Hide keyboard shortcuts

Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

1# This file is part of obs_base. 

2# 

3# Developed for the LSST Data Management System. 

4# This product includes software developed by the LSST Project 

5# (https://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__ = ["RepoConverter"] 

24 

25from dataclasses import dataclass 

26from collections import defaultdict 

27from abc import ABC, abstractmethod 

28import fnmatch 

29import re 

30from typing import ( 

31 Dict, 

32 Iterator, 

33 List, 

34 MutableMapping, 

35 Optional, 

36 Set, 

37 Tuple, 

38 Union, 

39 TYPE_CHECKING, 

40) 

41 

42from lsst.utils import doImport 

43from lsst.daf.butler import DataCoordinate, FileDataset, DatasetType 

44from lsst.sphgeom import RangeSet, Region 

45from .repoWalker import RepoWalker 

46 

47if TYPE_CHECKING: 47 ↛ 48line 47 didn't jump to line 48, because the condition on line 47 was never true

48 from ..mapping import Mapping as CameraMapperMapping # disambiguate from collections.abc.Mapping 

49 from .convertRepo import ConvertRepoTask 

50 from .scanner import PathElementHandler 

51 from lsst.daf.butler import StorageClass, Registry, SkyPixDimension, FormatterParameter 

52 

53 

54@dataclass 

55class ConversionSubset: 

56 """A helper class for `ConvertRepoTask` and `RepoConverter` that maintains 

57 lists of related data ID values that should be included in the conversion. 

58 

59 Parameters 

60 ---------- 

61 instrument : `str` 

62 Instrument name used in Gen3 data IDs. 

63 visits : `set` of `int` 

64 Visit IDs that define the filter. 

65 """ 

66 

67 def __init__(self, instrument: str, visits: Set[int]): 

68 self.instrument = instrument 

69 self.visits = visits 

70 self.regions = None 

71 self.tracts = {} 

72 self.skypix = {} 

73 

74 def addSkyMap(self, registry: Registry, name: str): 

75 """Populate the included tract IDs for the given skymap from those that 

76 overlap the visits the `ConversionSubset` was initialized with. 

77 

78 Parameters 

79 ---------- 

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

81 Registry that can be queried for visit/tract overlaps. 

82 name : `str` 

83 SkyMap name used in Gen3 data IDs. 

84 """ 

85 tracts = set() 

86 self.tracts[name] = tracts 

87 for visit in self.visits: 

88 for dataId in registry.queryDimensions(["tract"], expand=False, 

89 dataId={"skymap": name, 

90 "instrument": self.instrument, 

91 "visit": visit}): 

92 tracts.add(dataId["tract"]) 

93 

94 def addSkyPix(self, registry: Registry, dimension: SkyPixDimension): 

95 """Populate the included skypix IDs for the given dimension from those 

96 that overlap the visits the `ConversionSubset` was initialized with. 

97 

98 Parameters 

99 ---------- 

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

101 Registry that can be queried for visit regions. 

102 name : `str` 

103 SkyMap name used in Gen3 data IDs. 

104 """ 

105 if self.regions is None: 

106 self.regions = [] 

107 for visit in self.visits: 

108 dataId = registry.expandDataId(instrument=self.instrument, visit=visit) 

109 self.regions.append(dataId.region) 

110 ranges = RangeSet() 

111 for region in self.regions: 

112 ranges = ranges.union(dimension.pixelization.envelope(region)) 

113 self.skypix[dimension] = ranges 

114 

115 def isRelated(self, dataId: DataCoordinate) -> bool: 

116 """Test whether the given data ID is related to this subset and hence 

117 should be included in a repository conversion. 

118 

119 Parameters 

120 ---------- 

121 dataId : `lsst.daf.butler.DataCoordinate` 

122 Data ID to test. 

123 

124 Returns 

125 ------- 

126 related : `bool` 

127 `True` if this data ID should be included in a repository 

128 conversion. 

129 

130 Notes 

131 ----- 

132 More formally, this tests that the given data ID is not unrelated; 

133 if a data ID does not involve tracts, visits, or skypix dimensions, 

134 we always include it. 

135 """ 

136 if self.visits is None: 

137 # We're not filtering at all. 

