Coverage for python/lsst/pipe/base/executionButlerBuilder.py: 10%

145 statements  

« prev     ^ index     » next       coverage.py v7.2.5, created at 2023-05-04 05:01 -0700

1# This file is part of pipe_base. 

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/>. 

21from __future__ import annotations 

22 

23__all__ = ("buildExecutionButler",) 

24 

25import io 

26import itertools 

27from collections import defaultdict 

28from typing import Callable, DefaultDict, Iterable, List, Mapping, Optional, Set, Tuple, Union 

29 

30from lsst.daf.butler import Butler, Config, DataCoordinate, DatasetRef, DatasetType, Registry 

31from lsst.daf.butler.core.repoRelocation import BUTLER_ROOT_TAG 

32from lsst.daf.butler.registry import ConflictingDefinitionError, MissingDatasetTypeError 

33from lsst.daf.butler.transfers import RepoExportContext 

34from lsst.resources import ResourcePath, ResourcePathExpression 

35from lsst.utils.introspection import get_class_of 

36 

37from .graph import QuantumGraph 

38from .pipeline import PipelineDatasetTypes 

39 

40DataSetTypeMap = Mapping[DatasetType, Set[DataCoordinate]] 

41 

42 

43def _validate_dataset_type( 

44 candidate: DatasetType, previous: dict[Union[str, DatasetType], DatasetType], registry: Registry 

45) -> DatasetType: 

46 """Check the dataset types and return a consistent variant if there are 

47 different compatible options. 

48 

49 Parameters 

50 ---------- 

51 candidate : `lsst.daf.butler.DatasetType` 

52 The candidate dataset type. 

53 previous : `dict` [Union[`str`, `DatasetType`], `DatasetType`] 

54 Previous dataset types found, indexed by name and also by 

55 dataset type. The latter provides a quick way of returning a 

56 previously checked dataset type. 

57 registry : `lsst.daf.butler.Registry` 

58 Main registry whose dataset type registration should override the 

59 given one if it exists. 

60 

61 Returns 

62 ------- 

63 datasetType : `lsst.daf.butler.DatasetType` 

64 The dataset type to be used. This can be different from the 

65 given ``candidate`` if a previous dataset type was encountered 

66 with the same name and this one is compatible with it. 

67 

68 Raises 

69 ------ 

70 ConflictingDefinitionError 

71 Raised if a candidate dataset type has the same name as one 

72 previously encountered but is not compatible with it. 

73 

74 Notes 

75 ----- 

76 This function ensures that if a dataset type is given that has the 

77 same name as a previously encountered dataset type but differs solely 

78 in a way that is interchangeable (through a supported storage class) 

79 then we will always return the first dataset type encountered instead 

80 of the new variant. We assume that the butler will handle the 

81 type conversion itself later. 

82 """ 

83 # First check that if we have previously vetted this dataset type. 

84 # Return the vetted form immediately if we have. 

85 checked = previous.get(candidate) 

86 if checked: 

87 return checked 

88 

89 # Have not previously encountered this dataset type. 

90 name = candidate.name 

91 if prevDsType := previous.get(name): 

92 # Check compatibility. For now assume both directions have to 

93 # be acceptable. 

94 if prevDsType.is_compatible_with(candidate) and candidate.is_compatible_with(prevDsType): 

95 # Ensure that if this dataset type is used again we will return 

96 # the version that we were first given with this name. Store 

97 # it for next time and return the previous one. 

98 previous[candidate] = prevDsType 

99 return prevDsType 

100 else: 

101 raise ConflictingDefinitionError( 

102 f"Dataset type incompatibility in graph: {prevDsType} not compatible with {candidate}" 

103 ) 

104 

105 # We haven't seen this dataset type in this graph before, but it may 

106 # already be in the registry. 

107 try: 

108 registryDsType = registry.getDatasetType(name) 

109 previous[candidate] = registryDsType 

110 return registryDsType 

111 except MissingDatasetTypeError: 

112 pass 

113 # Dataset type is totally new. Store it by name and by dataset type so 

114 # it will be validated immediately next time it comes up. 

