Coverage for tests/test_butler.py: 14%
1248 statements
« prev ^ index » next coverage.py v6.5.0, created at 2022-10-07 09:47 +0000
« prev ^ index » next coverage.py v6.5.0, created at 2022-10-07 09:47 +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 socket
34import string
35import tempfile
36import time
37import unittest
38from tempfile import gettempdir
39from threading import Thread
41try:
42 import boto3
43 import botocore
44 from moto import mock_s3
45except ImportError:
46 boto3 = None
48 def mock_s3(cls):
49 """A no-op decorator in case moto mock_s3 can not be imported."""
50 return cls
53try:
54 # It's possible but silly to have testing.postgresql installed without
55 # having the postgresql server installed (because then nothing in
56 # testing.postgresql would work), so we use the presence of that module
57 # to test whether we can expect the server to be available.
58 import testing.postgresql
59except ImportError:
60 testing = None
63try:
64 from cheroot import wsgi
65 from wsgidav.wsgidav_app import WsgiDAVApp
66except ImportError:
67 WsgiDAVApp = None
69import astropy.time
70import sqlalchemy
71from lsst.daf.butler import (
72 Butler,
73 ButlerConfig,
74 CollectionType,
75 Config,
76 DatasetIdGenEnum,
77 DatasetRef,
78 DatasetType,
79 FileDataset,
80 FileTemplate,
81 FileTemplateValidationError,
82 StorageClassFactory,
83 ValidationError,
84 script,
85)
86from lsst.daf.butler.core.repoRelocation import BUTLER_ROOT_TAG
87from lsst.daf.butler.registry import (
88 CollectionError,
89 CollectionTypeError,
90 ConflictingDefinitionError,
91 DataIdValueError,
92 MissingCollectionError,
93)
94from lsst.daf.butler.tests import MetricsExample, MultiDetectorFormatter
95from lsst.daf.butler.tests.utils import makeTestTempDir, removeTestTempDir, safeTestTempDir
96from lsst.resources import ResourcePath
97from lsst.resources.http import _is_webdav_endpoint
98from lsst.resources.s3utils import setAwsEnvCredentials, unsetAwsEnvCredentials
99from lsst.utils import doImport
100from lsst.utils.introspection import get_full_type_name
102TESTDIR = os.path.abspath(os.path.dirname(__file__))
105def makeExampleMetrics():
106 return MetricsExample(
107 {"AM1": 5.2, "AM2": 30.6},
108 {"a": [1, 2, 3], "b": {"blue": 5, "red": "green"}},
109 [563, 234, 456.7, 752, 8, 9, 27],
110 )
113class TransactionTestError(Exception):
114 """Specific error for testing transactions, to prevent misdiagnosing
115 that might otherwise occur when a standard exception is used.
116 """
118 pass
121class ButlerConfigTests(unittest.TestCase):
122 """Simple tests for ButlerConfig that are not tested in any other test
123 cases."""
125 def testSearchPath(self):
126 configFile = os.path.join(TESTDIR, "config", "basic", "butler.yaml")
127 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
128 config1 = ButlerConfig(configFile)
129 self.assertNotIn("testConfigs", "\n".join(cm.output))
131 overrideDirectory = os.path.join(TESTDIR, "config", "testConfigs")
132 with self.assertLogs("lsst.daf.butler", level="DEBUG") as cm:
133 config2 = ButlerConfig(configFile, searchPaths=[overrideDirectory])
134 self.assertIn("testConfigs", "\n".join(cm.output))
136 key = ("datastore", "records", "table")
137 self.assertNotEqual(config1[key], config2[key])
138 self.assertEqual(config2[key], "override_record")
141class ButlerPutGetTests:
142 """Helper method for running a suite of put/get tests from different
143 butler configurations."""
145 root = None
146 default_run = "ingésτ😺"
148 @staticmethod
149 def addDatasetType(datasetTypeName, dimensions, storageClass, registry):
150 """Create a DatasetType and register it"""
151 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
152 registry.registerDatasetType(datasetType)
153 return datasetType
155 @classmethod
156 def setUpClass(cls):
157 cls.storageClassFactory = StorageClassFactory()
158 cls.storageClassFactory.addFromConfig(cls.configFile)
160 def assertGetComponents(self, butler, datasetRef, components, reference, collections=None):
161 datasetType = datasetRef.datasetType
162 dataId = datasetRef.dataId
163 deferred = butler.getDirectDeferred(datasetRef)
165 for component in components:
166 compTypeName = datasetType.componentTypeName(component)
167 result = butler.get(compTypeName, dataId, collections=collections)
168 self.assertEqual(result, getattr(reference, component))
169 result_deferred = deferred.get(component=component)
170 self.assertEqual(result_deferred, result)
172 def tearDown(self):
173 removeTestTempDir(self.root)
175 def create_butler(self, run, storageClass, datasetTypeName):
176 butler = Butler(self.tmpConfigFile, run=run)
178 collections = set(butler.registry.queryCollections())
179 self.assertEqual(collections, set([run]))
181 # Create and register a DatasetType
182 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
184 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
186 # Add needed Dimensions
187 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
188 butler.registry.insertDimensionData(
189 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
190 )
191 butler.registry.insertDimensionData(
192 "visit_system", {"instrument": "DummyCamComp", "id": 1, "name": "default"}
193 )
194 visit_start = astropy.time.Time("2020-01-01 08:00:00.123456789", scale="tai")
195 visit_end = astropy.time.Time("2020-01-01 08:00:36.66", scale="tai")
196 butler.registry.insertDimensionData(
197 "visit",
198 {
199 "instrument": "DummyCamComp",
200 "id": 423,
201 "name": "fourtwentythree",
202 "physical_filter": "d-r",
203 "visit_system": 1,
204 "datetime_begin": visit_start,
205 "datetime_end": visit_end,
206 },
207 )
209 # Add more visits for some later tests
210 for visit_id in (424, 425):
211 butler.registry.insertDimensionData(
212 "visit",
213 {
214 "instrument": "DummyCamComp",
215 "id": visit_id,
216 "name": f"fourtwentyfour_{visit_id}",
217 "physical_filter": "d-r",
218 "visit_system": 1,
219 },
220 )
221 return butler, datasetType
223 def runPutGetTest(self, storageClass, datasetTypeName):
224 # New datasets will be added to run and tag, but we will only look in
225 # tag when looking up datasets.
226 run = self.default_run
227 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
229 # Create and store a dataset
230 metric = makeExampleMetrics()
231 dataId = {"instrument": "DummyCamComp", "visit": 423}
233 # Create a DatasetRef for put
234 refIn = DatasetRef(datasetType, dataId, id=None)
236 # Put with a preexisting id should fail
237 with self.assertRaises(ValueError):
238 butler.put(metric, DatasetRef(datasetType, dataId, id=100))
240 # Put and remove the dataset once as a DatasetRef, once as a dataId,
241 # and once with a DatasetType
243 # Keep track of any collections we add and do not clean up
244 expected_collections = {run}
246 counter = 0
247 for args in ((refIn,), (datasetTypeName, dataId), (datasetType, dataId)):
248 # Since we are using subTest we can get cascading failures
249 # here with the first attempt failing and the others failing
250 # immediately because the dataset already exists. Work around
251 # this by using a distinct run collection each time
252 counter += 1
253 this_run = f"put_run_{counter}"
254 butler.registry.registerCollection(this_run, type=CollectionType.RUN)
255 expected_collections.update({this_run})
257 with self.subTest(args=args):
258 ref = butler.put(metric, *args, run=this_run)
259 self.assertIsInstance(ref, DatasetRef)
261 # Test getDirect
262 metricOut = butler.getDirect(ref)
263 self.assertEqual(metric, metricOut)
264 # Test get
265 metricOut = butler.get(ref.datasetType.name, dataId, collections=this_run)
266 self.assertEqual(metric, metricOut)
267 # Test get with a datasetRef
268 metricOut = butler.get(ref, collections=this_run)
269 self.assertEqual(metric, metricOut)
270 # Test getDeferred with dataId
271 metricOut = butler.getDeferred(ref.datasetType.name, dataId, collections=this_run).get()
272 self.assertEqual(metric, metricOut)
273 # Test getDeferred with a datasetRef
274 metricOut = butler.getDeferred(ref, collections=this_run).get()
275 self.assertEqual(metric, metricOut)
276 # and deferred direct with ref
277 metricOut = butler.getDirectDeferred(ref).get()
278 self.assertEqual(metric, metricOut)
280 # Check we can get components
281 if storageClass.isComposite():
282 self.assertGetComponents(
283 butler, ref, ("summary", "data", "output"), metric, collections=this_run
284 )
286 # Can the artifacts themselves be retrieved?
287 if not butler.datastore.isEphemeral:
288 root_uri = ResourcePath(self.root)
290 for preserve_path in (True, False):
291 destination = root_uri.join(f"artifacts/{preserve_path}_{counter}/")
292 # Use copy so that we can test that overwrite
293 # protection works (using "auto" for File URIs would
294 # use hard links and subsequent transfer would work
295 # because it knows they are the same file).
296 transferred = butler.retrieveArtifacts(
297 [ref], destination, preserve_path=preserve_path, transfer="copy"
298 )
299 self.assertGreater(len(transferred), 0)
300 artifacts = list(ResourcePath.findFileResources([destination]))
301 self.assertEqual(set(transferred), set(artifacts))
303 for artifact in transferred:
304 path_in_destination = artifact.relative_to(destination)
305 self.assertIsNotNone(path_in_destination)
307 # when path is not preserved there should not be
308 # any path separators.
309 num_seps = path_in_destination.count("/")
310 if preserve_path:
311 self.assertGreater(num_seps, 0)
312 else:
313 self.assertEqual(num_seps, 0)
315 primary_uri, secondary_uris = butler.datastore.getURIs(ref)
316 n_uris = len(secondary_uris)
317 if primary_uri:
318 n_uris += 1
319 self.assertEqual(
320 len(artifacts),
321 n_uris,
322 "Comparing expected artifacts vs actual:"
323 f" {artifacts} vs {primary_uri} and {secondary_uris}",
324 )
326 if preserve_path:
327 # No need to run these twice
328 with self.assertRaises(ValueError):
329 butler.retrieveArtifacts([ref], destination, transfer="move")
331 with self.assertRaises(FileExistsError):
332 butler.retrieveArtifacts([ref], destination)
334 transferred_again = butler.retrieveArtifacts(
335 [ref], destination, preserve_path=preserve_path, overwrite=True
336 )
337 self.assertEqual(set(transferred_again), set(transferred))
339 # Now remove the dataset completely.
340 butler.pruneDatasets([ref], purge=True, unstore=True)
341 # Lookup with original args should still fail.
342 with self.assertRaises(LookupError):
343 butler.datasetExists(*args, collections=this_run)
344 # getDirect() should still fail.
345 with self.assertRaises(FileNotFoundError):
346 butler.getDirect(ref)
347 # Registry shouldn't be able to find it by dataset_id anymore.
348 self.assertIsNone(butler.registry.getDataset(ref.id))
350 # Do explicit registry removal since we know they are
351 # empty
352 butler.registry.removeCollection(this_run)
353 expected_collections.remove(this_run)
355 # Put the dataset again, since the last thing we did was remove it
356 # and we want to use the default collection.
