=self.default_run) 

1288 self.assertIsInstance(butler, Butler) 

1289 

1290 # And again with a Path object with the butler yaml 

1291 path = pathlib.Path(self.tmpConfigFile) 

1292 butler = Butler(path, writeable=False) 

1293 self.assertIsInstance(butler, Butler) 

1294 

1295 # And again with a Path object without the butler yaml 

1296 # (making sure we skip it if the tmp config doesn't end 

1297 # in butler.yaml -- which is the case for a subclass) 

1298 if self.tmpConfigFile.endswith("butler.yaml"): 

1299 path = pathlib.Path(os.path.dirname(self.tmpConfigFile)) 

1300 butler = Butler(path, writeable=False) 

1301 self.assertIsInstance(butler, Butler) 

1302 

1303 def testExportTransferCopy(self): 

1304 """Test local export using all transfer modes""" 

1305 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1306 exportButler = self.runPutGetTest(storageClass, "test_metric") 

1307 # Test that the repo actually has at least one dataset. 

1308 datasets = list(exportButler.registry.queryDatasets(..., collections=...)) 

1309 self.assertGreater(len(datasets), 0) 

1310 uris = [exportButler.getURI(d) for d in datasets] 

1311 datastoreRoot = exportButler.datastore.root 

1312 

1313 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris] 

1314 

1315 for path in pathsInStore: 

1316 # Assume local file system 

1317 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}") 

1318 

1319 for transfer in ("copy", "link", "symlink", "relsymlink"): 

1320 with safeTestTempDir(TESTDIR) as exportDir: 

1321 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export: 

1322 export.saveDatasets(datasets) 

1323 for path in pathsInStore: 

1324 self.assertTrue( 

1325 self.checkFileExists(exportDir, path), 

1326 f"Check that mode {transfer} exported files", 

1327 ) 

1328 

1329 def testPruneDatasets(self): 

1330 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1331 butler = Butler(self.tmpConfigFile, writeable=True) 

1332 # Load registry data with dimensions to hang datasets off of. 

1333 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry")) 

1334 butler.import_(filename=os.path.join(registryDataDir, "base.yaml")) 

1335 # Add some RUN-type collections. 

1336 run1 = "run1" 

1337 butler.registry.registerRun(run1) 

1338 run2 = "run2" 

1339 butler.registry.registerRun(run2) 

1340 # put some datasets. ref1 and ref2 have the same data ID, and are in 

1341 # different runs. ref3 has a different data ID. 

1342 metric = makeExampleMetrics() 

1343 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"]) 

1344 datasetType = self.addDatasetType( 

1345 "prune_collections_test_dataset", dimensions, storageClass, butler.registry 

1346 ) 

1347 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1) 

1348 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2) 

1349 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1) 

1350 

1351 # Simple prune. 

1352 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True) 

1353 with self.assertRaises(LookupError): 

1354 butler.datasetExists(ref1.datasetType, ref1.dataId, collections=run1) 

1355 

1356 # Put data back. 

1357 ref1 = butler.put(metric, ref1.unresolved(), run=run1) 

1358 ref2 = butler.put(metric, ref2.unresolved(), run=run2) 

1359 ref3 = butler.put(metric, ref3.unresolved(), run=run1) 

1360 

1361 # Check that in normal mode, deleting the record will lead to 

1362 # trash not touching the file. 

1363 uri1 = butler.datastore.getURI(ref1) 

1364 butler.datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table 

1365 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref1.id}) 

1366 butler.datastore.trash(ref1) 

1367 butler.datastore.emptyTrash() 

1368 self.assertTrue(uri1.exists()) 

1369 uri1.remove() # Clean it up. 

1370 

1371 # Simulate execution butler setup by deleting the datastore 

1372 # record but keeping the file around and trusting. 

1373 butler.datastore.trustGetRequest = True 

1374 uri2 = butler.datastore.getURI(ref2) 

1375 uri3 = butler.datastore.getURI(ref3) 

1376 self.assertTrue(uri2.exists()) 

1377 self.assertTrue(uri3.exists()) 

1378 

1379 # Remove the datastore record. 