115 previous[name] = candidate 

116 previous[candidate] = candidate 

117 return candidate 

118 

119 

120def _accumulate( 

121 butler: Butler, 

122 graph: QuantumGraph, 

123 dataset_types: PipelineDatasetTypes, 

124) -> Tuple[Set[DatasetRef], DataSetTypeMap]: 

125 # accumulate the DatasetRefs that will be transferred to the execution 

126 # registry 

127 

128 # exports holds all the existing data that will be migrated to the 

129 # execution butler 

130 exports: Set[DatasetRef] = set() 

131 

132 # inserts is the mapping of DatasetType to dataIds for what is to be 

133 # inserted into the registry. These are the products that are expected 

134 # to be produced during processing of the QuantumGraph 

135 inserts: DefaultDict[DatasetType, Set[DataCoordinate]] = defaultdict(set) 

136 

137 # It is possible to end up with a graph that has different storage 

138 # classes attached to the same dataset type name. This is okay but 

139 # must we must ensure that only a single dataset type definition is 

140 # accumulated in the loop below. This data structure caches every dataset 

141 # type encountered and stores the compatible alternative. 

142 datasetTypes: dict[Union[str, DatasetType], DatasetType] = {} 

143 

144 # Add inserts for initOutputs (including initIntermediates); these are 

145 # defined fully by their DatasetType, because they have no dimensions. 

146 # initInputs are part of Quantum and that's the only place the graph stores 

147 # the dataset IDs, so we process them there even though each Quantum for a 

148 # task has the same ones. 

149 for dataset_type in itertools.chain(dataset_types.initIntermediates, dataset_types.initOutputs): 

150 if dataset_type.component() is not None: 

151 dataset_type = dataset_type.makeCompositeDatasetType() 

152 dataset_type = _validate_dataset_type(dataset_type, datasetTypes, butler.registry) 

153 inserts[dataset_type].add(DataCoordinate.makeEmpty(dataset_type.dimensions.universe)) 

154 

155 # Output references may be resolved even if they do not exist. Find all 

156 # actually existing refs. 

157 check_refs: Set[DatasetRef] = set() 

158 for quantum in (n.quantum for n in graph): 

159 for attrName in ("initInputs", "inputs", "outputs"): 

160 attr: Mapping[DatasetType, Union[DatasetRef, List[DatasetRef]]] = getattr(quantum, attrName) 

161 for type, refs in attr.items(): 

162 # This if block is because init inputs has a different 

163 # signature for its items 

164 if not isinstance(refs, list): 

165 refs = [refs] 

166 for ref in refs: 

167 if ref.id is not None: 

168 # We could check existence of individual components, 

169 # but it should be less work to check their parent. 

170 if ref.isComponent(): 

171 ref = ref.makeCompositeRef() 

172 check_refs.add(ref) 

173 exist_map = butler.datastore.knows_these(check_refs) 

174 existing_ids = set(ref.id for ref, exists in exist_map.items() if exists) 

175 del exist_map 

176 

177 for quantum in (n.quantum for n in graph): 

178 for attrName in ("initInputs", "inputs", "outputs"): 

179 attr = getattr(quantum, attrName) 

180 

181 for type, refs in attr.items(): 

182 if not isinstance(refs, list): 

183 refs = [refs] 

184 if type.component() is not None: 

185 type = type.makeCompositeDatasetType() 

186 type = _validate_dataset_type(type, datasetTypes, butler.registry) 

187 # iterate over all the references, if it exists and should be 

188 # exported, if not it should be inserted into the new registry 

189 for ref in refs: 

190 # Component dataset ID is the same as its parent ID, so 

191 # checking component in existing_ids works OK. 

192 if ref.id is not None and ref.id in existing_ids: 

193 # If this is a component we want the composite to be 

194 # exported. 

195 if ref.isComponent(): 

196 ref = ref.makeCompositeRef() 

197 # Make sure we export this with the registry's dataset 

198 # type, since transfer_from doesn't handle storage 

199 # class differences (maybe it should, but it's not 

200 # bad to be defensive here even if that changes). 