357 ref = butler.put(metric, refIn)
359 # Get with parameters
360 stop = 4
361 sliced = butler.get(ref, 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 # getDeferred with parameters
367 sliced = butler.getDeferred(ref, parameters={"slice": slice(stop)}).get()
368 self.assertNotEqual(metric, sliced)
369 self.assertEqual(metric.summary, sliced.summary)
370 self.assertEqual(metric.output, sliced.output)
371 self.assertEqual(metric.data[:stop], sliced.data)
372 # getDeferred with deferred parameters
373 sliced = butler.getDeferred(ref).get(parameters={"slice": slice(stop)})
374 self.assertNotEqual(metric, sliced)
375 self.assertEqual(metric.summary, sliced.summary)
376 self.assertEqual(metric.output, sliced.output)
377 self.assertEqual(metric.data[:stop], sliced.data)
379 if storageClass.isComposite():
380 # Check that components can be retrieved
381 metricOut = butler.get(ref.datasetType.name, dataId)
382 compNameS = ref.datasetType.componentTypeName("summary")
383 compNameD = ref.datasetType.componentTypeName("data")
384 summary = butler.get(compNameS, dataId)
385 self.assertEqual(summary, metric.summary)
386 data = butler.get(compNameD, dataId)
387 self.assertEqual(data, metric.data)
389 if "counter" in storageClass.derivedComponents:
390 count = butler.get(ref.datasetType.componentTypeName("counter"), dataId)
391 self.assertEqual(count, len(data))
393 count = butler.get(
394 ref.datasetType.componentTypeName("counter"), dataId, parameters={"slice": slice(stop)}
395 )
396 self.assertEqual(count, stop)
398 compRef = butler.registry.findDataset(compNameS, dataId, collections=butler.collections)
399 summary = butler.getDirect(compRef)
400 self.assertEqual(summary, metric.summary)
402 # Create a Dataset type that has the same name but is inconsistent.
403 inconsistentDatasetType = DatasetType(
404 datasetTypeName, datasetType.dimensions, self.storageClassFactory.getStorageClass("Config")
405 )
407 # Getting with a dataset type that does not match registry fails
408 with self.assertRaises(ValueError):
409 butler.get(inconsistentDatasetType, dataId)
411 # Combining a DatasetRef with a dataId should fail
412 with self.assertRaises(ValueError):
413 butler.get(ref, dataId)
414 # Getting with an explicit ref should fail if the id doesn't match
415 with self.assertRaises(ValueError):
416 butler.get(DatasetRef(ref.datasetType, ref.dataId, id=101))
418 # Getting a dataset with unknown parameters should fail
419 with self.assertRaises(KeyError):
420 butler.get(ref, parameters={"unsupported": True})
422 # Check we have a collection
423 collections = set(butler.registry.queryCollections())
424 self.assertEqual(collections, expected_collections)
426 # Clean up to check that we can remove something that may have
427 # already had a component removed
428 butler.pruneDatasets([ref], unstore=True, purge=True)
430 # Check that we can configure a butler to accept a put even
431 # if it already has the dataset in registry.
432 ref = butler.put(metric, refIn)
434 # Repeat put will fail.
435 with self.assertRaises(ConflictingDefinitionError):
436 butler.put(metric, refIn)
438 # Remove the datastore entry.
439 butler.pruneDatasets([ref], unstore=True, purge=False, disassociate=False)
441 # Put will still fail
442 with self.assertRaises(ConflictingDefinitionError):
443 butler.put(metric, refIn)
445 # Allow the put to succeed
446 butler._allow_put_of_predefined_dataset = True
447 ref2 = butler.put(metric, refIn)
448 self.assertEqual(ref2.id, ref.id)
450 # A second put will still fail but with a different exception
451 # than before.
452 with self.assertRaises(ConflictingDefinitionError):
453 butler.put(metric, refIn)
455 # Reset the flag to avoid confusion
456 butler._allow_put_of_predefined_dataset = False
458 # Leave the dataset in place since some downstream tests require
459 # something to be present
461 return butler
463 def testDeferredCollectionPassing(self):
464 # Construct a butler with no run or collection, but make it writeable.
465 butler = Butler(self.tmpConfigFile, writeable=True)
466 # Create and register a DatasetType
467 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
468 datasetType = self.addDatasetType(
469 "example", dimensions, self.storageClassFactory.getStorageClass("StructuredData"), butler.registry
470 )
471 # Add needed Dimensions
472 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
473 butler.registry.insertDimensionData(
474 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
475 )
476 butler.registry.insertDimensionData(
477 "visit",
478 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
479 )
480 dataId = {"instrument": "DummyCamComp", "visit": 423}
481 # Create dataset.
482 metric = makeExampleMetrics()
483 # Register a new run and put dataset.
484 run = "deferred"
485 self.assertTrue(butler.registry.registerRun(run))
486 # Second time it will be allowed but indicate no-op
487 self.assertFalse(butler.registry.registerRun(run))
488 ref = butler.put(metric, datasetType, dataId, run=run)
489 # Putting with no run should fail with TypeError.
490 with self.assertRaises(CollectionError):
491 butler.put(metric, datasetType, dataId)
492 # Dataset should exist.
493 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
494 # We should be able to get the dataset back, but with and without
495 # a deferred dataset handle.
496 self.assertEqual(metric, butler.get(datasetType, dataId, collections=[run]))
497 self.assertEqual(metric, butler.getDeferred(datasetType, dataId, collections=[run]).get())
498 # Trying to find the dataset without any collection is a TypeError.
499 with self.assertRaises(CollectionError):
500 butler.datasetExists(datasetType, dataId)
501 with self.assertRaises(CollectionError):
502 butler.get(datasetType, dataId)
503 # Associate the dataset with a different collection.
504 butler.registry.registerCollection("tagged")
505 butler.registry.associate("tagged", [ref])
506 # Deleting the dataset from the new collection should make it findable
507 # in the original collection.
508 butler.pruneDatasets([ref], tags=["tagged"])
509 self.assertTrue(butler.datasetExists(datasetType, dataId, collections=[run]))
512class ButlerTests(ButlerPutGetTests):
513 """Tests for Butler."""
515 useTempRoot = True
517 def setUp(self):
518 """Create a new butler root for each test."""
519 self.root = makeTestTempDir(TESTDIR)
520 Butler.makeRepo(self.root, config=Config(self.configFile))
521 self.tmpConfigFile = os.path.join(self.root, "butler.yaml")
523 def testConstructor(self):
524 """Independent test of constructor."""
525 butler = Butler(self.tmpConfigFile, run=self.default_run)
526 self.assertIsInstance(butler, Butler)
528 # Check that butler.yaml is added automatically.
529 if self.tmpConfigFile.endswith(end := "/butler.yaml"):
530 config_dir = self.tmpConfigFile[: -len(end)]
531 butler = Butler(config_dir, run=self.default_run)
532 self.assertIsInstance(butler, Butler)
534 # Even with a ResourcePath.
535 butler = Butler(ResourcePath(config_dir, forceDirectory=True), run=self.default_run)
536 self.assertIsInstance(butler, Butler)
538 collections = set(butler.registry.queryCollections())
539 self.assertEqual(collections, {self.default_run})
541 # Check that some special characters can be included in run name.
542 special_run = "u@b.c-A"
543 butler_special = Butler(butler=butler, run=special_run)
544 collections = set(butler_special.registry.queryCollections("*@*"))
545 self.assertEqual(collections, {special_run})
547 butler2 = Butler(butler=butler, collections=["other"])
548 self.assertEqual(butler2.collections, ("other",))
549 self.assertIsNone(butler2.run)
550 self.assertIs(butler.datastore, butler2.datastore)
552 # Test that we can use an environment variable to find this
553 # repository.
554 butler_index = Config()
555 butler_index["label"] = self.tmpConfigFile
556 for suffix in (".yaml", ".json"):
557 # Ensure that the content differs so that we know that
558 # we aren't reusing the cache.
559 bad_label = f"s3://bucket/not_real{suffix}"
560 butler_index["bad_label"] = bad_label
561 with ResourcePath.temporary_uri(suffix=suffix) as temp_file:
562 butler_index.dumpToUri(temp_file)
563 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": str(temp_file)}):
564 self.assertEqual(Butler.get_known_repos(), set(("label", "bad_label")))
565 uri = Butler.get_repo_uri("bad_label")
566 self.assertEqual(uri, ResourcePath(bad_label))
567 uri = Butler.get_repo_uri("label")
568 butler = Butler(uri, writeable=False)
569 self.assertIsInstance(butler, Butler)
570 butler = Butler("label", writeable=False)
571 self.assertIsInstance(butler, Butler)
572 with self.assertRaisesRegex(FileNotFoundError, "aliases:.*bad_label"):
573 Butler("not_there", writeable=False)
574 with self.assertRaises(KeyError) as cm:
575 Butler.get_repo_uri("missing")
576 self.assertIn("not known to", str(cm.exception))
577 with unittest.mock.patch.dict(os.environ, {"DAF_BUTLER_REPOSITORY_INDEX": "file://not_found/x.yaml"}):
578 with self.assertRaises(FileNotFoundError):
579 Butler.get_repo_uri("label")
580 self.assertEqual(Butler.get_known_repos(), set())
581 with self.assertRaises(KeyError) as cm:
582 # No environment variable set.
583 Butler.get_repo_uri("label")
584 self.assertIn("No repository index defined", str(cm.exception))
585 with self.assertRaisesRegex(FileNotFoundError, "no known aliases"):
586 # No aliases registered.
587 Butler("not_there")
588 self.assertEqual(Butler.get_known_repos(), set())
590 def testBasicPutGet(self):
591 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
592 self.runPutGetTest(storageClass, "test_metric")
594 def testCompositePutGetConcrete(self):
596 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadCompNoDisassembly")
597 butler = self.runPutGetTest(storageClass, "test_metric")
599 # Should *not* be disassembled
600 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
601 self.assertEqual(len(datasets), 1)
602 uri, components = butler.getURIs(datasets[0])
603 self.assertIsInstance(uri, ResourcePath)
604 self.assertFalse(components)
605 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
606 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
608 # Predicted dataset
609 dataId = {"instrument": "DummyCamComp", "visit": 424}
610 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
611 self.assertFalse(components)
612 self.assertIsInstance(uri, ResourcePath)
613 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
614 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
616 def testCompositePutGetVirtual(self):
617 storageClass = self.storageClassFactory.getStorageClass("StructuredCompositeReadComp")
618 butler = self.runPutGetTest(storageClass, "test_metric_comp")
620 # Should be disassembled
621 datasets = list(butler.registry.queryDatasets(..., collections=self.default_run))
622 self.assertEqual(len(datasets), 1)
623 uri, components = butler.getURIs(datasets[0])
625 if butler.datastore.isEphemeral:
626 # Never disassemble in-memory datastore
627 self.assertIsInstance(uri, ResourcePath)
628 self.assertFalse(components)
629 self.assertEqual(uri.fragment, "", f"Checking absence of fragment in {uri}")
630 self.assertIn("423", str(uri), f"Checking visit is in URI {uri}")
631 else:
632 self.assertIsNone(uri)
633 self.assertEqual(set(components), set(storageClass.components))
634 for compuri in components.values():
635 self.assertIsInstance(compuri, ResourcePath)
636 self.assertIn("423", str(compuri), f"Checking visit is in URI {compuri}")
637 self.assertEqual(compuri.fragment, "", f"Checking absence of fragment in {compuri}")
639 # Predicted dataset
640 dataId = {"instrument": "DummyCamComp", "visit": 424}
641 uri, components = butler.getURIs(datasets[0].datasetType, dataId=dataId, predict=True)
643 if butler.datastore.isEphemeral:
644 # Never disassembled
645 self.assertIsInstance(uri, ResourcePath)
646 self.assertFalse(components)
647 self.assertIn("424", str(uri), f"Checking visit is in URI {uri}")
648 self.assertEqual(uri.fragment, "predicted", f"Checking for fragment in {uri}")
649 else:
650 self.assertIsNone(uri)
651 self.assertEqual(set(components), set(storageClass.components))
652 for compuri in components.values():
653 self.assertIsInstance(compuri, ResourcePath)
654 self.assertIn("424", str(compuri), f"Checking visit is in URI {compuri}")
655 self.assertEqual(compuri.fragment, "predicted", f"Checking for fragment in {compuri}")
657 def testStorageClassOverrideGet(self):
658 """Test storage class conversion on get with override."""