138 return True 

139 if "visit" in dataId.graph and dataId["visit"] not in self.visits: 

140 return False 

141 if "tract" in dataId.graph and dataId["tract"] not in self.tracts[dataId["skymap"]]: 

142 return False 

143 for dimension, ranges in self.skypix.items(): 

144 if dimension in dataId.graph and not ranges.intersects(dataId[dimension]): 

145 return False 

146 return True 

147 

148 # Class attributes that will be shadowed by public instance attributes; 

149 # defined here only for documentation purposes. 

150 

151 instrument: str 

152 """The name of the instrument, as used in Gen3 data IDs (`str`). 

153 """ 

154 

155 visits: Set[int] 

156 """The set of visit IDs that should be included in the conversion (`set` 

157 of `int`). 

158 """ 

159 

160 regions: Optional[List[Region]] 

161 """Regions for all visits (`list` of `lsst.sphgeom.Region`). 

162 

163 Set to `None` before it has been initialized. Any code that attempts to 

164 use it when it is `None` has a logic bug. 

165 """ 

166 

167 tracts: Dict[str, Set[int]] 

168 """Tracts that should be included in the conversion, grouped by skymap 

169 name (`dict` mapping `str` to `set` of `int`). 

170 """ 

171 

172 skypix: Dict[SkyPixDimension, RangeSet] 

173 """SkyPix ranges that should be included in the conversion, grouped by 

174 dimension (`dict` mapping `SkyPixDimension` to `lsst.sphgeom.RangeSet`). 

175 """ 

176 

177 

178class RepoConverter(ABC): 

179 """An abstract base class for objects that help `ConvertRepoTask` convert 

180 datasets from a single Gen2 repository. 

181 

182 Parameters 

183 ---------- 

184 task : `ConvertRepoTask` 

185 Task instance that is using this helper object. 

186 root : `str` 

187 Root of the Gen2 repo being converted. 

188 collections : `list` of `str` 

189 Gen3 collections with which all converted datasets should be 

190 associated. 

191 subset : `ConversionSubset, optional 

192 Helper object that implements a filter that restricts the data IDs that 

193 are converted. 

194 

195 Notes 

196 ----- 

197 `RepoConverter` defines the only public API users of its subclasses should 

198 use (`prep`, `insertDimensionRecords`, and `ingest`). These delegate to 

199 several abstract methods that subclasses must implement. In some cases, 

200 subclasses may reimplement the public methods as well, but are expected to 

201 delegate to ``super()`` either at the beginning or end of their own 

202 implementation. 

203 """ 

204 

205 def __init__(self, *, task: ConvertRepoTask, root: str, run: Optional[str], 

206 subset: Optional[ConversionSubset] = None): 

207 self.task = task 

208 self.root = root 

209 self.subset = subset 

210 self._run = run 

211 self._repoWalker = None # Created in prep 

212 self._fileDatasets: MutableMapping[DatasetType, List[FileDataset]] = defaultdict(list) 

213 

214 @abstractmethod 

215 def isDatasetTypeSpecial(self, datasetTypeName: str) -> bool: 

216 """Test whether the given dataset is handled specially by this 

217 converter and hence should be ignored by generic base-class logic that 

218 searches for dataset types to convert. 

219 

220 Parameters 

221 ---------- 

222 datasetTypeName : `str` 

223 Name of the dataset type to test. 

224 

225 Returns 

226 ------- 

227 special : `bool` 

228 `True` if the dataset type is special. 

229 """ 

230 raise NotImplementedError() 

231 

232 @abstractmethod 

233 def iterMappings(self) -> Iterator[Tuple[str, CameraMapperMapping]]: 

234 """Iterate over all `CameraMapper` `Mapping` objects that should be 

235 considered for conversion by this repository. 

236 

237 This this should include any datasets that may appear in the 

238 repository, including those that are special (see 

239 `isDatasetTypeSpecial`) and those that are being ignored (see 

240 `ConvertRepoTask.isDatasetTypeIncluded`); this allows the converter 

241 to identify and hence skip these datasets quietly instead of warning 

242 about them as unrecognized. 

243 

244 Yields 

245 ------ 

246 datasetTypeName: `str` 

247 Name of the dataset type. 

248 mapping : `lsst.obs.base.mapping.Mapping` 

249 Mapping object used by the Gen2 `CameraMapper` to describe the 

250 dataset type. 