1380 butler.datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table 

1381 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref2.id}) 

1382 self.assertTrue(uri2.exists()) 

1383 butler.datastore.trash([ref2, ref3]) 

1384 # Immediate removal for ref2 file 

1385 self.assertFalse(uri2.exists()) 

1386 # But ref3 has to wait for the empty. 

1387 self.assertTrue(uri3.exists()) 

1388 butler.datastore.emptyTrash() 

1389 self.assertFalse(uri3.exists()) 

1390 

1391 # Clear out the datasets from registry. 

1392 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True) 

1393 

1394 def testPytypeCoercion(self): 

1395 """Test python type coercion on Butler.get and put.""" 

1396 

1397 # Store some data with the normal example storage class. 

1398 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1399 datasetTypeName = "test_metric" 

1400 butler = self.runPutGetTest(storageClass, datasetTypeName) 

1401 

1402 dataId = {"instrument": "DummyCamComp", "visit": 423} 

1403 metric = butler.get(datasetTypeName, dataId=dataId) 

1404 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample") 

1405 

1406 datasetType_ori = butler.registry.getDatasetType(datasetTypeName) 

1407 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents") 

1408 

1409 # Now need to hack the registry dataset type definition. 

1410 # There is no API for this. 

1411 manager = butler.registry._managers.datasets 

1412 manager._db.update( 

1413 manager._static.dataset_type, 

1414 {"name": datasetTypeName}, 

1415 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"}, 

1416 ) 

1417 

1418 # Force reset of dataset type cache 

1419 butler.registry.refresh() 

1420 

1421 datasetType_new = butler.registry.getDatasetType(datasetTypeName) 

1422 self.assertEqual(datasetType_new.name, datasetType_ori.name) 

1423 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel") 

1424 

1425 metric_model = butler.get(datasetTypeName, dataId=dataId) 

1426 self.assertNotEqual(type(metric_model), type(metric)) 

1427 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel") 

1428 

1429 # Put the model and read it back to show that everything now 

1430 # works as normal. 

1431 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424) 

1432 metric_model_new = butler.get(metric_ref) 

1433 self.assertEqual(metric_model_new, metric_model) 

1434 

1435 # Hack the storage class again to something that will fail on the 

1436 # get with no conversion class. 

1437 manager._db.update( 

1438 manager._static.dataset_type, 

1439 {"name": datasetTypeName}, 

1440 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"}, 

1441 ) 

1442 butler.registry.refresh() 

1443 

1444 with self.assertRaises(ValueError): 

1445 butler.get(datasetTypeName, dataId=dataId) 

1446 

1447 

1448@unittest.skipUnless(testing is not None, "testing.postgresql module not found") 

1449class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase): 

1450 """PosixDatastore specialization of a butler using Postgres""" 

1451 

1452 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1453 fullConfigKey = ".datastore.formatters" 

1454 validationCanFail = True 

1455 datastoreStr = ["/tmp"] 

1456 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"] 

1457 registryStr = "PostgreSQL@test" 

1458 

1459 @staticmethod 

1460 def _handler(postgresql): 

1461 engine = sqlalchemy.engine.create_engine(postgresql.url()) 

1462 with engine.begin() as connection: 

1463 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;")) 

1464 

1465 @classmethod 

1466 def setUpClass(cls): 

1467 # Create the postgres test server. 

1468 cls.postgresql = testing.postgresql.PostgresqlFactory( 

1469 cache_initialized_db=True, on_initialized=cls._handler 

1470 ) 

1471 super().setUpClass() 

1472 

1473 @classmethod 

1474 def tearDownClass(cls): 

1475 # Clean up any lingering SQLAlchemy engines/connections 

1476 # so they're closed before we shut down the server. 

