Coverage for python/lsst/daf/butler/registry/bridge/monolithic.py: 28%

103 statements  

« prev     ^ index     » next       coverage.py v7.3.2, created at 2023-12-06 10:53 +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 software is dual licensed under the GNU General Public License and also 

10# under a 3-clause BSD license. Recipients may choose which of these licenses 

11# to use; please see the files gpl-3.0.txt and/or bsd_license.txt, 

12# respectively. If you choose the GPL option then the following text applies 

13# (but note that there is still no warranty even if you opt for BSD instead): 

14# 

15# This program is free software: you can redistribute it and/or modify 

16# it under the terms of the GNU General Public License as published by 

17# the Free Software Foundation, either version 3 of the License, or 

18# (at your option) any later version. 

19# 

20# This program is distributed in the hope that it will be useful, 

21# but WITHOUT ANY WARRANTY; without even the implied warranty of 

22# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

23# GNU General Public License for more details. 

24# 

25# You should have received a copy of the GNU General Public License 

26# along with this program. If not, see <http://www.gnu.org/licenses/>. 

27from __future__ import annotations 

28 

29from ... import ddl 

30 

31__all__ = ("MonolithicDatastoreRegistryBridgeManager", "MonolithicDatastoreRegistryBridge") 

32 

33import copy 

34from collections import namedtuple 

35from collections.abc import Iterable, Iterator 

36from contextlib import contextmanager 

37from typing import TYPE_CHECKING, cast 

38 

39import sqlalchemy 

40 

41from ..._named import NamedValueSet 

42from ...datastore.stored_file_info import StoredDatastoreItemInfo 

43from ..interfaces import ( 

44 DatasetIdRef, 

45 DatastoreRegistryBridge, 

46 DatastoreRegistryBridgeManager, 

47 FakeDatasetRef, 

48 OpaqueTableStorage, 

49 VersionTuple, 

50) 

51from ..opaque import ByNameOpaqueTableStorage 

52from .ephemeral import EphemeralDatastoreRegistryBridge 

53 

54if TYPE_CHECKING: 

55 from ...datastore import DatastoreTransaction 

56 from ...dimensions import DimensionUniverse 

57 from ..interfaces import ( 

58 Database, 

59 DatasetRecordStorageManager, 

60 OpaqueTableStorageManager, 

61 StaticTablesContext, 

62 ) 

63 

64_TablesTuple = namedtuple( 

65 "_TablesTuple", 

66 [ 

67 "dataset_location", 

68 "dataset_location_trash", 

69 ], 

70) 

71 

72# This has to be updated on every schema change 

73_VERSION = VersionTuple(0, 2, 0) 

74 

75 

76def _makeTableSpecs(datasets: type[DatasetRecordStorageManager]) -> _TablesTuple: 

77 """Construct specifications for tables used by the monolithic datastore 

78 bridge classes. 

79 

80 Parameters 

81 ---------- 

82 universe : `DimensionUniverse` 

83 All dimensions known to the `Registry`. 

84 datasets : subclass of `DatasetRecordStorageManager` 

85 Manager class for datasets; used only to create foreign key fields. 

86 

87 Returns 

88 ------- 

89 specs : `_TablesTuple` 

90 A named tuple containing `ddl.TableSpec` instances. 

91 """ 

92 # We want the dataset_location and dataset_location_trash tables 

93 # to have the same definition, aside from the behavior of their link 

94 # to the dataset table: the trash table has no foreign key constraint. 

95 dataset_location_spec = ddl.TableSpec( 

96 doc=( 

97 "A table that provides information on whether a dataset is stored in " 

98 "one or more Datastores. The presence or absence of a record in this " 

99 "table itself indicates whether the dataset is present in that " 

100 "Datastore. " 

101 ), 

102 fields=NamedValueSet( 

103 [ 

104 ddl.FieldSpec( 

105 name="datastore_name", 

106 dtype=sqlalchemy.String, 

107 length=256, 

108 primaryKey=True, 

109 nullable=False, 

110 doc="Name of the Datastore this entry corresponds to.", 

111 ), 

112 ] 

113 ), 

114 ) 

115 dataset_location = copy.deepcopy(dataset_location_spec) 

116 datasets.addDatasetForeignKey(dataset_location, primaryKey=True) 

117 dataset_location_trash = copy.deepcopy(dataset_location_spec) 

118 datasets.addDatasetForeignKey(dataset_location_trash, primaryKey=True, constraint=False) 

119 return _TablesTuple( 

120 dataset_location=dataset_location, 

121 dataset_location_trash=dataset_location_trash, 

122 ) 

123 

124 

125class MonolithicDatastoreRegistryBridge(DatastoreRegistryBridge): 

126 """An implementation of `DatastoreRegistryBridge` that uses the same two 

127 tables for all non-ephemeral datastores. 

