Coverage for python/lsst/daf/butler/transfers/_yaml.py: 13%

193 statements  

« prev     ^ index     » next       coverage.py v7.2.7, created at 2023-08-05 01:26 +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/>. 

21 

22from __future__ import annotations 

23 

24__all__ = ["YamlRepoExportBackend", "YamlRepoImportBackend"] 

25 

26import uuid 

27import warnings 

28from collections import UserDict, defaultdict 

29from collections.abc import Iterable, Mapping 

30from datetime import datetime 

31from typing import IO, TYPE_CHECKING, Any 

32 

33import astropy.time 

34import yaml 

35from lsst.resources import ResourcePath 

36from lsst.utils import doImportType 

37from lsst.utils.introspection import find_outside_stacklevel 

38from lsst.utils.iteration import ensure_iterable 

39 

40from ..core import ( 

41 DatasetAssociation, 

42 DatasetId, 

43 DatasetRef, 

44 DatasetType, 

45 Datastore, 

46 DimensionElement, 

47 DimensionRecord, 

48 DimensionUniverse, 

49 FileDataset, 

50 Timespan, 

51) 

52from ..core.named import NamedValueSet 

53from ..registry import CollectionType, Registry 

54from ..registry.interfaces import ChainedCollectionRecord, CollectionRecord, RunRecord, VersionTuple 

55from ..registry.versions import IncompatibleVersionError 

56from ._interfaces import RepoExportBackend, RepoImportBackend 

57 

58if TYPE_CHECKING: 

59 from lsst.resources import ResourcePathExpression 

60 

61EXPORT_FORMAT_VERSION = VersionTuple(1, 0, 2) 

62"""Export format version. 

63 

64Files with a different major version or a newer minor version cannot be read by 

65this version of the code. 

66""" 

67 

68 

69class _RefMapper(UserDict[int, uuid.UUID]): 

70 """Create a local dict subclass which creates new deterministic UUID for 

71 missing keys. 

72 """ 

73 

74 _namespace = uuid.UUID("4d4851f4-2890-4d41-8779-5f38a3f5062b") 

75 

76 def __missing__(self, key: int) -> uuid.UUID: 

77 newUUID = uuid.uuid3(namespace=self._namespace, name=str(key)) 

78 self[key] = newUUID 

79 return newUUID 

80 

81 

82_refIntId2UUID = _RefMapper() 

83 

84 

85def _uuid_representer(dumper: yaml.Dumper, data: uuid.UUID) -> yaml.Node: 

86 """Generate YAML representation for UUID. 

87 

88 This produces a scalar node with a tag "!uuid" and value being a regular 

89 string representation of UUID. 

90 """ 

91 return dumper.represent_scalar("!uuid", str(data)) 

92 

93 

94def _uuid_constructor(loader: yaml.Loader, node: yaml.Node) -> uuid.UUID | None: 

95 if node.value is not None: 

96 return uuid.UUID(hex=node.value) 

97 return None 

98 

99 

100yaml.Dumper.add_representer(uuid.UUID, _uuid_representer) 

101yaml.SafeLoader.add_constructor("!uuid", _uuid_constructor) 

102 

103 

104class YamlRepoExportBackend(RepoExportBackend): 

105 """A repository export implementation that saves to a YAML file. 

106 

107 Parameters 

108 ---------- 

109 stream 

110 A writeable file-like object. 

111 """ 

112 

113 def __init__(self, stream: IO, universe: DimensionUniverse): 

114 self.stream = stream 

115 self.universe = universe 

116 self.data: list[dict[str, Any]] = [] 

117 

118 def saveDimensionData(self, element: DimensionElement, *data: DimensionRecord) -> None: 

119 # Docstring inherited from RepoExportBackend.saveDimensionData. 

120 data_dicts = [record.toDict(splitTimespan=True) for record in data] 

121 self.data.append( 

122 { 

123 "type": "dimension", 

124 "element": element.name, 

125 "records": data_dicts, 

126 } 

127 ) 

128 

129 def saveCollection(self, record: CollectionRecord, doc: str | None) -> None: 

130 # Docstring inherited from RepoExportBackend.saveCollections. 