1477 gc.collect() 

1478 cls.postgresql.clear_cache() 

1479 super().tearDownClass() 

1480 

1481 def setUp(self): 

1482 self.server = self.postgresql() 

1483 

1484 # Need to add a registry section to the config. 

1485 self._temp_config = False 

1486 config = Config(self.configFile) 

1487 config["registry", "db"] = self.server.url() 

1488 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh: 

1489 config.dump(fh) 

1490 self.configFile = fh.name 

1491 self._temp_config = True 

1492 super().setUp() 

1493 

1494 def tearDown(self): 

1495 self.server.stop() 

1496 if self._temp_config and os.path.exists(self.configFile): 

1497 os.remove(self.configFile) 

1498 super().tearDown() 

1499 

1500 def testMakeRepo(self): 

1501 # The base class test assumes that it's using sqlite and assumes 

1502 # the config file is acceptable to sqlite. 

1503 raise unittest.SkipTest("Postgres config is not compatible with this test.") 

1504 

1505 

1506class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase): 

1507 """InMemoryDatastore specialization of a butler""" 

1508 

1509 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml") 

1510 fullConfigKey = None 

1511 useTempRoot = False 

1512 validationCanFail = False 

1513 datastoreStr = ["datastore='InMemory"] 

1514 datastoreName = ["InMemoryDatastore@"] 

1515 registryStr = "/gen3.sqlite3" 

1516 

1517 def testIngest(self): 

1518 pass 

1519 

1520 

1521class ChainedDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase): 

1522 """PosixDatastore specialization""" 

1523 

1524 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml") 

1525 fullConfigKey = ".datastore.datastores.1.formatters" 

1526 validationCanFail = True 

1527 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"] 

1528 datastoreName = [ 

1529 "InMemoryDatastore@", 

1530 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1", 

1531 "SecondDatastore", 

1532 ] 

1533 registryStr = "/gen3.sqlite3" 

1534 

1535 

1536class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase): 

1537 """Test that a yaml file in one location can refer to a root in another.""" 

1538 

1539 datastoreStr = ["dir1"] 

1540 # Disable the makeRepo test since we are deliberately not using 

1541 # butler.yaml as the config name. 

1542 fullConfigKey = None 

1543 

1544 def setUp(self): 

1545 self.root = makeTestTempDir(TESTDIR) 

1546 

1547 # Make a new repository in one place 

1548 self.dir1 = os.path.join(self.root, "dir1") 

1549 Butler.makeRepo(self.dir1, config=Config(self.configFile)) 

1550 

1551 # Move the yaml file to a different place and add a "root" 

1552 self.dir2 = os.path.join(self.root, "dir2") 

1553 os.makedirs(self.dir2, exist_ok=True) 

1554 configFile1 = os.path.join(self.dir1, "butler.yaml") 

1555 config = Config(configFile1) 

1556 config["root"] = self.dir1 

1557 configFile2 = os.path.join(self.dir2, "butler2.yaml") 

1558 config.dumpToUri(configFile2) 

1559 os.remove(configFile1) 

1560 self.tmpConfigFile = configFile2 

1561 

1562 def testFileLocations(self): 

1563 self.assertNotEqual(self.dir1, self.dir2) 

1564 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml"))) 

1565 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml"))) 

1566 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3"))) 

1567 

1568 

1569class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase): 

1570 """Test that a config file created by makeRepo outside of repo works.""" 

1571 

1572 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1573 

1574 def setUp(self): 

1575 self.root = makeTestTempDir(TESTDIR) 

1576 self.root2 = makeTestTempDir(TESTDIR) 

1577 

1578 self.tmpConfigFile = os.path.join(self.root2, "different.yaml") 

1579 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile) 

1580 

1581 def tearDown(self): 

1582 if os.path.exists(self.root2): 

1583 shutil.rmtree(self.root2, ignore_errors=True) 

1584 super().tearDown() 

1585 

1586 def testConfigExistence(self): 

1587 c = Config(self.tmpConfigFile) 

1588 uri_config = ResourcePath(c["root"]) 

1589 uri_expected = ResourcePath(self.root, forceDirectory=True) 

1590 self.assertEqual(uri_config.geturl(), uri_expected.geturl()) 

1591 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path") 

1592 

1593 def testPutGet(self): 

1594 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents") 

1595 self.runPutGetTest(storageClass, "test_metric") 

1596 

1597 

1598class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase): 

1599 """Test that a config file created by makeRepo outside of repo works.""" 