128 

129 Parameters 

130 ---------- 

131 datastoreName : `str` 

132 Name of the `Datastore` as it should appear in `Registry` tables 

133 referencing it. 

134 db : `Database` 

135 Object providing a database connection and generic distractions. 

136 tables : `_TablesTuple` 

137 Named tuple containing `sqlalchemy.schema.Table` instances. 

138 """ 

139 

140 def __init__(self, datastoreName: str, *, db: Database, tables: _TablesTuple): 

141 super().__init__(datastoreName) 

142 self._db = db 

143 self._tables = tables 

144 

145 def _refsToRows(self, refs: Iterable[DatasetIdRef]) -> list[dict]: 

146 """Transform an iterable of `DatasetRef` or `FakeDatasetRef` objects to 

147 a list of dictionaries that match the schema of the tables used by this 

148 class. 

149 

150 Parameters 

151 ---------- 

152 refs : `~collections.abc.Iterable` [ `DatasetRef` or `FakeDatasetRef` ] 

153 Datasets to transform. 

154 

155 Returns 

156 ------- 

157 rows : `list` [ `dict` ] 

158 List of dictionaries, with "datastoreName" and "dataset_id" keys. 

159 """ 

160 return [{"datastore_name": self.datastoreName, "dataset_id": ref.id} for ref in refs] 

161 

162 def ensure(self, refs: Iterable[DatasetIdRef]) -> None: 

163 # Docstring inherited from DatastoreRegistryBridge 

164 self._db.ensure(self._tables.dataset_location, *self._refsToRows(refs)) 

165 

166 def insert(self, refs: Iterable[DatasetIdRef]) -> None: 

167 # Docstring inherited from DatastoreRegistryBridge 

168 self._db.insert(self._tables.dataset_location, *self._refsToRows(refs)) 

169 

170 def forget(self, refs: Iterable[DatasetIdRef]) -> None: 

171 # Docstring inherited from DatastoreRegistryBridge 

172 rows = self._refsToRows(self.check(refs)) 

173 self._db.delete(self._tables.dataset_location, ["datastore_name", "dataset_id"], *rows) 

174 

175 def moveToTrash(self, refs: Iterable[DatasetIdRef], transaction: DatastoreTransaction | None) -> None: 

176 # Docstring inherited from DatastoreRegistryBridge 

177 # TODO: avoid self.check() call via queries like 

178 # INSERT INTO dataset_location_trash 

179 # SELECT datastore_name, dataset_id FROM dataset_location 

180 # WHERE datastore_name=? AND dataset_id IN (?); 

181 # DELETE FROM dataset_location 

182 # WHERE datastore_name=? AND dataset_id IN (?); 

183 # ...but the Database interface doesn't support those kinds of queries 

184 # right now. 

185 rows = self._refsToRows(self.check(refs)) 

186 with self._db.transaction(): 

187 self._db.delete(self._tables.dataset_location, ["datastore_name", "dataset_id"], *rows) 

188 self._db.insert(self._tables.dataset_location_trash, *rows) 

189 

190 def check(self, refs: Iterable[DatasetIdRef]) -> Iterable[DatasetIdRef]: 

191 # Docstring inherited from DatastoreRegistryBridge 

192 byId = {ref.id: ref for ref in refs} 

193 sql = ( 

194 sqlalchemy.sql.select(self._tables.dataset_location.columns.dataset_id) 

195 .select_from(self._tables.dataset_location) 

196 .where( 

197 sqlalchemy.sql.and_( 

198 self._tables.dataset_location.columns.datastore_name == self.datastoreName, 

199 self._tables.dataset_location.columns.dataset_id.in_(byId.keys()), 

200 ) 