201 if type != ref.datasetType: 

202 ref = ref.overrideStorageClass(type.storageClass) 

203 assert ref.datasetType == type, "Dataset types should not differ in other ways." 

204 exports.add(ref) 

205 else: 

206 if ref.isComponent(): 

207 # We can't insert a component, and a component will 

208 # be part of some other upstream dataset, so it 

209 # should be safe to skip them here 

210 continue 

211 inserts[type].add(ref.dataId) 

212 return exports, inserts 

213 

214 

215def _discoverCollections(butler: Butler, collections: Iterable[str]) -> set[str]: 

216 # Recurse through any discovered collections to make sure all collections 

217 # are exported. This exists because I ran into a situation where some 

218 # collections were not properly being discovered and exported. This 

219 # method may be able to be removed in the future if collection export 

220 # logic changes 

221 collections = set(collections) 

222 while True: 

223 discoveredCollections = set( 

224 butler.registry.queryCollections(collections, flattenChains=True, includeChains=True) 

225 ) 

226 if len(discoveredCollections) > len(collections): 

227 collections = discoveredCollections 

228 else: 

229 break 

230 return collections 

231 

232 

233def _export(butler: Butler, collections: Optional[Iterable[str]], inserts: DataSetTypeMap) -> io.StringIO: 

234 # This exports relevant dimension records and collections using daf butler 

235 # objects, however it reaches in deep and does not use the public methods 

236 # so that it can export it to a string buffer and skip disk access. This 

237 # does not export the datasets themselves, since we use transfer_from for 

238 # that. 

239 yamlBuffer = io.StringIO() 

240 # Yaml is hard coded, since the class controls both ends of the 

241 # export/import 

242 BackendClass = get_class_of(butler._config["repo_transfer_formats", "yaml", "export"]) 

243 backend = BackendClass(yamlBuffer, universe=butler.registry.dimensions) 

244 exporter = RepoExportContext(butler.registry, butler.datastore, backend, directory=None, transfer=None) 

245 

246 # Need to ensure that the dimension records for outputs are 

247 # transferred. 

248 for _, dataIds in inserts.items(): 

249 exporter.saveDataIds(dataIds) 

250 

251 # Look for any defined collection, if not get the defaults 

252 if collections is None: 

253 collections = butler.registry.defaults.collections 

254 

255 # look up all collections associated with those inputs, this follows 

256 # all chains to make sure everything is properly exported 

257 for c in _discoverCollections(butler, collections): 

258 exporter.saveCollection(c) 

259 exporter._finish() 

260 

261 # reset the string buffer to the beginning so the read operation will 

262 # actually *see* the data that was exported 

263 yamlBuffer.seek(0) 

264 return yamlBuffer 

265 

266 

267def _setupNewButler( 

268 butler: Butler, 

269 outputLocation: ResourcePath, 

270 dirExists: bool, 

271 datastoreRoot: Optional[ResourcePath] = None, 

272) -> Butler: 

273 """Set up the execution butler 

274 

275 Parameters 

276 ---------- 

277 butler : `Butler` 

278 The original butler, upon which the execution butler is based. 

279 outputLocation : `ResourcePath` 

280 Location of the execution butler. 

281 dirExists : `bool` 

282 Does the ``outputLocation`` exist, and if so, should it be clobbered? 

283 datastoreRoot : `ResourcePath`, optional 

284 Path for the execution butler datastore. If not specified, then the 

285 original butler's datastore will be used. 

286 

287 Returns 

288 ------- 

289 execution_butler : `Butler` 

290 Execution butler. 

291 """ 

292 # Set up the new butler object at the specified location 

293 if dirExists: 

294 # Remove the existing table, if the code got this far and this exists 

295 # clobber must be true 

296 executionRegistry = outputLocation.join("gen3.sqlite3") 

297 if executionRegistry.exists(): 

298 executionRegistry.remove() 

299 else: 

300 outputLocation.mkdir() 

301 

302 # Copy the existing butler config, modifying the location of the 

303 # registry to the specified location. 

304 # Preserve the root path from the existing butler so things like 

305 # file data stores continue to look at the old location. 