131 data: dict[str, Any] = { 

132 "type": "collection", 

133 "collection_type": record.type.name, 

134 "name": record.name, 

135 } 

136 if doc is not None: 

137 data["doc"] = doc 

138 if isinstance(record, RunRecord): 

139 data["host"] = record.host 

140 data["timespan_begin"] = record.timespan.begin 

141 data["timespan_end"] = record.timespan.end 

142 elif isinstance(record, ChainedCollectionRecord): 

143 data["children"] = list(record.children) 

144 self.data.append(data) 

145 

146 def saveDatasets(self, datasetType: DatasetType, run: str, *datasets: FileDataset) -> None: 

147 # Docstring inherited from RepoExportBackend.saveDatasets. 

148 self.data.append( 

149 { 

150 "type": "dataset_type", 

151 "name": datasetType.name, 

152 "dimensions": [d.name for d in datasetType.dimensions], 

153 "storage_class": datasetType.storageClass_name, 

154 "is_calibration": datasetType.isCalibration(), 

155 } 

156 ) 

157 self.data.append( 

158 { 

159 "type": "dataset", 

160 "dataset_type": datasetType.name, 

161 "run": run, 

162 "records": [ 

163 { 

164 "dataset_id": [ref.id for ref in sorted(dataset.refs)], 

165 "data_id": [ref.dataId.byName() for ref in sorted(dataset.refs)], 

166 "path": dataset.path, 

167 "formatter": dataset.formatter, 

168 # TODO: look up and save other collections 

169 } 

170 for dataset in datasets 

171 ], 

172 } 

173 ) 

174 

175 def saveDatasetAssociations( 

176 self, collection: str, collectionType: CollectionType, associations: Iterable[DatasetAssociation] 

177 ) -> None: 

178 # Docstring inherited from RepoExportBackend.saveDatasetAssociations. 

179 if collectionType is CollectionType.TAGGED: 

180 self.data.append( 

181 { 

182 "type": "associations", 

183 "collection": collection, 

184 "collection_type": collectionType.name, 

185 "dataset_ids": [assoc.ref.id for assoc in associations], 

186 } 

187 ) 

188 elif collectionType is CollectionType.CALIBRATION: 

189 idsByTimespan: dict[Timespan, list[DatasetId]] = defaultdict(list) 

190 for association in associations: 

191 assert association.timespan is not None 

192 idsByTimespan[association.timespan].append(association.ref.id) 

193 self.data.append( 

194 { 

195 "type": "associations", 

196 "collection": collection, 

197 "collection_type": collectionType.name, 

198 "validity_ranges": [ 

199 { 

200 "timespan": timespan, 

201 "dataset_ids": dataset_ids, 

202 } 

203 for timespan, dataset_ids in idsByTimespan.items() 

204 ], 

205 } 

206 ) 

207 

208 def finish(self) -> None: 

209 # Docstring inherited from RepoExportBackend. 

210 yaml.dump( 

211 { 

212 "description": "Butler Data Repository Export", 

213 "version": str(EXPORT_FORMAT_VERSION), 

214 "universe_version": self.universe.version, 

215 "universe_namespace": self.universe.namespace, 

216 "data": self.data, 

217 }, 

218 stream=self.stream, 

219 sort_keys=False, 

220 ) 

221 

222 

223class YamlRepoImportBackend(RepoImportBackend): 

224 """A repository import implementation that reads from a YAML file. 

225 

226 Parameters 

227 ---------- 

228 stream 

229 A readable file-like object. 

230 registry : `Registry` 

231 The registry datasets will be imported into. Only used to retreive 

232 dataset types during construction; all write happen in `register` 

233 and `load`. 

234 """ 

235 

236 def __init__(self, stream: IO, registry: Registry): 

237 # We read the file fully and convert its contents to Python objects 

238 # instead of loading incrementally so we can spot some problems early; 

239 # because `register` can't be put inside a transaction, we'd rather not 

240 # run that at all if there's going to be problem later in `load`. 