1600 

1601 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1602 

1603 def setUp(self): 

1604 self.root = makeTestTempDir(TESTDIR) 

1605 self.root2 = makeTestTempDir(TESTDIR) 

1606 

1607 self.tmpConfigFile = self.root2 

1608 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile) 

1609 

1610 def testConfigExistence(self): 

1611 # Append the yaml file else Config constructor does not know the file 

1612 # type. 

1613 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml") 

1614 super().testConfigExistence() 

1615 

1616 

1617class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase): 

1618 """Test that a config file created by makeRepo outside of repo works.""" 

1619 

1620 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1621 

1622 def setUp(self): 

1623 self.root = makeTestTempDir(TESTDIR) 

1624 self.root2 = makeTestTempDir(TESTDIR) 

1625 

1626 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl() 

1627 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile) 

1628 

1629 

1630@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!") 

1631class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase): 

1632 """S3Datastore specialization of a butler; an S3 storage Datastore + 

1633 a local in-memory SqlRegistry. 

1634 """ 

1635 

1636 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml") 

1637 fullConfigKey = None 

1638 validationCanFail = True 

1639 

1640 bucketName = "anybucketname" 

1641 """Name of the Bucket that will be used in the tests. The name is read from 

1642 the config file used with the tests during set-up. 

1643 """ 

1644 

1645 root = "butlerRoot/" 

1646 """Root repository directory expected to be used in case useTempRoot=False. 

1647 Otherwise the root is set to a 20 characters long randomly generated string 

1648 during set-up. 

1649 """ 

1650 

1651 datastoreStr = [f"datastore={root}"] 

1652 """Contains all expected root locations in a format expected to be 

1653 returned by Butler stringification. 

1654 """ 

1655 

1656 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"] 

1657 """The expected format of the S3 Datastore string.""" 

1658 

1659 registryStr = "/gen3.sqlite3" 

1660 """Expected format of the Registry string.""" 

1661 

1662 mock_s3 = mock_s3() 

1663 """The mocked s3 interface from moto.""" 

1664 

1665 def genRoot(self): 

1666 """Returns a random string of len 20 to serve as a root 

1667 name for the temporary bucket repo. 

1668 

1669 This is equivalent to tempfile.mkdtemp as this is what self.root 

1670 becomes when useTempRoot is True. 

1671 """ 

1672 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20)) 

1673 return rndstr + "/" 

1674 

1675 def setUp(self): 

1676 config = Config(self.configFile) 

1677 uri = ResourcePath(config[".datastore.datastore.root"]) 

1678 self.bucketName = uri.netloc 

1679 

1680 # Enable S3 mocking of tests. 

1681 self.mock_s3.start() 

1682 

1683 # set up some fake credentials if they do not exist 

1684 self.usingDummyCredentials = setAwsEnvCredentials() 

1685 

1686 if self.useTempRoot: 

1687 self.root = self.genRoot() 

1688 rooturi = f"s3://{self.bucketName}/{self.root}" 

1689 config.update({"datastore": {"datastore": {"root": rooturi}}}) 

1690 

1691 # need local folder to store registry database 

1692 self.reg_dir = makeTestTempDir(TESTDIR) 

1693 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3" 

1694 

1695 # MOTO needs to know that we expect Bucket bucketname to exist 

1696 # (this used to be the class attribute bucketName) 

1697 s3 = boto3.resource("s3") 

1698 s3.create_bucket(Bucket=self.bucketName) 

1699 

1700 self.datastoreStr = f"datastore={self.root}" 

1701 self.datastoreName = [f"FileDatastore@{rooturi}"] 