659 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
660 datasetTypeName = "anything"
661 run = self.default_run
663 butler, datasetType = self.create_butler(run, storageClass, datasetTypeName)
665 # Create and store a dataset.
666 metric = makeExampleMetrics()
667 dataId = {"instrument": "DummyCamComp", "visit": 423}
669 ref = butler.put(metric, datasetType, dataId)
671 # Return native type.
672 retrieved = butler.get(ref)
673 self.assertEqual(retrieved, metric)
675 # Specify an override.
676 new_sc = self.storageClassFactory.getStorageClass("MetricsConversion")
677 model = butler.getDirect(ref, storageClass=new_sc)
678 self.assertNotEqual(type(model), type(retrieved))
679 self.assertIs(type(model), new_sc.pytype)
680 self.assertEqual(retrieved, model)
682 # Defer but override later.
683 deferred = butler.getDirectDeferred(ref)
684 model = deferred.get(storageClass=new_sc)
685 self.assertIs(type(model), new_sc.pytype)
686 self.assertEqual(retrieved, model)
688 # Defer but override up front.
689 deferred = butler.getDirectDeferred(ref, storageClass=new_sc)
690 model = deferred.get()
691 self.assertIs(type(model), new_sc.pytype)
692 self.assertEqual(retrieved, model)
694 # Retrieve a component. Should be a tuple.
695 data = butler.get("anything.data", dataId, storageClass="StructuredDataDataTestTuple")
696 self.assertIs(type(data), tuple)
697 self.assertEqual(data, tuple(retrieved.data))
699 # Parameter on the write storage class should work regardless
700 # of read storage class.
701 data = butler.get(
702 "anything.data",
703 dataId,
704 storageClass="StructuredDataDataTestTuple",
705 parameters={"slice": slice(2, 4)},
706 )
707 self.assertEqual(len(data), 2)
709 # Try a parameter that is known to the read storage class but not
710 # the write storage class.
711 with self.assertRaises(KeyError):
712 butler.get(
713 "anything.data",
714 dataId,
715 storageClass="StructuredDataDataTestTuple",
716 parameters={"xslice": slice(2, 4)},
717 )
719 def testPytypePutCoercion(self):
720 """Test python type coercion on Butler.get and put."""
722 # Store some data with the normal example storage class.
723 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
724 datasetTypeName = "test_metric"
725 butler, _ = self.create_butler(self.default_run, storageClass, datasetTypeName)
727 dataId = {"instrument": "DummyCamComp", "visit": 423}
729 # Put a dict and this should coerce to a MetricsExample
730 test_dict = {"summary": {"a": 1}, "output": {"b": 2}}
731 metric_ref = butler.put(test_dict, datasetTypeName, dataId=dataId, visit=424)
732 test_metric = butler.getDirect(metric_ref)
733 self.assertEqual(get_full_type_name(test_metric), "lsst.daf.butler.tests.MetricsExample")
734 self.assertEqual(test_metric.summary, test_dict["summary"])
735 self.assertEqual(test_metric.output, test_dict["output"])
737 # Check that the put still works if a DatasetType is given with
738 # a definition matching this python type.
739 registry_type = butler.registry.getDatasetType(datasetTypeName)
740 this_type = DatasetType(datasetTypeName, registry_type.dimensions, "StructuredDataDictJson")
741 metric2_ref = butler.put(test_dict, this_type, dataId=dataId, visit=425)
742 self.assertEqual(metric2_ref.datasetType, registry_type)
744 # The get will return the type expected by registry.
745 test_metric2 = butler.getDirect(metric2_ref)
746 self.assertEqual(get_full_type_name(test_metric2), "lsst.daf.butler.tests.MetricsExample")
748 # Make a new DatasetRef with the compatible but different DatasetType.
749 # This should now return a dict.
750 new_ref = DatasetRef(this_type, metric2_ref.dataId, id=metric2_ref.id, run=metric2_ref.run)
751 test_dict2 = butler.getDirect(new_ref)
752 self.assertEqual(get_full_type_name(test_dict2), "dict")
754 # Get it again with the wrong dataset type definition using get()
755 # rather than getDirect(). This should be consistent with getDirect()
756 # behavior and return the type of the DatasetType.
757 test_dict3 = butler.get(this_type, dataId=dataId, visit=425)
758 self.assertEqual(get_full_type_name(test_dict3), "dict")
760 def testIngest(self):
761 butler = Butler(self.tmpConfigFile, run=self.default_run)
763 # Create and register a DatasetType
764 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "detector"])
766 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDictYaml")
767 datasetTypeName = "metric"
769 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
771 # Add needed Dimensions
772 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
773 butler.registry.insertDimensionData(
774 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
775 )
776 for detector in (1, 2):
777 butler.registry.insertDimensionData(
778 "detector", {"instrument": "DummyCamComp", "id": detector, "full_name": f"detector{detector}"}
779 )
781 butler.registry.insertDimensionData(
782 "visit",
783 {"instrument": "DummyCamComp", "id": 423, "name": "fourtwentythree", "physical_filter": "d-r"},
784 {"instrument": "DummyCamComp", "id": 424, "name": "fourtwentyfour", "physical_filter": "d-r"},
785 )
787 formatter = doImport("lsst.daf.butler.formatters.yaml.YamlFormatter")
788 dataRoot = os.path.join(TESTDIR, "data", "basic")
789 datasets = []
790 for detector in (1, 2):
791 detector_name = f"detector_{detector}"
792 metricFile = os.path.join(dataRoot, f"{detector_name}.yaml")
793 dataId = {"instrument": "DummyCamComp", "visit": 423, "detector": detector}
794 # Create a DatasetRef for ingest
795 refIn = DatasetRef(datasetType, dataId, id=None)
797 datasets.append(FileDataset(path=metricFile, refs=[refIn], formatter=formatter))
799 butler.ingest(*datasets, transfer="copy")
801 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 423}
802 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 423}
804 metrics1 = butler.get(datasetTypeName, dataId1)
805 metrics2 = butler.get(datasetTypeName, dataId2)
806 self.assertNotEqual(metrics1, metrics2)
808 # Compare URIs
809 uri1 = butler.getURI(datasetTypeName, dataId1)
810 uri2 = butler.getURI(datasetTypeName, dataId2)
811 self.assertNotEqual(uri1, uri2)
813 # Now do a multi-dataset but single file ingest
814 metricFile = os.path.join(dataRoot, "detectors.yaml")
815 refs = []
816 for detector in (1, 2):
817 detector_name = f"detector_{detector}"
818 dataId = {"instrument": "DummyCamComp", "visit": 424, "detector": detector}
819 # Create a DatasetRef for ingest
820 refs.append(DatasetRef(datasetType, dataId, id=None))
822 datasets = []
823 datasets.append(FileDataset(path=metricFile, refs=refs, formatter=MultiDetectorFormatter))
825 butler.ingest(*datasets, transfer="copy", record_validation_info=False)
827 # Check that the datastore recorded no file size.
828 # Not all datastores can support this.
829 try:
830 infos = butler.datastore.getStoredItemsInfo(datasets[0].refs[0])
831 self.assertEqual(infos[0].file_size, -1)
832 except AttributeError:
833 pass
835 dataId1 = {"instrument": "DummyCamComp", "detector": 1, "visit": 424}
836 dataId2 = {"instrument": "DummyCamComp", "detector": 2, "visit": 424}
838 multi1 = butler.get(datasetTypeName, dataId1)
839 multi2 = butler.get(datasetTypeName, dataId2)
841 self.assertEqual(multi1, metrics1)
842 self.assertEqual(multi2, metrics2)
844 # Compare URIs
845 uri1 = butler.getURI(datasetTypeName, dataId1)
846 uri2 = butler.getURI(datasetTypeName, dataId2)
847 self.assertEqual(uri1, uri2, f"Cf. {uri1} with {uri2}")
849 # Test that removing one does not break the second
850 # This line will issue a warning log message for a ChainedDatastore
851 # that uses an InMemoryDatastore since in-memory can not ingest
852 # files.
853 butler.pruneDatasets([datasets[0].refs[0]], unstore=True, disassociate=False)
854 self.assertFalse(butler.datasetExists(datasetTypeName, dataId1))
855 self.assertTrue(butler.datasetExists(datasetTypeName, dataId2))
856 multi2b = butler.get(datasetTypeName, dataId2)
857 self.assertEqual(multi2, multi2b)
859 def testPruneCollections(self):
860 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
861 butler = Butler(self.tmpConfigFile, writeable=True)
862 # Load registry data with dimensions to hang datasets off of.
863 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
864 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
865 # Add some RUN-type collections.
866 run1 = "run1"
867 butler.registry.registerRun(run1)
868 run2 = "run2"
869 butler.registry.registerRun(run2)
870 # put some datasets. ref1 and ref2 have the same data ID, and are in
871 # different runs. ref3 has a different data ID.
872 metric = makeExampleMetrics()
873 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
874 datasetType = self.addDatasetType(
875 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
876 )
877 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
878 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
879 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
881 # Try to delete a RUN collection without purge, or with purge and not
882 # unstore.
883 with self.assertRaises(TypeError):
884 butler.pruneCollection(run1)
885 with self.assertRaises(TypeError):
886 butler.pruneCollection(run2, purge=True)
887 # Add a TAGGED collection and associate ref3 only into it.
888 tag1 = "tag1"
889 registered = butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
890 self.assertTrue(registered)
891 # Registering a second time should be allowed.
892 registered = butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
893 self.assertFalse(registered)
894 butler.registry.associate(tag1, [ref3])
895 # Add a CHAINED collection that searches run1 and then run2. It
896 # logically contains only ref1, because ref2 is shadowed due to them
897 # having the same data ID and dataset type.
898 chain1 = "chain1"
899 butler.registry.registerCollection(chain1, type=CollectionType.CHAINED)
900 butler.registry.setCollectionChain(chain1, [run1, run2])
901 # Try to delete RUN collections, which should fail with complete
902 # rollback because they're still referenced by the CHAINED
903 # collection.
904 with self.assertRaises(sqlalchemy.exc.IntegrityError):
905 butler.pruneCollection(run1, purge=True, unstore=True)
906 with self.assertRaises(sqlalchemy.exc.IntegrityError):
907 butler.pruneCollection(run2, purge=True, unstore=True)
908 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
909 existence = butler.datastore.mexists([ref1, ref2, ref3])
910 self.assertTrue(existence[ref1])
911 self.assertTrue(existence[ref2])
912 self.assertTrue(existence[ref3])
913 # Try to delete CHAINED and TAGGED collections with purge; should not
914 # work.