201 ) 

202 ) 

203 with self._db.query(sql) as sql_result: 

204 sql_rows = sql_result.fetchall() 

205 for row in sql_rows: 

206 yield byId[row.dataset_id] 

207 

208 @contextmanager 

209 def emptyTrash( 

210 self, 

211 records_table: OpaqueTableStorage | None = None, 

212 record_class: type[StoredDatastoreItemInfo] | None = None, 

213 record_column: str | None = None, 

214 ) -> Iterator[tuple[Iterable[tuple[DatasetIdRef, StoredDatastoreItemInfo | None]], set[str] | None]]: 

215 # Docstring inherited from DatastoreRegistryBridge 

216 

217 if records_table is None: 

218 raise ValueError("This implementation requires a records table.") 

219 

220 assert isinstance( 

221 records_table, ByNameOpaqueTableStorage 

222 ), f"Records table must support hidden attributes. Got {type(records_table)}." 

223 

224 if record_class is None: 

225 raise ValueError("Record class must be provided if records table is given.") 

226 

227 # Helper closure to generate the common join+where clause. 

228 def join_records( 

229 select: sqlalchemy.sql.Select, location_table: sqlalchemy.schema.Table 

230 ) -> sqlalchemy.sql.Select: 

231 # mypy needs to be sure 

232 assert isinstance(records_table, ByNameOpaqueTableStorage) 

233 return select.select_from( 

234 records_table._table.join( 

235 location_table, 

236 onclause=records_table._table.columns.dataset_id == location_table.columns.dataset_id, 

237 ) 

238 ).where(location_table.columns.datastore_name == self.datastoreName) 

239 

240 # SELECT records.dataset_id, records.path FROM records 

241 # JOIN records on dataset_location.dataset_id == records.dataset_id 

242 # WHERE dataset_location.datastore_name = datastoreName 

243 

244 # It's possible that we may end up with a ref listed in the trash 

245 # table that is not listed in the records table. Such an 

246 # inconsistency would be missed by this query. 

247 info_in_trash = join_records(records_table._table.select(), self._tables.dataset_location_trash) 

248 

249 # Run query, transform results into a list of dicts that we can later 

250 # use to delete. 

251 with self._db.query(info_in_trash) as sql_result: 

252 rows = [dict(row, datastore_name=self.datastoreName) for row in sql_result.mappings()] 

253 

254 # It is possible for trashed refs to be linked to artifacts that 

255 # are still associated with refs that are not to be trashed. We 

256 # need to be careful to consider those and indicate to the caller 

257 # that those artifacts should be retained. Can only do this check 

258 # if the caller provides a column name that can map to multiple 

259 # refs. 

260 preserved: set[str] | None = None 

261 if record_column is not None: 

262 # Some helper subqueries 

263 items_not_in_trash = join_records( 

264 sqlalchemy.sql.select(records_table._table.columns[record_column]), 

265 self._tables.dataset_location, 

266 ).alias("items_not_in_trash") 

267 items_in_trash = join_records( 

268 sqlalchemy.sql.select(records_table._table.columns[record_column]), 

269 self._tables.dataset_location_trash, 

270 ).alias("items_in_trash") 

271 

272 # A query for paths that are referenced by datasets in the trash 

273 # and datasets not in the trash. 

274 items_to_preserve = sqlalchemy.sql.select(items_in_trash.columns[record_column]).select_from( 

275 items_not_in_trash.join( 

276 items_in_trash, 

277 onclause=items_in_trash.columns[record_column] 

278 == items_not_in_trash.columns[record_column], 

279 ) 

280 ) 

281 with self._db.query(items_to_preserve) as sql_result: 

282 preserved = {row[record_column] for row in sql_result.mappings()} 

283 

284 # Convert results to a tuple of id+info and a record of the artifacts 

285 # that should not be deleted from datastore. The id+info tuple is 

286 # solely to allow logging to report the relevant ID. 

287 id_info = ((FakeDatasetRef(row["dataset_id"]), record_class.from_record(row)) for row in rows) 

288 

289 # Start contextmanager, return results 

290 yield ((id_info, preserved)) 

291 

292 # No exception raised in context manager block. 