241 wrapper = yaml.safe_load(stream) 

242 if wrapper["version"] == 0: 

243 # Grandfather-in 'version: 0' -> 1.0.0, which is what we wrote 

244 # before we really tried to do versioning here. 

245 fileVersion = VersionTuple(1, 0, 0) 

246 else: 

247 fileVersion = VersionTuple.fromString(wrapper["version"]) 

248 if fileVersion.major != EXPORT_FORMAT_VERSION.major: 

249 raise IncompatibleVersionError( 

250 f"Cannot read repository export file with version={fileVersion} " 

251 f"({EXPORT_FORMAT_VERSION.major}.x.x required)." 

252 ) 

253 if fileVersion.minor > EXPORT_FORMAT_VERSION.minor: 

254 raise IncompatibleVersionError( 

255 f"Cannot read repository export file with version={fileVersion} " 

256 f"< {EXPORT_FORMAT_VERSION.major}.{EXPORT_FORMAT_VERSION.minor}.x required." 

257 ) 

258 self.runs: dict[str, tuple[str | None, Timespan]] = {} 

259 self.chains: dict[str, list[str]] = {} 

260 self.collections: dict[str, CollectionType] = {} 

261 self.collectionDocs: dict[str, str] = {} 

262 self.datasetTypes: NamedValueSet[DatasetType] = NamedValueSet() 

263 self.dimensions: Mapping[DimensionElement, list[DimensionRecord]] = defaultdict(list) 

264 self.tagAssociations: dict[str, list[DatasetId]] = defaultdict(list) 

265 self.calibAssociations: dict[str, dict[Timespan, list[DatasetId]]] = defaultdict(dict) 

266 self.refsByFileId: dict[DatasetId, DatasetRef] = {} 

267 self.registry: Registry = registry 

268 

269 universe_version = wrapper.get("universe_version", 0) 

270 universe_namespace = wrapper.get("universe_namespace", "daf_butler") 

271 

272 # If this is data exported before the reorganization of visits 

273 # and visit systems and that new schema is in use, some filtering 

274 # will be needed. The entry in the visit dimension record will be 

275 # silently dropped when visit is created but the 

276 # visit_system_membership must be constructed. 

277 migrate_visit_system = False 

278 if ( 

279 universe_version < 2 

280 and universe_namespace == "daf_butler" 

281 and "visit_system_membership" in self.registry.dimensions 

282 ): 

283 migrate_visit_system = True 

284 

285 datasetData = [] 

286 for data in wrapper["data"]: 

287 if data["type"] == "dimension": 

288 # convert all datetime values to astropy 

289 for record in data["records"]: 

290 for key in record: 

291 # Some older YAML files were produced with native 

292 # YAML support for datetime, we support reading that 

293 # data back. Newer conversion uses _AstropyTimeToYAML 

294 # class with special YAML tag. 

295 if isinstance(record[key], datetime): 

296 record[key] = astropy.time.Time(record[key], scale="utc") 

297 element = self.registry.dimensions[data["element"]] 

298 RecordClass: type[DimensionRecord] = element.RecordClass 

299 self.dimensions[element].extend(RecordClass(**r) for r in data["records"]) 

300 

301 if data["element"] == "visit" and migrate_visit_system: 

302 # Must create the visit_system_membership records. 