915 with self.assertRaises(TypeError):
916 butler.pruneCollection(tag1, purge=True, unstore=True)
917 with self.assertRaises(TypeError):
918 butler.pruneCollection(chain1, purge=True, unstore=True)
919 # Remove the tagged collection with unstore=False. This should not
920 # affect the datasets.
921 butler.pruneCollection(tag1)
922 with self.assertRaises(MissingCollectionError):
923 butler.registry.getCollectionType(tag1)
924 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
925 existence = butler.datastore.mexists([ref1, ref2, ref3])
926 self.assertTrue(existence[ref1])
927 self.assertTrue(existence[ref2])
928 self.assertTrue(existence[ref3])
929 # Add the tagged collection back in, and remove it with unstore=True.
930 # This should remove ref3 only from the datastore.
931 butler.registry.registerCollection(tag1, type=CollectionType.TAGGED)
932 butler.registry.associate(tag1, [ref3])
933 butler.pruneCollection(tag1, unstore=True)
934 with self.assertRaises(MissingCollectionError):
935 butler.registry.getCollectionType(tag1)
936 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
937 existence = butler.datastore.mexists([ref1, ref2, ref3])
938 self.assertTrue(existence[ref1])
939 self.assertTrue(existence[ref2])
940 self.assertFalse(existence[ref3])
941 # Delete the chain with unstore=False. The datasets should not be
942 # affected at all.
943 butler.pruneCollection(chain1)
944 with self.assertRaises(MissingCollectionError):
945 butler.registry.getCollectionType(chain1)
946 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
947 existence = butler.datastore.mexists([ref1, ref2, ref3])
948 self.assertTrue(existence[ref1])
949 self.assertTrue(existence[ref2])
950 self.assertFalse(existence[ref3])
951 # Redefine and then delete the chain with unstore=True. Only ref1
952 # should be unstored (ref3 has already been unstored, but otherwise
953 # would be now).
954 butler.registry.registerCollection(chain1, type=CollectionType.CHAINED)
955 butler.registry.setCollectionChain(chain1, [run1, run2])
956 butler.pruneCollection(chain1, unstore=True)
957 with self.assertRaises(MissingCollectionError):
958 butler.registry.getCollectionType(chain1)
959 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref1, ref2, ref3])
960 existence = butler.datastore.mexists([ref1, ref2, ref3])
961 self.assertFalse(existence[ref1])
962 self.assertTrue(existence[ref2])
963 self.assertFalse(existence[ref3])
964 # Remove run1. This removes ref1 and ref3 from the registry (they're
965 # already gone from the datastore, which is fine).
966 butler.pruneCollection(run1, purge=True, unstore=True)
967 with self.assertRaises(MissingCollectionError):
968 butler.registry.getCollectionType(run1)
969 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [ref2])
970 self.assertTrue(butler.datastore.exists(ref2))
971 # Remove run2. This removes ref2 from the registry and the datastore.
972 butler.pruneCollection(run2, purge=True, unstore=True)
973 with self.assertRaises(MissingCollectionError):
974 butler.registry.getCollectionType(run2)
975 self.assertCountEqual(set(butler.registry.queryDatasets(..., collections=...)), [])
977 # Now that the collections have been pruned we can remove the
978 # dataset type
979 butler.registry.removeDatasetType(datasetType.name)
981 def testPickle(self):
982 """Test pickle support."""
983 butler = Butler(self.tmpConfigFile, run=self.default_run)
984 butlerOut = pickle.loads(pickle.dumps(butler))
985 self.assertIsInstance(butlerOut, Butler)
986 self.assertEqual(butlerOut._config, butler._config)
987 self.assertEqual(butlerOut.collections, butler.collections)
988 self.assertEqual(butlerOut.run, butler.run)
990 def testGetDatasetTypes(self):
991 butler = Butler(self.tmpConfigFile, run=self.default_run)
992 dimensions = butler.registry.dimensions.extract(["instrument", "visit", "physical_filter"])
993 dimensionEntries = [
994 (
995 "instrument",
996 {"instrument": "DummyCam"},
997 {"instrument": "DummyHSC"},
998 {"instrument": "DummyCamComp"},
999 ),
1000 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1001 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1002 ]
1003 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1004 # Add needed Dimensions
1005 for args in dimensionEntries:
1006 butler.registry.insertDimensionData(*args)
1008 # When a DatasetType is added to the registry entries are not created
1009 # for components but querying them can return the components.
1010 datasetTypeNames = {"metric", "metric2", "metric4", "metric33", "pvi", "paramtest"}
1011 components = set()
1012 for datasetTypeName in datasetTypeNames:
1013 # Create and register a DatasetType
1014 self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1016 for componentName in storageClass.components:
1017 components.add(DatasetType.nameWithComponent(datasetTypeName, componentName))
1019 fromRegistry: set[DatasetType] = set()
1020 for parent_dataset_type in butler.registry.queryDatasetTypes():
1021 fromRegistry.add(parent_dataset_type)
1022 fromRegistry.update(parent_dataset_type.makeAllComponentDatasetTypes())
1023 self.assertEqual({d.name for d in fromRegistry}, datasetTypeNames | components)
1025 # Now that we have some dataset types registered, validate them
1026 butler.validateConfiguration(
1027 ignore=[
1028 "test_metric_comp",
1029 "metric3",
1030 "metric5",
1031 "calexp",
1032 "DummySC",
1033 "datasetType.component",
1034 "random_data",
1035 "random_data_2",
1036 ]
1037 )
1039 # Add a new datasetType that will fail template validation
1040 self.addDatasetType("test_metric_comp", dimensions, storageClass, butler.registry)
1041 if self.validationCanFail:
1042 with self.assertRaises(ValidationError):
1043 butler.validateConfiguration()
1045 # Rerun validation but with a subset of dataset type names
1046 butler.validateConfiguration(datasetTypeNames=["metric4"])
1048 # Rerun validation but ignore the bad datasetType
1049 butler.validateConfiguration(
1050 ignore=[
1051 "test_metric_comp",
1052 "metric3",
1053 "metric5",
1054 "calexp",
1055 "DummySC",
1056 "datasetType.component",
1057 "random_data",
1058 "random_data_2",
1059 ]
1060 )
1062 def testTransaction(self):
1063 butler = Butler(self.tmpConfigFile, run=self.default_run)
1064 datasetTypeName = "test_metric"
1065 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1066 dimensionEntries = (
1067 ("instrument", {"instrument": "DummyCam"}),
1068 ("physical_filter", {"instrument": "DummyCam", "name": "d-r", "band": "R"}),
1069 ("visit", {"instrument": "DummyCam", "id": 42, "name": "fortytwo", "physical_filter": "d-r"}),
1070 )
1071 storageClass = self.storageClassFactory.getStorageClass("StructuredData")
1072 metric = makeExampleMetrics()
1073 dataId = {"instrument": "DummyCam", "visit": 42}
1074 # Create and register a DatasetType
1075 datasetType = self.addDatasetType(datasetTypeName, dimensions, storageClass, butler.registry)
1076 with self.assertRaises(TransactionTestError):
1077 with butler.transaction():
1078 # Add needed Dimensions
1079 for args in dimensionEntries:
1080 butler.registry.insertDimensionData(*args)
1081 # Store a dataset
1082 ref = butler.put(metric, datasetTypeName, dataId)
1083 self.assertIsInstance(ref, DatasetRef)
1084 # Test getDirect
1085 metricOut = butler.getDirect(ref)
1086 self.assertEqual(metric, metricOut)
1087 # Test get
1088 metricOut = butler.get(datasetTypeName, dataId)
1089 self.assertEqual(metric, metricOut)
1090 # Check we can get components
1091 self.assertGetComponents(butler, ref, ("summary", "data", "output"), metric)
1092 raise TransactionTestError("This should roll back the entire transaction")
1093 with self.assertRaises(DataIdValueError, msg=f"Check can't expand DataId {dataId}"):
1094 butler.registry.expandDataId(dataId)
1095 # Should raise LookupError for missing data ID value
1096 with self.assertRaises(LookupError, msg=f"Check can't get by {datasetTypeName} and {dataId}"):
1097 butler.get(datasetTypeName, dataId)
1098 # Also check explicitly if Dataset entry is missing
1099 self.assertIsNone(butler.registry.findDataset(datasetType, dataId, collections=butler.collections))
1100 # Direct retrieval should not find the file in the Datastore
1101 with self.assertRaises(FileNotFoundError, msg=f"Check {ref} can't be retrieved directly"):
1102 butler.getDirect(ref)
1104 def testMakeRepo(self):
1105 """Test that we can write butler configuration to a new repository via
1106 the Butler.makeRepo interface and then instantiate a butler from the
1107 repo root.
1108 """
1109 # Do not run the test if we know this datastore configuration does
1110 # not support a file system root
1111 if self.fullConfigKey is None:
1112 return
1114 # create two separate directories
1115 root1 = tempfile.mkdtemp(dir=self.root)
1116 root2 = tempfile.mkdtemp(dir=self.root)
1118 butlerConfig = Butler.makeRepo(root1, config=Config(self.configFile))
1119 limited = Config(self.configFile)
1120 butler1 = Butler(butlerConfig)
1121 butlerConfig = Butler.makeRepo(root2, standalone=True, config=Config(self.configFile))
1122 full = Config(self.tmpConfigFile)
1123 butler2 = Butler(butlerConfig)
1124 # Butlers should have the same configuration regardless of whether
1125 # defaults were expanded.
1126 self.assertEqual(butler1._config, butler2._config)
1127 # Config files loaded directly should not be the same.
1128 self.assertNotEqual(limited, full)
1129 # Make sure "limited" doesn't have a few keys we know it should be
1130 # inheriting from defaults.
1131 self.assertIn(self.fullConfigKey, full)
1132 self.assertNotIn(self.fullConfigKey, limited)
1134 # Collections don't appear until something is put in them
1135 collections1 = set(butler1.registry.queryCollections())
1136 self.assertEqual(collections1, set())
1137 self.assertEqual(set(butler2.registry.queryCollections()), collections1)
1139 # Check that a config with no associated file name will not
1140 # work properly with relocatable Butler repo
1141 butlerConfig.configFile = None
1142 with self.assertRaises(ValueError):
1143 Butler(butlerConfig)
1145 with self.assertRaises(FileExistsError):
1146 Butler.makeRepo(self.root, standalone=True, config=Config(self.configFile), overwrite=False)
1148 def testStringification(self):
1149 butler = Butler(self.tmpConfigFile, run=self.default_run)
1150 butlerStr = str(butler)
1152 if self.datastoreStr is not None:
1153 for testStr in self.datastoreStr:
1154 self.assertIn(testStr, butlerStr)
1155 if self.registryStr is not None:
1156 self.assertIn(self.registryStr, butlerStr)
1158 datastoreName = butler.datastore.name
1159 if self.datastoreName is not None:
1160 for testStr in self.datastoreName:
1161 self.assertIn(testStr, datastoreName)
1163 def testButlerRewriteDataId(self):
1164 """Test that dataIds can be rewritten based on dimension records."""
1166 butler = Butler(self.tmpConfigFile, run=self.default_run)
1168 storageClass = self.storageClassFactory.getStorageClass("StructuredDataDict")
1169 datasetTypeName = "random_data"
1171 # Create dimension records.