1702 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False) 

1703 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml") 

1704 

1705 def tearDown(self): 

1706 s3 = boto3.resource("s3") 

1707 bucket = s3.Bucket(self.bucketName) 

1708 try: 

1709 bucket.objects.all().delete() 

1710 except botocore.exceptions.ClientError as e: 

1711 if e.response["Error"]["Code"] == "404": 

1712 # the key was not reachable - pass 

1713 pass 

1714 else: 

1715 raise 

1716 

1717 bucket = s3.Bucket(self.bucketName) 

1718 bucket.delete() 

1719 

1720 # Stop the S3 mock. 

1721 self.mock_s3.stop() 

1722 

1723 # unset any potentially set dummy credentials 

1724 if self.usingDummyCredentials: 

1725 unsetAwsEnvCredentials() 

1726 

1727 if self.reg_dir is not None and os.path.exists(self.reg_dir): 

1728 shutil.rmtree(self.reg_dir, ignore_errors=True) 

1729 

1730 if self.useTempRoot and os.path.exists(self.root): 

1731 shutil.rmtree(self.root, ignore_errors=True) 

1732 

1733 super().tearDown() 

1734 

1735 

1736class PosixDatastoreTransfers(unittest.TestCase): 

1737 """Test data transfers between butlers. 

1738 

1739 Test for different managers. UUID to UUID and integer to integer are 

1740 tested. UUID to integer is not supported since we do not currently 

1741 want to allow that. Integer to UUID is supported with the caveat 

1742 that UUID4 will be generated and this will be incorrect for raw 

1743 dataset types. The test ignores that. 

1744 """ 

1745 

1746 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml") 

1747 

1748 @classmethod 

1749 def setUpClass(cls): 

1750 cls.storageClassFactory = StorageClassFactory() 

1751 cls.storageClassFactory.addFromConfig(cls.configFile) 

1752 

1753 def setUp(self): 

1754 self.root = makeTestTempDir(TESTDIR) 

1755 self.config = Config(self.configFile) 

1756 

1757 def tearDown(self): 

1758 removeTestTempDir(self.root) 

1759 

1760 def create_butler(self, manager, label): 

1761 config = Config(self.configFile) 

1762 config["registry", "managers", "datasets"] = manager 

1763 return Butler(Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True) 

1764 

1765 def create_butlers(self, manager1, manager2): 

1766 self.source_butler = self.create_butler(manager1, "1") 

1767 self.target_butler = self.create_butler(manager2, "2") 

1768 

1769 def testTransferUuidToUuid(self): 

1770 self.create_butlers( 

1771 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1772 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1773 ) 

1774 # Setting id_gen_map should have no effect here 

1775 self.assertButlerTransfers(id_gen_map={"random_data_2": DatasetIdGenEnum.DATAID_TYPE}) 

1776 

1777 def testTransferMissing(self): 

1778 """Test transfers where datastore records are missing. 

1779 

1780 This is how execution butler works. 

1781 """ 

1782 self.create_butlers( 

1783 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1784 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1785 ) 

1786 

1787 # Configure the source butler to allow trust. 

1788 self.source_butler.datastore.trustGetRequest = True 

1789 

1790 self.assertButlerTransfers(purge=True) 

1791 

1792 def testTransferMissingDisassembly(self): 

1793 """Test transfers where datastore records are missing. 

1794 

1795 This is how execution butler works. 

1796 """ 

1797 self.create_butlers( 

1798 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1799 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID", 

1800 ) 

1801 

1802 # Configure the source butler to allow trust. 

1803 self.source_butler.datastore.trustGetRequest = True 

1804 

1805 # Test disassembly. 

1806 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite") 

1807 

1808 def assertButlerTransfers(self, id_gen_map=None, purge=False, storageClassName="StructuredData"): 

1809 """Test that a run can be transferred to another butler.""" 