251 """ 

252 raise NotImplementedError() 

253 

254 @abstractmethod 

255 def makeRepoWalkerTarget(self, datasetTypeName: str, template: str, keys: Dict[str, type], 

256 storageClass: StorageClass, 

257 formatter: FormatterParameter = None, 

258 targetHandler: Optional[PathElementHandler] = None, 

259 ) -> RepoWalker.Target: 

260 """Make a struct that identifies a dataset type to be extracted by 

261 walking the repo directory structure. 

262 

263 Parameters 

264 ---------- 

265 datasetTypeName : `str` 

266 Name of the dataset type (the same in both Gen2 and Gen3). 

267 template : `str` 

268 The full Gen2 filename template. 

269 keys : `dict` [`str`, `type`] 

270 A dictionary mapping Gen2 data ID key to the type of its value. 

271 storageClass : `lsst.daf.butler.StorageClass` 

272 Gen3 storage class for this dataset type. 

273 formatter : `lsst.daf.butler.Formatter` or `str`, optional 

274 A Gen 3 formatter class or fully-qualified name. 

275 targetHandler : `PathElementHandler`, optional 

276 Specialist target handler to use for this dataset type. 

277 

278 Returns 

279 ------- 

280 target : `RepoWalker.Target` 

281 A struct containing information about the target dataset (much of 

282 it simplify forwarded from the arguments). 

283 """ 

284 raise NotImplementedError() 

285 

286 def getSpecialDirectories(self) -> List[str]: 

287 """Return a list of directory paths that should not be searched for 

288 files. 

289 

290 These may be directories that simply do not contain datasets (or 

291 contain datasets in another repository), or directories whose datasets 

292 are handled specially by a subclass. 

293 

294 Returns 

295 ------- 

296 directories : `list` [`str`] 

297 The full paths of directories to skip, relative to the repository 

298 root. 

299 """ 

300 return [] 

301 

302 def prep(self): 

303 """Perform preparatory work associated with the dataset types to be 

304 converted from this repository (but not the datasets themselves). 

305 

306 Notes 

307 ----- 

308 This should be a relatively fast operation that should not depend on 

309 the size of the repository. 

310 

311 Subclasses may override this method, but must delegate to the base 

312 class implementation at some point in their own logic. 

313 More often, subclasses will specialize the behavior of `prep` by 

314 overriding other methods to which the base class implementation 

315 delegates. These include: 

316 - `iterMappings` 

317 - `isDatasetTypeSpecial` 

318 - `getSpecialDirectories` 

319 - `makeRepoWalkerTarget` 

320 

321 This should not perform any write operations to the Gen3 repository. 

322 It is guaranteed to be called before `insertDimensionData`. 

323 """ 

324 self.task.log.info(f"Preparing other dataset types from root {self.root}.") 

325 walkerInputs: List[Union[RepoWalker.Target, RepoWalker.Skip]] = [] 

326 for datasetTypeName, mapping in self.iterMappings(): 

327 try: 

328 template = mapping.template 

329 except RuntimeError: 

330 # No template for this dataset in this mapper, so there's no 

331 # way there should be instances of this dataset in this repo. 

332 continue 

333 extensions = [""] 

334 skip = False 

335 message = None 

336 storageClass = None 

337 if (not self.task.isDatasetTypeIncluded(datasetTypeName) 

338 or self.isDatasetTypeSpecial(datasetTypeName)): 

339 # User indicated not to include this data, but we still want 

340 # to recognize files of that type to avoid warning about them. 

341 skip = True 

342 else: 

343 storageClass = self._guessStorageClass(datasetTypeName, mapping) 

344 if storageClass is None: 

345 # This may be a problem, but only if we actually encounter any 

346 # files corresponding to this dataset. Of course, we need 

347 # to be able to parse those files in order to recognize that 

348 # situation. 