1172 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1173 butler.registry.insertDimensionData(
1174 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1175 )
1176 butler.registry.insertDimensionData(
1177 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
1178 )
1180 dimensions = butler.registry.dimensions.extract(["instrument", "exposure"])
1181 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
1182 butler.registry.registerDatasetType(datasetType)
1184 n_exposures = 5
1185 dayobs = 20210530
1187 for i in range(n_exposures):
1188 butler.registry.insertDimensionData(
1189 "exposure",
1190 {
1191 "instrument": "DummyCamComp",
1192 "id": i,
1193 "obs_id": f"exp{i}",
1194 "seq_num": i,
1195 "day_obs": dayobs,
1196 "physical_filter": "d-r",
1197 },
1198 )
1200 # Write some data.
1201 for i in range(n_exposures):
1202 metric = {"something": i, "other": "metric", "list": [2 * x for x in range(i)]}
1204 # Use the seq_num for the put to test rewriting.
1205 dataId = {"seq_num": i, "day_obs": dayobs, "instrument": "DummyCamComp", "physical_filter": "d-r"}
1206 ref = butler.put(metric, datasetTypeName, dataId=dataId)
1208 # Check that the exposure is correct in the dataId
1209 self.assertEqual(ref.dataId["exposure"], i)
1211 # and check that we can get the dataset back with the same dataId
1212 new_metric = butler.get(datasetTypeName, dataId=dataId)
1213 self.assertEqual(new_metric, metric)
1216class FileDatastoreButlerTests(ButlerTests):
1217 """Common tests and specialization of ButlerTests for butlers backed
1218 by datastores that inherit from FileDatastore.
1219 """
1221 def checkFileExists(self, root, relpath):
1222 """Checks if file exists at a given path (relative to root).
1224 Test testPutTemplates verifies actual physical existance of the files
1225 in the requested location.
1226 """
1227 uri = ResourcePath(root, forceDirectory=True)
1228 return uri.join(relpath).exists()
1230 def testPutTemplates(self):
1231 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1232 butler = Butler(self.tmpConfigFile, run=self.default_run)
1234 # Add needed Dimensions
1235 butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
1236 butler.registry.insertDimensionData(
1237 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
1238 )
1239 butler.registry.insertDimensionData(
1240 "visit", {"instrument": "DummyCamComp", "id": 423, "name": "v423", "physical_filter": "d-r"}
1241 )
1242 butler.registry.insertDimensionData(
1243 "visit", {"instrument": "DummyCamComp", "id": 425, "name": "v425", "physical_filter": "d-r"}
1244 )
1246 # Create and store a dataset
1247 metric = makeExampleMetrics()
1249 # Create two almost-identical DatasetTypes (both will use default
1250 # template)
1251 dimensions = butler.registry.dimensions.extract(["instrument", "visit"])
1252 butler.registry.registerDatasetType(DatasetType("metric1", dimensions, storageClass))
1253 butler.registry.registerDatasetType(DatasetType("metric2", dimensions, storageClass))
1254 butler.registry.registerDatasetType(DatasetType("metric3", dimensions, storageClass))
1256 dataId1 = {"instrument": "DummyCamComp", "visit": 423}
1257 dataId2 = {"instrument": "DummyCamComp", "visit": 423, "physical_filter": "d-r"}
1259 # Put with exactly the data ID keys needed
1260 ref = butler.put(metric, "metric1", dataId1)
1261 uri = butler.getURI(ref)
1262 self.assertTrue(
1263 self.checkFileExists(
1264 butler.datastore.root, f"{self.default_run}/metric1/??#?/d-r/DummyCamComp_423.pickle"
1265 ),
1266 f"Checking existence of {uri}",
1267 )
1269 # Check the template based on dimensions
1270 butler.datastore.templates.validateTemplates([ref])
1272 # Put with extra data ID keys (physical_filter is an optional
1273 # dependency); should not change template (at least the way we're
1274 # defining them to behave now; the important thing is that they
1275 # must be consistent).
1276 ref = butler.put(metric, "metric2", dataId2)
1277 uri = butler.getURI(ref)
1278 self.assertTrue(
1279 self.checkFileExists(
1280 butler.datastore.root, f"{self.default_run}/metric2/d-r/DummyCamComp_v423.pickle"
1281 ),
1282 f"Checking existence of {uri}",
1283 )
1285 # Check the template based on dimensions
1286 butler.datastore.templates.validateTemplates([ref])
1288 # Use a template that has a typo in dimension record metadata.
1289 # Easier to test with a butler that has a ref with records attached.
1290 template = FileTemplate("a/{visit.name}/{id}_{visit.namex:?}.fits")
1291 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1292 path = template.format(ref)
1293 self.assertEqual(path, f"a/v423/{ref.id}_fits")
1295 template = FileTemplate("a/{visit.name}/{id}_{visit.namex}.fits")
1296 with self.assertRaises(KeyError):
1297 with self.assertLogs("lsst.daf.butler.core.fileTemplates", "INFO"):
1298 template.format(ref)
1300 # Now use a file template that will not result in unique filenames
1301 with self.assertRaises(FileTemplateValidationError):
1302 butler.put(metric, "metric3", dataId1)
1304 def testImportExport(self):
1305 # Run put/get tests just to create and populate a repo.
1306 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1307 self.runImportExportTest(storageClass)
1309 @unittest.expectedFailure
1310 def testImportExportVirtualComposite(self):
1311 # Run put/get tests just to create and populate a repo.
1312 storageClass = self.storageClassFactory.getStorageClass("StructuredComposite")
1313 self.runImportExportTest(storageClass)
1315 def runImportExportTest(self, storageClass):
1316 """This test does an export to a temp directory and an import back
1317 into a new temp directory repo. It does not assume a posix datastore"""
1318 exportButler = self.runPutGetTest(storageClass, "test_metric")
1319 print("Root:", exportButler.datastore.root)
1320 # Test that the repo actually has at least one dataset.
1321 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1322 self.assertGreater(len(datasets), 0)
1323 # Add a DimensionRecord that's unused by those datasets.
1324 skymapRecord = {"name": "example_skymap", "hash": (50).to_bytes(8, byteorder="little")}
1325 exportButler.registry.insertDimensionData("skymap", skymapRecord)
1326 # Export and then import datasets.
1327 with safeTestTempDir(TESTDIR) as exportDir:
1328 exportFile = os.path.join(exportDir, "exports.yaml")
1329 with exportButler.export(filename=exportFile, directory=exportDir, transfer="auto") as export:
1330 export.saveDatasets(datasets)
1331 # Export the same datasets again. This should quietly do
1332 # nothing because of internal deduplication, and it shouldn't
1333 # complain about being asked to export the "htm7" elements even
1334 # though there aren't any in these datasets or in the database.
1335 export.saveDatasets(datasets, elements=["htm7"])
1336 # Save one of the data IDs again; this should be harmless
1337 # because of internal deduplication.
1338 export.saveDataIds([datasets[0].dataId])
1339 # Save some dimension records directly.
1340 export.saveDimensionData("skymap", [skymapRecord])
1341 self.assertTrue(os.path.exists(exportFile))
1342 with safeTestTempDir(TESTDIR) as importDir:
1343 # We always want this to be a local posix butler
1344 Butler.makeRepo(importDir, config=Config(os.path.join(TESTDIR, "config/basic/butler.yaml")))
1345 # Calling script.butlerImport tests the implementation of the
1346 # butler command line interface "import" subcommand. Functions
1347 # in the script folder are generally considered protected and
1348 # should not be used as public api.
1349 with open(exportFile, "r") as f:
1350 script.butlerImport(
1351 importDir,
1352 export_file=f,
1353 directory=exportDir,
1354 transfer="auto",
1355 skip_dimensions=None,
1356 reuse_ids=False,
1357 )
1358 importButler = Butler(importDir, run=self.default_run)
1359 for ref in datasets:
1360 with self.subTest(ref=ref):
1361 # Test for existence by passing in the DatasetType and
1362 # data ID separately, to avoid lookup by dataset_id.
1363 self.assertTrue(importButler.datasetExists(ref.datasetType, ref.dataId))
1364 self.assertEqual(
1365 list(importButler.registry.queryDimensionRecords("skymap")),
1366 [importButler.registry.dimensions["skymap"].RecordClass(**skymapRecord)],
1367 )
1369 def testRemoveRuns(self):
1370 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1371 butler = Butler(self.tmpConfigFile, writeable=True)
1372 # Load registry data with dimensions to hang datasets off of.
1373 registryDataDir = os.path.normpath(os.path.join(os.path.dirname(__file__), "data", "registry"))
1374 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1375 # Add some RUN-type collection.
1376 run1 = "run1"
1377 butler.registry.registerRun(run1)
1378 run2 = "run2"
1379 butler.registry.registerRun(run2)
1380 # put a dataset in each
1381 metric = makeExampleMetrics()
1382 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1383 datasetType = self.addDatasetType(
1384 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1385 )
1386 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1387 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1388 uri1 = butler.getURI(ref1, collections=[run1])
1389 uri2 = butler.getURI(ref2, collections=[run2])
1390 # Remove from both runs with different values for unstore.
1391 butler.removeRuns([run1], unstore=True)
1392 butler.removeRuns([run2], unstore=False)
1393 # Should be nothing in registry for either one, and datastore should
1394 # not think either exists.
1395 with self.assertRaises(MissingCollectionError):
1396 butler.registry.getCollectionType(run1)
1397 with self.assertRaises(MissingCollectionError):
1398 butler.registry.getCollectionType(run2)
1399 self.assertFalse(butler.datastore.exists(ref1))
1400 self.assertFalse(butler.datastore.exists(ref2))
1401 # The ref we unstored should be gone according to the URI, but the
1402 # one we forgot should still be around.
1403 self.assertFalse(uri1.exists())
1404 self.assertTrue(uri2.exists())
1407class PosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1408 """PosixDatastore specialization of a butler"""
1410 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1411 fullConfigKey = ".datastore.formatters"
1412 validationCanFail = True
1413 datastoreStr = ["/tmp"]
1414 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1415 registryStr = "/gen3.sqlite3"
1417 def testPathConstructor(self):
1418 """Independent test of constructor using PathLike."""
1419 butler = Butler(self.tmpConfigFile, run=self.default_run)
1420 self.assertIsInstance(butler, Butler)
1422 # And again with a Path object with the butler yaml
1423 path = pathlib.Path(self.tmpConfigFile)
1424 butler = Butler(path, writeable=False)
1425 self.assertIsInstance(butler, Butler)
1427 # And again with a Path object without the butler yaml
1428 # (making sure we skip it if the tmp config doesn't end
1429 # in butler.yaml -- which is the case for a subclass)
1430 if self.tmpConfigFile.endswith("butler.yaml"):
1431 path = pathlib.Path(os.path.dirname(self.tmpConfigFile))
1432 butler = Butler(path, writeable=False)
1433 self.assertIsInstance(butler, Butler)
1435 def testExportTransferCopy(self):
1436 """Test local export using all transfer modes"""
1437 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1438 exportButler = self.runPutGetTest(storageClass, "test_metric")
1439 # Test that the repo actually has at least one dataset.
1440 datasets = list(exportButler.registry.queryDatasets(..., collections=...))