303 element = self.registry.dimensions["visit_system_membership"] 

304 RecordClass = element.RecordClass 

305 self.dimensions[element].extend( 

306 RecordClass(instrument=r["instrument"], visit_system=r["visit_system"], visit=r["id"]) 

307 for r in data["records"] 

308 ) 

309 

310 elif data["type"] == "collection": 

311 collectionType = CollectionType.from_name(data["collection_type"]) 

312 if collectionType is CollectionType.RUN: 

313 self.runs[data["name"]] = ( 

314 data["host"], 

315 Timespan(begin=data["timespan_begin"], end=data["timespan_end"]), 

316 ) 

317 elif collectionType is CollectionType.CHAINED: 

318 children = [] 

319 for child in data["children"]: 

320 if not isinstance(child, str): 

321 warnings.warn( 

322 f"CHAINED collection {data['name']} includes restrictions on child " 

323 "collection searches, which are no longer suppored and will be ignored.", 

324 stacklevel=find_outside_stacklevel("lsst.daf.butler"), 

325 ) 

326 # Old form with dataset type restrictions only, 

327 # supported for backwards compatibility. 

328 child, _ = child 

329 children.append(child) 

330 self.chains[data["name"]] = children 

331 else: 

332 self.collections[data["name"]] = collectionType 

333 doc = data.get("doc") 

334 if doc is not None: 

335 self.collectionDocs[data["name"]] = doc 

336 elif data["type"] == "run": 

337 # Also support old form of saving a run with no extra info. 

338 self.runs[data["name"]] = (None, Timespan(None, None)) 

339 elif data["type"] == "dataset_type": 

340 dimensions = data["dimensions"] 

341 if migrate_visit_system and "visit" in dimensions and "visit_system" in dimensions: 

342 dimensions.remove("visit_system") 

343 self.datasetTypes.add( 

344 DatasetType( 

345 data["name"], 

346 dimensions=dimensions, 

347 storageClass=data["storage_class"], 

348 universe=self.registry.dimensions, 

349 isCalibration=data.get("is_calibration", False), 

350 ) 

351 ) 

352 elif data["type"] == "dataset": 

353 # Save raw dataset data for a second loop, so we can ensure we 

354 # know about all dataset types first. 

355 datasetData.append(data) 

356 elif data["type"] == "associations": 

357 collectionType = CollectionType.from_name(data["collection_type"]) 

358 if collectionType is CollectionType.TAGGED: 

359 self.tagAssociations[data["collection"]].extend( 

360 [x if not isinstance(x, int) else _refIntId2UUID[x] for x in data["dataset_ids"]] 

361 ) 

362 elif collectionType is CollectionType.CALIBRATION: 

363 assocsByTimespan = self.calibAssociations[data["collection"]] 

364 for d in data["validity_ranges"]: 

365 if "timespan" in d: 

366 assocsByTimespan[d["timespan"]] = [ 

367 x if not isinstance(x, int) else _refIntId2UUID[x] for x in d["dataset_ids"] 

368 ] 

369 else: 

370 # TODO: this is for backward compatibility, should 

371 # be removed at some point. 

372 assocsByTimespan[Timespan(begin=d["begin"], end=d["end"])] = [ 

373 x if not isinstance(x, int) else _refIntId2UUID[x] for x in d["dataset_ids"] 

374 ] 

375 else: 

376 raise ValueError(f"Unexpected calibration type for association: {collectionType.name}.") 

377 else: 

378 raise ValueError(f"Unexpected dictionary type: {data['type']}.") 

379 # key is (dataset type name, run) 

380 self.datasets: Mapping[tuple[str, str], list[FileDataset]] = defaultdict(list) 

381 for data in datasetData: 

382 datasetType = self.datasetTypes.get(data["dataset_type"]) 

383 if datasetType is None: 

384 datasetType = self.registry.getDatasetType(data["dataset_type"]) 

385 self.datasets[data["dataset_type"], data["run"]].extend( 

386 FileDataset( 

387 d.get("path"), 

388 [ 

389 DatasetRef( 

390 datasetType, 

391 dataId, 

392 run=data["run"], 

393 id=refid if not isinstance(refid, int) else _refIntId2UUID[refid], 

394 ) 

395 for dataId, refid in zip( 

396 ensure_iterable(d["data_id"]), ensure_iterable(d["dataset_id"]), strict=True 

397 ) 

398 ], 

399 formatter=doImportType(d.get("formatter")) if "formatter" in d else None, 

400 ) 

401 for d in data["records"] 

402 ) 

403 

404 def register(self) -> None: 

405 # Docstring inherited from RepoImportBackend.register. 