349 message = f"no storage class found for {datasetTypeName}" 

350 skip = True 

351 # Handle files that are compressed on disk, but the gen2 template is just `.fits` 

352 if template.endswith(".fits"): 

353 extensions.extend((".gz", ".fz")) 

354 for extension in extensions: 

355 if skip: 

356 walkerInput = RepoWalker.Skip( 

357 template=template+extension, 

358 keys=mapping.keys(), 

359 message=message, 

360 ) 

361 self.task.log.debug("Skipping template in walker: %s", template) 

362 else: 

363 assert message is None 

364 targetHandler = self.task.config.targetHandlerClasses.get(datasetTypeName) 

365 if targetHandler is not None: 

366 targetHandler = doImport(targetHandler) 

367 walkerInput = self.makeRepoWalkerTarget( 

368 datasetTypeName=datasetTypeName, 

369 template=template+extension, 

370 keys=mapping.keys(), 

371 storageClass=storageClass, 

372 formatter=self.task.config.formatterClasses.get(datasetTypeName), 

373 targetHandler=targetHandler, 

374 ) 

375 self.task.log.debug("Adding template to walker: %s", template) 

376 walkerInputs.append(walkerInput) 

377 

378 for dirPath in self.getSpecialDirectories(): 

379 walkerInputs.append( 

380 RepoWalker.Skip( 

381 template=dirPath, # not really a template, but that's fine; it's relative to root. 

382 keys={}, 

383 message=None, 

384 isForFiles=True, 

385 ) 

386 ) 

387 fileIgnoreRegExTerms = [] 

388 for pattern in self.task.config.fileIgnorePatterns: 

389 fileIgnoreRegExTerms.append(fnmatch.translate(pattern)) 

390 if fileIgnoreRegExTerms: 

391 fileIgnoreRegEx = re.compile("|".join(fileIgnoreRegExTerms)) 

392 else: 

393 fileIgnoreRegEx = None 

394 self._repoWalker = RepoWalker(walkerInputs, fileIgnoreRegEx=fileIgnoreRegEx) 

395 

396 def iterDatasets(self) -> Iterator[FileDataset]: 

397 """Iterate over datasets in the repository that should be ingested into 

398 the Gen3 repository. 

399 

400 The base class implementation yields nothing; the datasets handled by 

401 the `RepoConverter` base class itself are read directly in 

402 `findDatasets`. 

403 

404 Subclasses should override this method if they support additional 

405 datasets that are handled some other way. 

406 

407 Yields 

408 ------ 

409 dataset : `FileDataset` 

410 Structures representing datasets to be ingested. Paths should be 

411 absolute. 

412 """ 

413 yield from () 

414 

415 def findDatasets(self): 

416 assert self._repoWalker, "prep() must be called before findDatasets." 

417 self.task.log.info("Adding special datasets in repo %s.", self.root) 

418 for dataset in self.iterDatasets(): 

419 assert len(dataset.refs) == 1 

420 self._fileDatasets[dataset.refs[0].datasetType].append(dataset) 

421 self.task.log.info("Finding datasets from files in repo %s.", self.root) 

422 self._fileDatasets.update( 

423 self._repoWalker.walk( 

424 self.root, 

425 log=self.task.log, 

426 predicate=(self.subset.isRelated if self.subset is not None else None) 

427 ) 

428 ) 

429 

430 def insertDimensionData(self): 

431 """Insert any dimension records uniquely derived from this repository 

432 into the registry. 

433 

434 Subclasses may override this method, but may not need to; the default 

435 implementation does nothing. 

436 

437 SkyMap and SkyPix dimensions should instead be handled by calling 

438 `ConvertRepoTask.useSkyMap` or `ConvertRepoTask.useSkyPix`, because 

439 these dimensions are in general shared by multiple Gen2 repositories. 

440 

441 This method is guaranteed to be called between `prep` and 

442 `expandDataIds`. 

443 """ 

444 pass 

445 

446 def expandDataIds(self): 

447 """Expand the data IDs for all datasets to be inserted. 

448 

449 Subclasses may override this method, but must delegate to the base 

450 class implementation if they do. 

451 

452 This involves queries to the registry, but not writes. It is 

453 guaranteed to be called between `insertDimensionData` and `ingest`. 

454 """ 

455 import itertools 

456 for datasetType, datasetsForType in self._fileDatasets.items(): 

457 self.task.log.info("Expanding data IDs for %s %s datasets.", len(datasetsForType), 

458 datasetType.name) 

459 expanded = [] 

460 for dataset in datasetsForType: 

461 for i, ref in enumerate(dataset.refs): 

462 try: 

463 dataId = self.task.registry.expandDataId(ref.dataId) 

464 dataset.refs[i] = ref.expanded(dataId) 

465 except LookupError as err: 

466 self.task.log.warn("Skipping ingestion for '%s': %s", dataset.path, err) 

467 # Remove skipped datasets from multi-extension FileDatasets 

468 dataset.refs[i] = None # We will strip off the `None`s after the loop. 