1441 self.assertGreater(len(datasets), 0)
1442 uris = [exportButler.getURI(d) for d in datasets]
1443 datastoreRoot = exportButler.datastore.root
1445 pathsInStore = [uri.relative_to(datastoreRoot) for uri in uris]
1447 for path in pathsInStore:
1448 # Assume local file system
1449 self.assertTrue(self.checkFileExists(datastoreRoot, path), f"Checking path {path}")
1451 for transfer in ("copy", "link", "symlink", "relsymlink"):
1452 with safeTestTempDir(TESTDIR) as exportDir:
1453 with exportButler.export(directory=exportDir, format="yaml", transfer=transfer) as export:
1454 export.saveDatasets(datasets)
1455 for path in pathsInStore:
1456 self.assertTrue(
1457 self.checkFileExists(exportDir, path),
1458 f"Check that mode {transfer} exported files",
1459 )
1461 def testPruneDatasets(self):
1462 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1463 butler = Butler(self.tmpConfigFile, writeable=True)
1464 # Load registry data with dimensions to hang datasets off of.
1465 registryDataDir = os.path.normpath(os.path.join(TESTDIR, "data", "registry"))
1466 butler.import_(filename=os.path.join(registryDataDir, "base.yaml"))
1467 # Add some RUN-type collections.
1468 run1 = "run1"
1469 butler.registry.registerRun(run1)
1470 run2 = "run2"
1471 butler.registry.registerRun(run2)
1472 # put some datasets. ref1 and ref2 have the same data ID, and are in
1473 # different runs. ref3 has a different data ID.
1474 metric = makeExampleMetrics()
1475 dimensions = butler.registry.dimensions.extract(["instrument", "physical_filter"])
1476 datasetType = self.addDatasetType(
1477 "prune_collections_test_dataset", dimensions, storageClass, butler.registry
1478 )
1479 ref1 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run1)
1480 ref2 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-G"}, run=run2)
1481 ref3 = butler.put(metric, datasetType, {"instrument": "Cam1", "physical_filter": "Cam1-R1"}, run=run1)
1483 # Simple prune.
1484 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1485 with self.assertRaises(LookupError):
1486 butler.datasetExists(ref1.datasetType, ref1.dataId, collections=run1)
1488 # Put data back.
1489 ref1 = butler.put(metric, ref1.unresolved(), run=run1)
1490 ref2 = butler.put(metric, ref2.unresolved(), run=run2)
1491 ref3 = butler.put(metric, ref3.unresolved(), run=run1)
1493 # Check that in normal mode, deleting the record will lead to
1494 # trash not touching the file.
1495 uri1 = butler.datastore.getURI(ref1)
1496 butler.datastore.bridge.moveToTrash([ref1], transaction=None) # Update the dataset_location table
1497 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref1.id})
1498 butler.datastore.trash(ref1)
1499 butler.datastore.emptyTrash()
1500 self.assertTrue(uri1.exists())
1501 uri1.remove() # Clean it up.
1503 # Simulate execution butler setup by deleting the datastore
1504 # record but keeping the file around and trusting.
1505 butler.datastore.trustGetRequest = True
1506 uri2 = butler.datastore.getURI(ref2)
1507 uri3 = butler.datastore.getURI(ref3)
1508 self.assertTrue(uri2.exists())
1509 self.assertTrue(uri3.exists())
1511 # Remove the datastore record.
1512 butler.datastore.bridge.moveToTrash([ref2], transaction=None) # Update the dataset_location table
1513 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref2.id})
1514 self.assertTrue(uri2.exists())
1515 butler.datastore.trash([ref2, ref3])
1516 # Immediate removal for ref2 file
1517 self.assertFalse(uri2.exists())
1518 # But ref3 has to wait for the empty.
1519 self.assertTrue(uri3.exists())
1520 butler.datastore.emptyTrash()
1521 self.assertFalse(uri3.exists())
1523 # Clear out the datasets from registry.
1524 butler.pruneDatasets([ref1, ref2, ref3], purge=True, unstore=True)
1526 def testPytypeCoercion(self):
1527 """Test python type coercion on Butler.get and put."""
1529 # Store some data with the normal example storage class.
1530 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1531 datasetTypeName = "test_metric"
1532 butler = self.runPutGetTest(storageClass, datasetTypeName)
1534 dataId = {"instrument": "DummyCamComp", "visit": 423}
1535 metric = butler.get(datasetTypeName, dataId=dataId)
1536 self.assertEqual(get_full_type_name(metric), "lsst.daf.butler.tests.MetricsExample")
1538 datasetType_ori = butler.registry.getDatasetType(datasetTypeName)
1539 self.assertEqual(datasetType_ori.storageClass.name, "StructuredDataNoComponents")
1541 # Now need to hack the registry dataset type definition.
1542 # There is no API for this.
1543 manager = butler.registry._managers.datasets
1544 manager._db.update(
1545 manager._static.dataset_type,
1546 {"name": datasetTypeName},
1547 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataNoComponentsModel"},
1548 )
1550 # Force reset of dataset type cache
1551 butler.registry.refresh()
1553 datasetType_new = butler.registry.getDatasetType(datasetTypeName)
1554 self.assertEqual(datasetType_new.name, datasetType_ori.name)
1555 self.assertEqual(datasetType_new.storageClass.name, "StructuredDataNoComponentsModel")
1557 metric_model = butler.get(datasetTypeName, dataId=dataId)
1558 self.assertNotEqual(type(metric_model), type(metric))
1559 self.assertEqual(get_full_type_name(metric_model), "lsst.daf.butler.tests.MetricsExampleModel")
1561 # Put the model and read it back to show that everything now
1562 # works as normal.
1563 metric_ref = butler.put(metric_model, datasetTypeName, dataId=dataId, visit=424)
1564 metric_model_new = butler.get(metric_ref)
1565 self.assertEqual(metric_model_new, metric_model)
1567 # Hack the storage class again to something that will fail on the
1568 # get with no conversion class.
1569 manager._db.update(
1570 manager._static.dataset_type,
1571 {"name": datasetTypeName},
1572 {datasetTypeName: datasetTypeName, "storage_class": "StructuredDataListYaml"},
1573 )
1574 butler.registry.refresh()
1576 with self.assertRaises(ValueError):
1577 butler.get(datasetTypeName, dataId=dataId)
1580@unittest.skipUnless(testing is not None, "testing.postgresql module not found")
1581class PostgresPosixDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1582 """PosixDatastore specialization of a butler using Postgres"""
1584 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1585 fullConfigKey = ".datastore.formatters"
1586 validationCanFail = True
1587 datastoreStr = ["/tmp"]
1588 datastoreName = [f"FileDatastore@{BUTLER_ROOT_TAG}"]
1589 registryStr = "PostgreSQL@test"
1591 @staticmethod
1592 def _handler(postgresql):
1593 engine = sqlalchemy.engine.create_engine(postgresql.url())
1594 with engine.begin() as connection:
1595 connection.execute(sqlalchemy.text("CREATE EXTENSION btree_gist;"))
1597 @classmethod
1598 def setUpClass(cls):
1599 # Create the postgres test server.
1600 cls.postgresql = testing.postgresql.PostgresqlFactory(
1601 cache_initialized_db=True, on_initialized=cls._handler
1602 )
1603 super().setUpClass()
1605 @classmethod
1606 def tearDownClass(cls):
1607 # Clean up any lingering SQLAlchemy engines/connections
1608 # so they're closed before we shut down the server.
1609 gc.collect()
1610 cls.postgresql.clear_cache()
1611 super().tearDownClass()
1613 def setUp(self):
1614 self.server = self.postgresql()
1616 # Need to add a registry section to the config.
1617 self._temp_config = False
1618 config = Config(self.configFile)
1619 config["registry", "db"] = self.server.url()
1620 with tempfile.NamedTemporaryFile("w", suffix=".yaml", delete=False) as fh:
1621 config.dump(fh)
1622 self.configFile = fh.name
1623 self._temp_config = True
1624 super().setUp()
1626 def tearDown(self):
1627 self.server.stop()
1628 if self._temp_config and os.path.exists(self.configFile):
1629 os.remove(self.configFile)
1630 super().tearDown()
1632 def testMakeRepo(self):
1633 # The base class test assumes that it's using sqlite and assumes
1634 # the config file is acceptable to sqlite.
1635 raise unittest.SkipTest("Postgres config is not compatible with this test.")
1638class InMemoryDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1639 """InMemoryDatastore specialization of a butler"""
1641 configFile = os.path.join(TESTDIR, "config/basic/butler-inmemory.yaml")
1642 fullConfigKey = None
1643 useTempRoot = False
1644 validationCanFail = False
1645 datastoreStr = ["datastore='InMemory"]
1646 datastoreName = ["InMemoryDatastore@"]
1647 registryStr = "/gen3.sqlite3"
1649 def testIngest(self):
1650 pass
1653class ChainedDatastoreButlerTestCase(ButlerTests, unittest.TestCase):
1654 """PosixDatastore specialization"""
1656 configFile = os.path.join(TESTDIR, "config/basic/butler-chained.yaml")
1657 fullConfigKey = ".datastore.datastores.1.formatters"
1658 validationCanFail = True
1659 datastoreStr = ["datastore='InMemory", "/FileDatastore_1/,", "/FileDatastore_2/'"]
1660 datastoreName = [
1661 "InMemoryDatastore@",
1662 f"FileDatastore@{BUTLER_ROOT_TAG}/FileDatastore_1",
1663 "SecondDatastore",
1664 ]
1665 registryStr = "/gen3.sqlite3"
1668class ButlerExplicitRootTestCase(PosixDatastoreButlerTestCase):
1669 """Test that a yaml file in one location can refer to a root in another."""
1671 datastoreStr = ["dir1"]
1672 # Disable the makeRepo test since we are deliberately not using
1673 # butler.yaml as the config name.
1674 fullConfigKey = None
1676 def setUp(self):
1677 self.root = makeTestTempDir(TESTDIR)
1679 # Make a new repository in one place
1680 self.dir1 = os.path.join(self.root, "dir1")
1681 Butler.makeRepo(self.dir1, config=Config(self.configFile))
1683 # Move the yaml file to a different place and add a "root"
1684 self.dir2 = os.path.join(self.root, "dir2")
1685 os.makedirs(self.dir2, exist_ok=True)
1686 configFile1 = os.path.join(self.dir1, "butler.yaml")
1687 config = Config(configFile1)
1688 config["root"] = self.dir1
1689 configFile2 = os.path.join(self.dir2, "butler2.yaml")
1690 config.dumpToUri(configFile2)
1691 os.remove(configFile1)
1692 self.tmpConfigFile = configFile2
1694 def testFileLocations(self):
1695 self.assertNotEqual(self.dir1, self.dir2)
1696 self.assertTrue(os.path.exists(os.path.join(self.dir2, "butler2.yaml")))
1697 self.assertFalse(os.path.exists(os.path.join(self.dir1, "butler.yaml")))
1698 self.assertTrue(os.path.exists(os.path.join(self.dir1, "gen3.sqlite3")))
1701class ButlerMakeRepoOutfileTestCase(ButlerPutGetTests, unittest.TestCase):
1702 """Test that a config file created by makeRepo outside of repo works."""