1810 

1811 storageClass = self.storageClassFactory.getStorageClass(storageClassName) 

1812 datasetTypeName = "random_data" 

1813 

1814 # Test will create 3 collections and we will want to transfer 

1815 # two of those three. 

1816 runs = ["run1", "run2", "other"] 

1817 

1818 # Also want to use two different dataset types to ensure that 

1819 # grouping works. 

1820 datasetTypeNames = ["random_data", "random_data_2"] 

1821 

1822 # Create the run collections in the source butler. 

1823 for run in runs: 

1824 self.source_butler.registry.registerCollection(run, CollectionType.RUN) 

1825 

1826 # Create dimensions in source butler. 

1827 n_exposures = 30 

1828 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"}) 

1829 self.source_butler.registry.insertDimensionData( 

1830 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"} 

1831 ) 

1832 self.source_butler.registry.insertDimensionData( 

1833 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"} 

1834 ) 

1835 

1836 for i in range(n_exposures): 

1837 self.source_butler.registry.insertDimensionData( 

1838 "exposure", 

1839 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"}, 

1840 ) 

1841 

1842 # Create dataset types in the source butler. 

1843 dimensions = self.source_butler.registry.dimensions.extract(["instrument", "exposure"]) 

1844 for datasetTypeName in datasetTypeNames: 

1845 datasetType = DatasetType(datasetTypeName, dimensions, storageClass) 

1846 self.source_butler.registry.registerDatasetType(datasetType) 

1847 

1848 # Write a dataset to an unrelated run -- this will ensure that 

1849 # we are rewriting integer dataset ids in the target if necessary. 

1850 # Will not be relevant for UUID. 

1851 run = "distraction" 

1852 butler = Butler(butler=self.source_butler, run=run) 

1853 butler.put( 

1854 makeExampleMetrics(), 

1855 datasetTypeName, 

1856 exposure=1, 

1857 instrument="DummyCamComp", 

1858 physical_filter="d-r", 

1859 ) 

1860 

1861 # Write some example metrics to the source 

1862 butler = Butler(butler=self.source_butler) 

1863 

1864 # Set of DatasetRefs that should be in the list of refs to transfer 

1865 # but which will not be transferred. 

1866 deleted = set() 

1867 

1868 n_expected = 20 # Number of datasets expected to be transferred 

1869 source_refs = [] 

1870 for i in range(n_exposures): 

1871 # Put a third of datasets into each collection, only retain 

1872 # two thirds. 

1873 index = i % 3 

1874 run = runs[index] 

1875 datasetTypeName = datasetTypeNames[i % 2] 

1876 

1877 metric_data = { 

1878 "summary": {"counter": i}, 

1879 "output": {"text": "metric"}, 

1880 "data": [2 * x for x in range(i)], 

1881 } 

1882 metric = MetricsExample(**metric_data) 

1883 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"} 

1884 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run) 

1885 

1886 # Remove the datastore record using low-level API 

1887 if purge: 

1888 # Remove records for a fraction. 

1889 if index == 1: 

1890 # For one of these delete the file as well. 

1891 # This allows the "missing" code to filter the 

1892 # file out. 

1893 if not deleted: 

1894 primary, uris = butler.datastore.getURIs(ref) 

1895 if primary: 

1896 primary.remove() 

1897 for uri in uris.values(): 

1898 uri.remove() 

1899 n_expected -= 1 

1900 deleted.add(ref) 

1901 

1902 # Remove the datastore record. 

1903 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref.id}) 

1904 

1905 if index < 2: 

1906 source_refs.append(ref) 

1907 if ref not in deleted: 

1908 new_metric = butler.get(ref.unresolved(), collections=run) 

1909 self.assertEqual(new_metric, metric) 

1910 

1911 # Create some bad dataset types to ensure we check for inconsistent 

1912 # definitions. 

1913 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList") 

1914 for datasetTypeName in datasetTypeNames: 

1915 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass) 

1916 self.target_butler.registry.registerDatasetType(datasetType) 

1917 with self.assertRaises(ConflictingDefinitionError) as cm: 

1918 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map) 

1919 self.assertIn("dataset type differs", str(cm.exception)) 

1920 

1921 # And remove the bad definitions. 