306 config = Config(butler._config) 

307 config["root"] = outputLocation.geturl() 

308 config["allow_put_of_predefined_dataset"] = True 

309 config["registry", "db"] = "sqlite:///<butlerRoot>/gen3.sqlite3" 

310 

311 # Remove any namespace that may be set in main registry. 

312 config.pop(("registry", "namespace"), None) 

313 

314 # Obscore manager cannot be used with execution butler. 

315 config.pop(("registry", "managers", "obscore"), None) 

316 

317 # record the current root of the datastore if it is specified relative 

318 # to the butler root 

319 if datastoreRoot is not None: 

320 config["datastore", "root"] = datastoreRoot.geturl() 

321 elif config.get(("datastore", "root")) == BUTLER_ROOT_TAG and butler._config.configDir is not None: 

322 config["datastore", "root"] = butler._config.configDir.geturl() 

323 config["datastore", "trust_get_request"] = True 

324 

325 # Requires that we use the dimension configuration from the original 

326 # butler and not use the defaults. 

327 config = Butler.makeRepo( 

328 root=outputLocation, 

329 config=config, 

330 dimensionConfig=butler.registry.dimensions.dimensionConfig, 

331 overwrite=True, 

332 forceConfigRoot=False, 

333 ) 

334 

335 # Return a newly created butler 

336 return Butler(config, writeable=True) 

337 

338 

339def _import( 

340 yamlBuffer: io.StringIO, 

341 newButler: Butler, 

342 inserts: DataSetTypeMap, 

343 run: Optional[str], 

344 butlerModifier: Optional[Callable[[Butler], Butler]], 

345) -> Butler: 

346 # This method takes the exports from the existing butler, imports 

347 # them into the newly created butler, and then inserts the datasets 

348 # that are expected to be produced. 

349 

350 # import the existing datasets using "split" mode. "split" is safe 

351 # because execution butler is assumed to be able to see all the file 

352 # locations that the main datastore can see. "split" supports some 

353 # absolute URIs in the datastore. 

354 newButler.import_(filename=yamlBuffer, format="yaml", reuseIds=True, transfer="split") 

355 

356 # If there is modifier callable, run it to make necessary updates 

357 # to the new butler. 

358 if butlerModifier is not None: 

359 newButler = butlerModifier(newButler) 

360 

361 # Register datasets to be produced and insert them into the registry 

362 for dsType, dataIds in inserts.items(): 

363 # Storage class differences should have already been resolved by calls 

364 # _validate_dataset_type in _export, resulting in the Registry dataset 

365 # type whenever that exists. 

366 newButler.registry.registerDatasetType(dsType) 

367 newButler.registry.insertDatasets(dsType, dataIds, run) 

368 

369 return newButler 

370 

371 

372def buildExecutionButler( 

373 butler: Butler, 

374 graph: QuantumGraph, 

375 outputLocation: ResourcePathExpression, 

376 run: Optional[str], 

377 *, 

378 clobber: bool = False, 

379 butlerModifier: Optional[Callable[[Butler], Butler]] = None, 

380 collections: Optional[Iterable[str]] = None, 

381 datastoreRoot: Optional[ResourcePathExpression] = None, 

382 transfer: str = "auto", 

383) -> Butler: 

384 r"""buildExecutionButler is a function that is responsible for exporting 

385 input `QuantumGraphs` into a new minimal `~lsst.daf.butler.Butler` which 

386 only contains datasets specified by the `QuantumGraph`. These datasets are 

387 both those that already exist in the input `~lsst.daf.butler.Butler`, and 

388 those that are expected to be produced during the execution of the 

389 `QuantumGraph`. 

390 

391 Parameters 

392 ---------- 

393 butler : `lsst.daf.butler.Bulter` 

394 This is the existing `~lsst.daf.butler.Butler` instance from which 

395 existing datasets will be exported. This should be the 

396 `~lsst.daf.butler.Butler` which was used to create any `QuantumGraphs` 

397 that will be converted with this object. 