469 dataset.refs[:] = itertools.filterfalse(lambda x: x is None, dataset.refs) 

470 if dataset.refs: 

471 expanded.append(dataset) 

472 

473 datasetsForType[:] = expanded 

474 

475 def ingest(self): 

476 """Insert converted datasets into the Gen3 repository. 

477 

478 Subclasses may override this method, but must delegate to the base 

479 class implementation at some point in their own logic. 

480 

481 This method is guaranteed to be called after `expandDataIds`. 

482 """ 

483 for datasetType, datasetsForType in self._fileDatasets.items(): 

484 self.task.registry.registerDatasetType(datasetType) 

485 try: 

486 run = self.getRun(datasetType.name) 

487 except LookupError: 

488 self.task.log.warn(f"No run configured for dataset type {datasetType.name}.") 

489 continue 

490 self.task.log.info("Ingesting %s %s datasets into run %s.", len(datasetsForType), 

491 datasetType.name, run) 

492 try: 

493 self.task.registry.registerRun(run) 

494 self.task.butler3.ingest(*datasetsForType, transfer=self.task.config.transfer, run=run) 

495 except LookupError as err: 

496 raise LookupError(f"Error expanding data ID for dataset type {datasetType.name}.") from err 

497 

498 def getRun(self, datasetTypeName: str) -> str: 

499 """Return the name of the run to insert instances of the given dataset 

500 type into in this collection. 

501 

502 Parameters 

503 ---------- 

504 datasetTypeName : `str` 

505 Name of the dataset type. 

506 

507 Returns 

508 ------- 

509 run : `str` 

510 Name of the `~lsst.daf.butler.CollectionType.RUN` collection. 

511 """ 

512 assert self._run is not None, "Method must be overridden if self._run is allowed to be None" 

513 return self._run 

514 

515 def _guessStorageClass(self, datasetTypeName: str, mapping: CameraMapperMapping 

516 ) -> Optional[StorageClass]: 

517 """Infer the Gen3 `StorageClass` from a dataset from a combination of 

518 configuration and Gen2 dataset type information. 

519 

520 datasetTypeName: `str` 

521 Name of the dataset type. 

522 mapping : `lsst.obs.base.mapping.Mapping` 

523 Mapping object used by the Gen2 `CameraMapper` to describe the 

524 dataset type. 

525 """ 

526 storageClassName = self.task.config.storageClasses.get(datasetTypeName) 

527 if storageClassName is None and mapping.python is not None: 

528 storageClassName = self.task.config.storageClasses.get(mapping.python, None) 

529 if storageClassName is None and mapping.persistable is not None: 

530 storageClassName = self.task.config.storageClasses.get(mapping.persistable, None) 

531 if storageClassName is None and mapping.python is not None: 

532 unqualified = mapping.python.split(".")[-1] 

533 storageClassName = self.task.config.storageClasses.get(unqualified, None) 

534 if storageClassName is not None: 

535 storageClass = self.task.butler3.storageClasses.getStorageClass(storageClassName) 

536 else: 

537 try: 

538 storageClass = self.task.butler3.storageClasses.getStorageClass(mapping.persistable) 

539 except KeyError: 

540 storageClass = None 

541 if storageClass is None and mapping.python is not None: 

542 try: 

543 storageClass = self.task.butler3.storageClasses.getStorageClass(unqualified) 

544 except KeyError: 

545 pass 

546 if storageClass is None: 

547 self.task.log.debug("No StorageClass found for %s; skipping.", datasetTypeName) 

548 else: 

549 self.task.log.debug("Using StorageClass %s for %s.", storageClass.name, datasetTypeName) 

550 return storageClass 

551 

552 # Class attributes that will be shadowed by public instance attributes; 

553 # defined here only for documentation purposes. 

554 

555 task: ConvertRepoTask 

556 """The parent task that constructed and uses this converter 

557 (`ConvertRepoTask`). 

558 """ 

559 

560 root: str 

561 """Root path to the Gen2 repository this converter manages (`str`). 

562 

563 This is a complete path, not relative to some other repository root. 

564 """ 

565 

566 subset: Optional[ConversionSubset] 

567 """An object that represents a filter to be applied to the datasets that 

568 are converted (`ConversionSubset` or `None`). 

569 """