1922 for datasetTypeName in datasetTypeNames: 

1923 self.target_butler.registry.removeDatasetType(datasetTypeName) 

1924 

1925 # Transfer without creating dataset types should fail. 

1926 with self.assertRaises(KeyError): 

1927 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map) 

1928 

1929 # Transfer without creating dimensions should fail. 

1930 with self.assertRaises(ConflictingDefinitionError) as cm: 

1931 self.target_butler.transfer_from( 

1932 self.source_butler, source_refs, id_gen_map=id_gen_map, register_dataset_types=True 

1933 ) 

1934 self.assertIn("dimension", str(cm.exception)) 

1935 

1936 # The failed transfer above leaves registry in an inconsistent 

1937 # state because the run is created but then rolled back without 

1938 # the collection cache being cleared. For now force a refresh. 

1939 # Can remove with DM-35498. 

1940 self.target_butler.registry.refresh() 

1941 

1942 # Now transfer them to the second butler, including dimensions. 

1943 with self.assertLogs(level=logging.DEBUG) as cm: 

1944 transferred = self.target_butler.transfer_from( 

1945 self.source_butler, 

1946 source_refs, 

1947 id_gen_map=id_gen_map, 

1948 register_dataset_types=True, 

1949 transfer_dimensions=True, 

1950 ) 

1951 self.assertEqual(len(transferred), n_expected) 

1952 log_output = ";".join(cm.output) 

1953 self.assertIn("found in datastore for chunk", log_output) 

1954 self.assertIn("Creating output run", log_output) 

1955 

1956 # Do the transfer twice to ensure that it will do nothing extra. 

1957 # Only do this if purge=True because it does not work for int 

1958 # dataset_id. 

1959 if purge: 

1960 # This should not need to register dataset types. 

1961 transferred = self.target_butler.transfer_from( 

1962 self.source_butler, source_refs, id_gen_map=id_gen_map 

1963 ) 

1964 self.assertEqual(len(transferred), n_expected) 

1965 

1966 # Also do an explicit low-level transfer to trigger some 

1967 # edge cases. 

1968 with self.assertLogs(level=logging.DEBUG) as cm: 

1969 self.target_butler.datastore.transfer_from(self.source_butler.datastore, source_refs) 

1970 log_output = ";".join(cm.output) 

1971 self.assertIn("no file artifacts exist", log_output) 

1972 

1973 with self.assertRaises(TypeError): 

1974 self.target_butler.datastore.transfer_from(self.source_butler, source_refs) 

1975 

1976 with self.assertRaises(ValueError): 

1977 self.target_butler.datastore.transfer_from( 

1978 self.source_butler.datastore, source_refs, transfer="split" 

1979 ) 

1980 

1981 # Now try to get the same refs from the new butler. 

1982 for ref in source_refs: 

1983 if ref not in deleted: 

1984 unresolved_ref = ref.unresolved() 

1985 new_metric = self.target_butler.get(unresolved_ref, collections=ref.run) 

1986 old_metric = self.source_butler.get(unresolved_ref, collections=ref.run) 

1987 self.assertEqual(new_metric, old_metric) 

1988 

1989 # Now prune run2 collection and create instead a CHAINED collection. 

1990 # This should block the transfer. 

1991 self.target_butler.removeRuns(["run2"], unstore=True) 

1992 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED) 

1993 with self.assertRaises(CollectionTypeError): 

1994 # Re-importing the run1 datasets can be problematic if they 

1995 # use integer IDs so filter those out. 

1996 to_transfer = [ref for ref in source_refs if ref.run == "run2"] 

1997 self.target_butler.transfer_from(self.source_butler, to_transfer, id_gen_map=id_gen_map) 

1998 

1999 

2000if __name__ == "__main__": 2000 ↛ 2001line 2000 didn't jump to line 2001, because the condition on line 2000 was never true

2001 unittest.main()