406 for datasetType in self.datasetTypes: 

407 self.registry.registerDatasetType(datasetType) 

408 for run in self.runs: 

409 self.registry.registerRun(run, doc=self.collectionDocs.get(run)) 

410 # No way to add extra run info to registry yet. 

411 for collection, collection_type in self.collections.items(): 

412 self.registry.registerCollection( 

413 collection, collection_type, doc=self.collectionDocs.get(collection) 

414 ) 

415 for chain, children in self.chains.items(): 

416 self.registry.registerCollection( 

417 chain, CollectionType.CHAINED, doc=self.collectionDocs.get(chain) 

418 ) 

419 self.registry.setCollectionChain(chain, children) 

420 

421 def load( 

422 self, 

423 datastore: Datastore | None, 

424 *, 

425 directory: ResourcePathExpression | None = None, 

426 transfer: str | None = None, 

427 skip_dimensions: set | None = None, 

428 ) -> None: 

429 # Docstring inherited from RepoImportBackend.load. 

430 for element, dimensionRecords in self.dimensions.items(): 

431 if skip_dimensions and element in skip_dimensions: 

432 continue 

433 # Using skip_existing=True here assumes that the records in the 

434 # database are either equivalent or at least preferable to the ones 

435 # being imported. It'd be ideal to check that, but that would mean 

436 # using syncDimensionData, which is not vectorized and is hence 

437 # unacceptably slo. 

438 self.registry.insertDimensionData(element, *dimensionRecords, skip_existing=True) 

439 # FileDatasets to ingest into the datastore (in bulk): 

440 fileDatasets = [] 

441 for records in self.datasets.values(): 

442 # Make a big flattened list of all data IDs and dataset_ids, while 

443 # remembering slices that associate them with the FileDataset 

444 # instances they came from. 

445 datasets: list[DatasetRef] = [] 

446 dataset_ids: list[DatasetId] = [] 

447 slices = [] 

448 for fileDataset in records: 

449 start = len(datasets) 

450 datasets.extend(fileDataset.refs) 

451 dataset_ids.extend(ref.id for ref in fileDataset.refs) 

452 stop = len(datasets) 

453 slices.append(slice(start, stop)) 

454 # Insert all of those DatasetRefs at once. 

455 # For now, we ignore the dataset_id we pulled from the file 

456 # and just insert without one to get a new autoincrement value. 

457 # Eventually (once we have origin in IDs) we'll preserve them. 

458 resolvedRefs = self.registry._importDatasets(datasets) 

459 # Populate our dictionary that maps int dataset_id values from the 

460 # export file to the new DatasetRefs 

461 for fileId, ref in zip(dataset_ids, resolvedRefs, strict=True): 

462 self.refsByFileId[fileId] = ref 

463 # Now iterate over the original records, and install the new 

464 # resolved DatasetRefs to replace the unresolved ones as we 

465 # reorganize the collection information. 

466 for sliceForFileDataset, fileDataset in zip(slices, records, strict=True): 

467 fileDataset.refs = resolvedRefs[sliceForFileDataset] 

468 if directory is not None: 

469 fileDataset.path = ResourcePath(directory, forceDirectory=True).join(fileDataset.path) 

470 fileDatasets.append(fileDataset) 

471 # Ingest everything into the datastore at once. 

472 if datastore is not None and fileDatasets: 

473 datastore.ingest(*fileDatasets, transfer=transfer) 

474 # Associate datasets with tagged collections. 

475 for collection, dataset_ids in self.tagAssociations.items(): 

476 self.registry.associate(collection, [self.refsByFileId[i] for i in dataset_ids]) 

477 # Associate datasets with calibration collections. 

478 for collection, idsByTimespan in self.calibAssociations.items(): 

479 for timespan, dataset_ids in idsByTimespan.items(): 

480 self.registry.certify(collection, [self.refsByFileId[i] for i in dataset_ids], timespan)