1704 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1706 def setUp(self):
1707 self.root = makeTestTempDir(TESTDIR)
1708 self.root2 = makeTestTempDir(TESTDIR)
1710 self.tmpConfigFile = os.path.join(self.root2, "different.yaml")
1711 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1713 def tearDown(self):
1714 if os.path.exists(self.root2):
1715 shutil.rmtree(self.root2, ignore_errors=True)
1716 super().tearDown()
1718 def testConfigExistence(self):
1719 c = Config(self.tmpConfigFile)
1720 uri_config = ResourcePath(c["root"])
1721 uri_expected = ResourcePath(self.root, forceDirectory=True)
1722 self.assertEqual(uri_config.geturl(), uri_expected.geturl())
1723 self.assertNotIn(":", uri_config.path, "Check for URI concatenated with normal path")
1725 def testPutGet(self):
1726 storageClass = self.storageClassFactory.getStorageClass("StructuredDataNoComponents")
1727 self.runPutGetTest(storageClass, "test_metric")
1730class ButlerMakeRepoOutfileDirTestCase(ButlerMakeRepoOutfileTestCase):
1731 """Test that a config file created by makeRepo outside of repo works."""
1733 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1735 def setUp(self):
1736 self.root = makeTestTempDir(TESTDIR)
1737 self.root2 = makeTestTempDir(TESTDIR)
1739 self.tmpConfigFile = self.root2
1740 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1742 def testConfigExistence(self):
1743 # Append the yaml file else Config constructor does not know the file
1744 # type.
1745 self.tmpConfigFile = os.path.join(self.tmpConfigFile, "butler.yaml")
1746 super().testConfigExistence()
1749class ButlerMakeRepoOutfileUriTestCase(ButlerMakeRepoOutfileTestCase):
1750 """Test that a config file created by makeRepo outside of repo works."""
1752 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
1754 def setUp(self):
1755 self.root = makeTestTempDir(TESTDIR)
1756 self.root2 = makeTestTempDir(TESTDIR)
1758 self.tmpConfigFile = ResourcePath(os.path.join(self.root2, "something.yaml")).geturl()
1759 Butler.makeRepo(self.root, config=Config(self.configFile), outfile=self.tmpConfigFile)
1762@unittest.skipIf(not boto3, "Warning: boto3 AWS SDK not found!")
1763class S3DatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1764 """S3Datastore specialization of a butler; an S3 storage Datastore +
1765 a local in-memory SqlRegistry.
1766 """
1768 configFile = os.path.join(TESTDIR, "config/basic/butler-s3store.yaml")
1769 fullConfigKey = None
1770 validationCanFail = True
1772 bucketName = "anybucketname"
1773 """Name of the Bucket that will be used in the tests. The name is read from
1774 the config file used with the tests during set-up.
1775 """
1777 root = "butlerRoot/"
1778 """Root repository directory expected to be used in case useTempRoot=False.
1779 Otherwise the root is set to a 20 characters long randomly generated string
1780 during set-up.
1781 """
1783 datastoreStr = [f"datastore={root}"]
1784 """Contains all expected root locations in a format expected to be
1785 returned by Butler stringification.
1786 """
1788 datastoreName = ["FileDatastore@s3://{bucketName}/{root}"]
1789 """The expected format of the S3 Datastore string."""
1791 registryStr = "/gen3.sqlite3"
1792 """Expected format of the Registry string."""
1794 mock_s3 = mock_s3()
1795 """The mocked s3 interface from moto."""
1797 def genRoot(self):
1798 """Returns a random string of len 20 to serve as a root
1799 name for the temporary bucket repo.
1801 This is equivalent to tempfile.mkdtemp as this is what self.root
1802 becomes when useTempRoot is True.
1803 """
1804 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1805 return rndstr + "/"
1807 def setUp(self):
1808 config = Config(self.configFile)
1809 uri = ResourcePath(config[".datastore.datastore.root"])
1810 self.bucketName = uri.netloc
1812 # Enable S3 mocking of tests.
1813 self.mock_s3.start()
1815 # set up some fake credentials if they do not exist
1816 self.usingDummyCredentials = setAwsEnvCredentials()
1818 if self.useTempRoot:
1819 self.root = self.genRoot()
1820 rooturi = f"s3://{self.bucketName}/{self.root}"
1821 config.update({"datastore": {"datastore": {"root": rooturi}}})
1823 # need local folder to store registry database
1824 self.reg_dir = makeTestTempDir(TESTDIR)
1825 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
1827 # MOTO needs to know that we expect Bucket bucketname to exist
1828 # (this used to be the class attribute bucketName)
1829 s3 = boto3.resource("s3")
1830 s3.create_bucket(Bucket=self.bucketName)
1832 self.datastoreStr = f"datastore={self.root}"
1833 self.datastoreName = [f"FileDatastore@{rooturi}"]
1834 Butler.makeRepo(rooturi, config=config, forceConfigRoot=False)
1835 self.tmpConfigFile = posixpath.join(rooturi, "butler.yaml")
1837 def tearDown(self):
1838 s3 = boto3.resource("s3")
1839 bucket = s3.Bucket(self.bucketName)
1840 try:
1841 bucket.objects.all().delete()
1842 except botocore.exceptions.ClientError as e:
1843 if e.response["Error"]["Code"] == "404":
1844 # the key was not reachable - pass
1845 pass
1846 else:
1847 raise
1849 bucket = s3.Bucket(self.bucketName)
1850 bucket.delete()
1852 # Stop the S3 mock.
1853 self.mock_s3.stop()
1855 # unset any potentially set dummy credentials
1856 if self.usingDummyCredentials:
1857 unsetAwsEnvCredentials()
1859 if self.reg_dir is not None and os.path.exists(self.reg_dir):
1860 shutil.rmtree(self.reg_dir, ignore_errors=True)
1862 if self.useTempRoot and os.path.exists(self.root):
1863 shutil.rmtree(self.root, ignore_errors=True)
1865 super().tearDown()
1868@unittest.skipIf(WsgiDAVApp is None, "Warning: wsgidav/cheroot not found!")
1869class WebdavDatastoreButlerTestCase(FileDatastoreButlerTests, unittest.TestCase):
1870 """WebdavDatastore specialization of a butler; a Webdav storage Datastore +
1871 a local in-memory SqlRegistry.
1872 """
1874 configFile = os.path.join(TESTDIR, "config/basic/butler-webdavstore.yaml")
1875 fullConfigKey = None
1876 validationCanFail = True
1878 serverName = "localhost"
1879 """Name of the server that will be used in the tests.
1880 """
1882 portNumber = 8080
1883 """Port on which the webdav server listens. Automatically chosen
1884 at setUpClass via the _getfreeport() method
1885 """
1887 root = "butlerRoot/"
1888 """Root repository directory expected to be used in case useTempRoot=False.
1889 Otherwise the root is set to a 20 characters long randomly generated string
1890 during set-up.
1891 """
1893 datastoreStr = [f"datastore={root}"]
1894 """Contains all expected root locations in a format expected to be
1895 returned by Butler stringification.
1896 """
1898 datastoreName = ["FileDatastore@https://{serverName}/{root}"]
1899 """The expected format of the WebdavDatastore string."""
1901 registryStr = "/gen3.sqlite3"
1902 """Expected format of the Registry string."""
1904 serverThread = None
1905 """Thread in which the local webdav server will run"""
1907 stopWebdavServer = False
1908 """This flag will cause the webdav server to
1909 gracefully shut down when True
1910 """
1912 def genRoot(self):
1913 """Returns a random string of len 20 to serve as a root
1914 name for the temporary bucket repo.
1916 This is equivalent to tempfile.mkdtemp as this is what self.root
1917 becomes when useTempRoot is True.
1918 """
1919 rndstr = "".join(random.choice(string.ascii_uppercase + string.digits) for _ in range(20))
1920 return rndstr + "/"
1922 @classmethod
1923 def setUpClass(cls):
1924 # Do the same as inherited class
1925 cls.storageClassFactory = StorageClassFactory()
1926 cls.storageClassFactory.addFromConfig(cls.configFile)
1928 cls.portNumber = cls._getfreeport()
1929 # Run a local webdav server on which tests will be run
1930 cls.serverThread = Thread(
1931 target=cls._serveWebdav, args=(cls, cls.portNumber, lambda: cls.stopWebdavServer), daemon=True
1932 )
1933 cls.serverThread.start()
1934 # Wait for it to start
1935 time.sleep(3)
1937 @classmethod
1938 def tearDownClass(cls):
1939 # Ask for graceful shut down of the webdav server
1940 cls.stopWebdavServer = True
1941 # Wait for the thread to exit
1942 cls.serverThread.join()
1943 super().tearDownClass()
1945 def setUp(self):
1946 config = Config(self.configFile)
1948 if self.useTempRoot:
1949 self.root = self.genRoot()
1950 self.rooturi = f"http://{self.serverName}:{self.portNumber}/{self.root}"
1951 config.update({"datastore": {"datastore": {"root": self.rooturi}}})
1953 # need local folder to store registry database
1954 self.reg_dir = makeTestTempDir(TESTDIR)
1955 config["registry", "db"] = f"sqlite:///{self.reg_dir}/gen3.sqlite3"
1957 self.datastoreStr = f"datastore={self.root}"
1958 self.datastoreName = [f"FileDatastore@{self.rooturi}"]
1960 if not _is_webdav_endpoint(self.rooturi):
1961 raise OSError("Webdav server not running properly: cannot run tests.")
1963 Butler.makeRepo(self.rooturi, config=config, forceConfigRoot=False)
1964 self.tmpConfigFile = posixpath.join(self.rooturi, "butler.yaml")
1966 def tearDown(self):
1967 # Clear temporary directory
1968 ResourcePath(self.rooturi).remove()
1969 ResourcePath(self.rooturi).session.close()
1971 if self.reg_dir is not None and os.path.exists(self.reg_dir):
1972 shutil.rmtree(self.reg_dir, ignore_errors=True)
1974 if self.useTempRoot and os.path.exists(self.root):
1975 shutil.rmtree(self.root, ignore_errors=True)
1977 super().tearDown()
1979 def _serveWebdav(self, port: int, stopWebdavServer):
1980 """Starts a local webdav-compatible HTTP server,
1981 Listening on http://localhost:port
1982 This server only runs when this test class is instantiated,
1983 and then shuts down. Must be started is a separate thread.
1985 Parameters
1986 ----------
1987 port : `int`
1988 The port number on which the server should listen
1989 """
1990 root_path = gettempdir()
1992 config = {
1993 "host": "0.0.0.0",
1994 "port": port,
1995 "provider_mapping": {"/": root_path},
1996 "http_authenticator": {"domain_controller": None},
1997 "simple_dc": {"user_mapping": {"*": True}},
1998 "verbose": 0,
1999 }
2000 app = WsgiDAVApp(config)
2002 server_args = {
2003 "bind_addr": (config["host"], config["port"]),
2004 "wsgi_app": app,
2005 }
2006 server = wsgi.Server(**server_args)
2007 server.prepare()
2009 try:
2010 # Start the actual server in a separate thread
2011 t = Thread(target=server.serve, daemon=True)
2012 t.start()
2013 # watch stopWebdavServer, and gracefully
2014 # shut down the server when True
2015 while True:
2016 if stopWebdavServer():
2017 break
2018 time.sleep(1)
2019 except KeyboardInterrupt:
2020 print("Caught Ctrl-C, shutting down...")