398 graph : `QuantumGraph` 

399 Graph containing nodes that are to be exported into an execution 

400 butler 

401 outputLocation : convertible to `ResourcePath` 

402 URI Location at which the execution butler is to be exported. May be 

403 specified as a string or a `ResourcePath` instance. 

404 run : `str`, optional 

405 The run collection that the exported datasets are to be placed in. If 

406 None, the default value in registry.defaults will be used. 

407 clobber : `bool`, Optional 

408 By default a butler will not be created if a file or directory 

409 already exists at the output location. If this is set to `True` 

410 what is at the location will be deleted prior to running the 

411 export. Defaults to `False` 

412 butlerModifier : `~typing.Callable`, Optional 

413 If supplied this should be a callable that accepts a 

414 `~lsst.daf.butler.Butler`, and returns an instantiated 

415 `~lsst.daf.butler.Butler`. This callable may be used to make any 

416 modifications to the `~lsst.daf.butler.Butler` desired. This 

417 will be called after importing all datasets that exist in the input 

418 `~lsst.daf.butler.Butler` but prior to inserting Datasets expected 

419 to be produced. Examples of what this method could do include 

420 things such as creating collections/runs/ etc. 

421 collections : `~typing.Iterable` of `str`, Optional 

422 An iterable of collection names that will be exported from the input 

423 `~lsst.daf.butler.Butler` when creating the execution butler. If not 

424 supplied the `~lsst.daf.butler.Butler`\ 's `~lsst.daf.butler.Registry` 

425 default collections will be used. 

426 datastoreRoot : convertible to `ResourcePath`, Optional 

427 Root directory for datastore of execution butler. If `None`, then the 

428 original butler's datastore will be used. 

429 transfer : `str` 

430 How (and whether) the input datasets should be added to the execution 

431 butler datastore. This should be a ``transfer`` string recognized by 

432 :func:`lsst.resources.ResourcePath.transfer_from`. 

433 ``"auto"`` means to ``"copy"`` if the ``datastoreRoot`` is specified. 

434 

435 Returns 

436 ------- 

437 executionButler : `lsst.daf.butler.Butler` 

438 An instance of the newly created execution butler 

439 

440 Raises 

441 ------ 

442 FileExistsError 

443 Raised if something exists in the filesystem at the specified output 

444 location and clobber is `False` 

445 NotADirectoryError 

446 Raised if specified output URI does not correspond to a directory 

447 """ 

448 # We know this must refer to a directory. 

449 outputLocation = ResourcePath(outputLocation, forceDirectory=True) 

450 if datastoreRoot is not None: 

451 datastoreRoot = ResourcePath(datastoreRoot, forceDirectory=True) 

452 

453 # Do this first to Fail Fast if the output exists 

454 if (dirExists := outputLocation.exists()) and not clobber: 

455 raise FileExistsError("Cannot create a butler at specified location, location exists") 

456 if not outputLocation.isdir(): 

457 raise NotADirectoryError("The specified output URI does not appear to correspond to a directory") 

458 

459 # Gather all DatasetTypes from the Python and check any that already exist 

460 # in the registry for consistency. This does not check that all dataset 

461 # types here exist, because they might want to register dataset types 

462 # later. It would be nice to also check that, but to that we would need to 

463 # be told whether they plan to register dataset types later (DM-30845). 

464 dataset_types = PipelineDatasetTypes.fromPipeline(graph.iterTaskGraph(), registry=butler.registry) 

465 

466 exports, inserts = _accumulate(butler, graph, dataset_types) 

467 yamlBuffer = _export(butler, collections, inserts) 

468 

469 newButler = _setupNewButler(butler, outputLocation, dirExists, datastoreRoot) 

470 

471 newButler = _import(yamlBuffer, newButler, inserts, run, butlerModifier) 

472 

473 if transfer == "auto" and datastoreRoot is not None: 

474 transfer = "copy" 

475 

476 # Transfer the existing datasets directly from the source butler. 

477 newButler.transfer_from( 

478 butler, 

479 exports, 

480 transfer=transfer, 

481 skip_missing=False, # Everything should exist. 

482 register_dataset_types=True, 

483 transfer_dimensions=True, 

484 ) 

485 

486 return newButler