293 if not rows: 

294 return 

295 

296 # Delete the rows from the records table 

297 records_table.delete(["dataset_id"], *[{"dataset_id": row["dataset_id"]} for row in rows]) 

298 

299 # Delete those rows from the trash table. 

300 self._db.delete( 

301 self._tables.dataset_location_trash, 

302 ["dataset_id", "datastore_name"], 

303 *[{"dataset_id": row["dataset_id"], "datastore_name": row["datastore_name"]} for row in rows], 

304 ) 

305 

306 

307class MonolithicDatastoreRegistryBridgeManager(DatastoreRegistryBridgeManager): 

308 """An implementation of `DatastoreRegistryBridgeManager` that uses the same 

309 two tables for all non-ephemeral datastores. 

310 

311 Parameters 

312 ---------- 

313 db : `Database` 

314 Object providing a database connection and generic distractions. 

315 tables : `_TablesTuple` 

316 Named tuple containing `sqlalchemy.schema.Table` instances. 

317 opaque : `OpaqueTableStorageManager` 

318 Manager object for opaque table storage in the `Registry`. 

319 universe : `DimensionUniverse` 

320 All dimensions know to the `Registry`. 

321 datasetIdColumnType : `type` 

322 Type for dataset ID column. 

323 """ 

324 

325 def __init__( 

326 self, 

327 *, 

328 db: Database, 

329 tables: _TablesTuple, 

330 opaque: OpaqueTableStorageManager, 

331 universe: DimensionUniverse, 

332 datasetIdColumnType: type, 

333 registry_schema_version: VersionTuple | None = None, 

334 ): 

335 super().__init__( 

336 opaque=opaque, 

337 universe=universe, 

338 datasetIdColumnType=datasetIdColumnType, 

339 registry_schema_version=registry_schema_version, 

340 ) 

341 self._db = db 

342 self._tables = tables 

343 self._ephemeral: dict[str, EphemeralDatastoreRegistryBridge] = {} 

344 

345 @classmethod 

346 def initialize( 

347 cls, 

348 db: Database, 

349 context: StaticTablesContext, 

350 *, 

351 opaque: OpaqueTableStorageManager, 

352 datasets: type[DatasetRecordStorageManager], 

353 universe: DimensionUniverse, 

354 registry_schema_version: VersionTuple | None = None, 

355 ) -> DatastoreRegistryBridgeManager: 

356 # Docstring inherited from DatastoreRegistryBridge 

357 tables = context.addTableTuple(_makeTableSpecs(datasets)) 

358 return cls( 

359 db=db, 

360 tables=cast(_TablesTuple, tables), 

361 opaque=opaque, 

362 universe=universe, 

363 datasetIdColumnType=datasets.getIdColumnType(), 

364 registry_schema_version=registry_schema_version, 

365 ) 

366 

367 def refresh(self) -> None: 

368 # Docstring inherited from DatastoreRegistryBridge 

369 # This implementation has no in-Python state that depends on which 

370 # datastores exist, so there's nothing to do. 

371 pass 

372 

373 def register(self, name: str, *, ephemeral: bool = False) -> DatastoreRegistryBridge: 

374 # Docstring inherited from DatastoreRegistryBridge 

375 if ephemeral: 

376 return self._ephemeral.setdefault(name, EphemeralDatastoreRegistryBridge(name)) 

377 return MonolithicDatastoreRegistryBridge(name, db=self._db, tables=self._tables) 

378 

379 def findDatastores(self, ref: DatasetIdRef) -> Iterable[str]: 

380 # Docstring inherited from DatastoreRegistryBridge 

381 sql = ( 

382 sqlalchemy.sql.select(self._tables.dataset_location.columns.datastore_name) 

383 .select_from(self._tables.dataset_location) 

384 .where(self._tables.dataset_location.columns.dataset_id == ref.id) 

385 ) 

386 with self._db.query(sql) as sql_result: 

387 sql_rows = sql_result.mappings().fetchall() 

388 for row in sql_rows: 

389 yield row[self._tables.dataset_location.columns.datastore_name] 

390 for name, bridge in self._ephemeral.items(): 

391 if ref in bridge: 

392 yield name 

393 

394 @classmethod 

395 def currentVersions(cls) -> list[VersionTuple]: 

396 # Docstring inherited from VersionedExtension. 

397 return [_VERSION]