2021 finally:
2022 server.stop()
2023 t.join()
2025 def _getfreeport():
2026 """
2027 Determines a free port using sockets.
2028 """
2029 free_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
2030 free_socket.bind(("127.0.0.1", 0))
2031 free_socket.listen()
2032 port = free_socket.getsockname()[1]
2033 free_socket.close()
2034 return port
2037class PosixDatastoreTransfers(unittest.TestCase):
2038 """Test data transfers between butlers.
2040 Test for different managers. UUID to UUID and integer to integer are
2041 tested. UUID to integer is not supported since we do not currently
2042 want to allow that. Integer to UUID is supported with the caveat
2043 that UUID4 will be generated and this will be incorrect for raw
2044 dataset types. The test ignores that.
2045 """
2047 configFile = os.path.join(TESTDIR, "config/basic/butler.yaml")
2049 @classmethod
2050 def setUpClass(cls):
2051 cls.storageClassFactory = StorageClassFactory()
2052 cls.storageClassFactory.addFromConfig(cls.configFile)
2054 def setUp(self):
2055 self.root = makeTestTempDir(TESTDIR)
2056 self.config = Config(self.configFile)
2058 def tearDown(self):
2059 removeTestTempDir(self.root)
2061 def create_butler(self, manager, label):
2062 config = Config(self.configFile)
2063 config["registry", "managers", "datasets"] = manager
2064 return Butler(Butler.makeRepo(f"{self.root}/butler{label}", config=config), writeable=True)
2066 def create_butlers(self, manager1, manager2):
2067 self.source_butler = self.create_butler(manager1, "1")
2068 self.target_butler = self.create_butler(manager2, "2")
2070 def testTransferUuidToUuid(self):
2071 self.create_butlers(
2072 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2073 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2074 )
2075 # Setting id_gen_map should have no effect here
2076 self.assertButlerTransfers(id_gen_map={"random_data_2": DatasetIdGenEnum.DATAID_TYPE})
2078 def testTransferIntToInt(self):
2079 with self.assertWarns(FutureWarning):
2080 self.create_butlers(
2081 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManager",
2082 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManager",
2083 )
2084 # int dataset ID only allows UNIQUE
2085 self.assertButlerTransfers()
2087 def testTransferIntToUuid(self):
2088 with self.assertWarns(FutureWarning):
2089 self.create_butlers(
2090 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManager",
2091 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2092 )
2093 self.assertButlerTransfers(id_gen_map={"random_data_2": DatasetIdGenEnum.DATAID_TYPE})
2095 def testTransferMissing(self):
2096 """Test transfers where datastore records are missing.
2098 This is how execution butler works.
2099 """
2100 self.create_butlers(
2101 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2102 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2103 )
2105 # Configure the source butler to allow trust.
2106 self.source_butler.datastore.trustGetRequest = True
2108 self.assertButlerTransfers(purge=True)
2110 def testTransferMissingDisassembly(self):
2111 """Test transfers where datastore records are missing.
2113 This is how execution butler works.
2114 """
2115 self.create_butlers(
2116 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2117 "lsst.daf.butler.registry.datasets.byDimensions.ByDimensionsDatasetRecordStorageManagerUUID",
2118 )
2120 # Configure the source butler to allow trust.
2121 self.source_butler.datastore.trustGetRequest = True
2123 # Test disassembly.
2124 self.assertButlerTransfers(purge=True, storageClassName="StructuredComposite")
2126 def assertButlerTransfers(self, id_gen_map=None, purge=False, storageClassName="StructuredData"):
2127 """Test that a run can be transferred to another butler."""
2129 storageClass = self.storageClassFactory.getStorageClass(storageClassName)
2130 datasetTypeName = "random_data"
2132 # Test will create 3 collections and we will want to transfer
2133 # two of those three.
2134 runs = ["run1", "run2", "other"]
2136 # Also want to use two different dataset types to ensure that
2137 # grouping works.
2138 datasetTypeNames = ["random_data", "random_data_2"]
2140 # Create the run collections in the source butler.
2141 for run in runs:
2142 self.source_butler.registry.registerCollection(run, CollectionType.RUN)
2144 # Create dimensions in source butler.
2145 n_exposures = 30
2146 self.source_butler.registry.insertDimensionData("instrument", {"name": "DummyCamComp"})
2147 self.source_butler.registry.insertDimensionData(
2148 "physical_filter", {"instrument": "DummyCamComp", "name": "d-r", "band": "R"}
2149 )
2150 self.source_butler.registry.insertDimensionData(
2151 "detector", {"instrument": "DummyCamComp", "id": 1, "full_name": "det1"}
2152 )
2154 for i in range(n_exposures):
2155 self.source_butler.registry.insertDimensionData(
2156 "exposure",
2157 {"instrument": "DummyCamComp", "id": i, "obs_id": f"exp{i}", "physical_filter": "d-r"},
2158 )
2160 # Create dataset types in the source butler.
2161 dimensions = self.source_butler.registry.dimensions.extract(["instrument", "exposure"])
2162 for datasetTypeName in datasetTypeNames:
2163 datasetType = DatasetType(datasetTypeName, dimensions, storageClass)
2164 self.source_butler.registry.registerDatasetType(datasetType)
2166 # Write a dataset to an unrelated run -- this will ensure that
2167 # we are rewriting integer dataset ids in the target if necessary.
2168 # Will not be relevant for UUID.
2169 run = "distraction"
2170 butler = Butler(butler=self.source_butler, run=run)
2171 butler.put(
2172 makeExampleMetrics(),
2173 datasetTypeName,
2174 exposure=1,
2175 instrument="DummyCamComp",
2176 physical_filter="d-r",
2177 )
2179 # Write some example metrics to the source
2180 butler = Butler(butler=self.source_butler)
2182 # Set of DatasetRefs that should be in the list of refs to transfer
2183 # but which will not be transferred.
2184 deleted = set()
2186 n_expected = 20 # Number of datasets expected to be transferred
2187 source_refs = []
2188 for i in range(n_exposures):
2189 # Put a third of datasets into each collection, only retain
2190 # two thirds.
2191 index = i % 3
2192 run = runs[index]
2193 datasetTypeName = datasetTypeNames[i % 2]
2195 metric_data = {
2196 "summary": {"counter": i},
2197 "output": {"text": "metric"},
2198 "data": [2 * x for x in range(i)],
2199 }
2200 metric = MetricsExample(**metric_data)
2201 dataId = {"exposure": i, "instrument": "DummyCamComp", "physical_filter": "d-r"}
2202 ref = butler.put(metric, datasetTypeName, dataId=dataId, run=run)
2204 # Remove the datastore record using low-level API
2205 if purge:
2206 # Remove records for a fraction.
2207 if index == 1:
2209 # For one of these delete the file as well.
2210 # This allows the "missing" code to filter the
2211 # file out.
2212 if not deleted:
2213 primary, uris = butler.datastore.getURIs(ref)
2214 if primary:
2215 primary.remove()
2216 for uri in uris.values():
2217 uri.remove()
2218 n_expected -= 1
2219 deleted.add(ref)
2221 # Remove the datastore record.
2222 butler.datastore._table.delete(["dataset_id"], {"dataset_id": ref.id})
2224 if index < 2:
2225 source_refs.append(ref)
2226 if ref not in deleted:
2227 new_metric = butler.get(ref.unresolved(), collections=run)
2228 self.assertEqual(new_metric, metric)
2230 # Create some bad dataset types to ensure we check for inconsistent
2231 # definitions.
2232 badStorageClass = self.storageClassFactory.getStorageClass("StructuredDataList")
2233 for datasetTypeName in datasetTypeNames:
2234 datasetType = DatasetType(datasetTypeName, dimensions, badStorageClass)
2235 self.target_butler.registry.registerDatasetType(datasetType)
2236 with self.assertRaises(ConflictingDefinitionError) as cm:
2237 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map)
2238 self.assertIn("dataset type differs", str(cm.exception))
2240 # And remove the bad definitions.
2241 for datasetTypeName in datasetTypeNames:
2242 self.target_butler.registry.removeDatasetType(datasetTypeName)
2244 # Transfer without creating dataset types should fail.
2245 with self.assertRaises(KeyError):
2246 self.target_butler.transfer_from(self.source_butler, source_refs, id_gen_map=id_gen_map)
2248 # Transfer without creating dimensions should fail.
2249 with self.assertRaises(ConflictingDefinitionError) as cm:
2250 self.target_butler.transfer_from(
2251 self.source_butler, source_refs, id_gen_map=id_gen_map, register_dataset_types=True
2252 )
2253 self.assertIn("dimension", str(cm.exception))
2255 # The failed transfer above leaves registry in an inconsistent
2256 # state because the run is created but then rolled back without
2257 # the collection cache being cleared. For now force a refresh.
2258 # Can remove with DM-35498.
2259 self.target_butler.registry.refresh()
2261 # Now transfer them to the second butler, including dimensions.
2262 with self.assertLogs(level=logging.DEBUG) as cm:
2263 transferred = self.target_butler.transfer_from(
2264 self.source_butler,
2265 source_refs,
2266 id_gen_map=id_gen_map,
2267 register_dataset_types=True,
2268 transfer_dimensions=True,
2269 )
2270 self.assertEqual(len(transferred), n_expected)
2271 log_output = ";".join(cm.output)
2272 self.assertIn("found in datastore for chunk", log_output)
2273 self.assertIn("Creating output run", log_output)
2275 # Do the transfer twice to ensure that it will do nothing extra.
2276 # Only do this if purge=True because it does not work for int
2277 # dataset_id.
2278 if purge:
2279 # This should not need to register dataset types.
2280 transferred = self.target_butler.transfer_from(
2281 self.source_butler, source_refs, id_gen_map=id_gen_map
2282 )
2283 self.assertEqual(len(transferred), n_expected)
2285 # Also do an explicit low-level transfer to trigger some
2286 # edge cases.
2287 with self.assertLogs(level=logging.DEBUG) as cm:
2288 self.target_butler.datastore.transfer_from(self.source_butler.datastore, source_refs)
2289 log_output = ";".join(cm.output)
2290 self.assertIn("no file artifacts exist", log_output)
2292 with self.assertRaises(TypeError):
2293 self.target_butler.datastore.transfer_from(self.source_butler, source_refs)
2295 with self.assertRaises(ValueError):
2296 self.target_butler.datastore.transfer_from(
2297 self.source_butler.datastore, source_refs, transfer="split"
2298 )
2300 # Now try to get the same refs from the new butler.
2301 for ref in source_refs:
2302 if ref not in deleted:
2303 unresolved_ref = ref.unresolved()
2304 new_metric = self.target_butler.get(unresolved_ref, collections=ref.run)
2305 old_metric = self.source_butler.get(unresolved_ref, collections=ref.run)
2306 self.assertEqual(new_metric, old_metric)
2308 # Now prune run2 collection and create instead a CHAINED collection.
2309 # This should block the transfer.
2310 self.target_butler.pruneCollection("run2", purge=True, unstore=True)
2311 self.target_butler.registry.registerCollection("run2", CollectionType.CHAINED)
2312 with self.assertRaises(CollectionTypeError):
2313 # Re-importing the run1 datasets can be problematic if they
2314 # use integer IDs so filter those out.
2315 to_transfer = [ref for ref in source_refs if ref.run == "run2"]
2316 self.target_butler.transfer_from(self.source_butler, to_transfer, id_gen_map=id_gen_map)
2319if __name__ == "__main__": 2319 ↛ 2320line 2319 didn't jump to line 2320, because the condition on line 2319 was never true
2320 unittest.main()