Coverage for python/lsst/obs/base/gen2to3/convertRepo.py : 29%

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
23__all__ = ["ConvertRepoConfig", "ConvertRepoTask", "ConvertRepoSkyMapConfig", "Rerun"]
25import os
26import fnmatch
27from dataclasses import dataclass
28from multiprocessing import Pool
29from typing import Iterable, Optional, List, Dict
31from lsst.daf.butler import (
32 Butler as Butler3,
33 CollectionType,
34 SkyPixDimension
35)
36from lsst.pex.config import Config, ConfigurableField, ConfigDictField, DictField, ListField, Field
37from lsst.pipe.base import Task
38from lsst.skymap import skyMapRegistry, BaseSkyMap
40from ..ingest import RawIngestTask
41from ..defineVisits import DefineVisitsTask
42from .repoConverter import ConversionSubset
43from .rootRepoConverter import RootRepoConverter
44from .calibRepoConverter import CalibRepoConverter
45from .standardRepoConverter import StandardRepoConverter
46from .._instrument import Instrument
49@dataclass
50class ConfiguredSkyMap:
51 """Struct containing information about a skymap that may appear in a Gen2
52 repository.
53 """
55 name: str
56 """Name of the skymap used in Gen3 data IDs.
57 """
59 sha1: bytes
60 """Hash computed by `BaseSkyMap.getSha1`.
61 """
63 instance: BaseSkyMap
64 """Name of the skymap used in Gen3 data IDs.
65 """
67 used: bool = False
68 """Whether this skymap has been found in at least one repository being
69 converted.
70 """
73@dataclass
74class Rerun:
75 """Specification for a Gen2 processing-output repository to convert.
76 """
78 path: str
79 """Absolute or relative (to the root repository) path to the Gen2
80 repository (`str`).
81 """
83 runName: str
84 """Name of the `~lsst.daf.butler.CollectionType.RUN` collection datasets
85 will be inserted into (`str`).
86 """
88 chainName: Optional[str]
89 """Name of a `~lsst.daf.butler.CollectionType.CHAINED` collection that will
90 combine this repository's datasets with those of its parent repositories
91 (`str`, optional).
92 """
94 parents: List[str]
95 """Collection names associated with parent repositories, used to define the
96 chained collection (`list` [ `str` ]).
98 Ignored if `chainName` is `None`. Runs used in the root repo are
99 automatically included.
100 """
103class ConvertRepoSkyMapConfig(Config):
104 """Sub-config used to hold the parameters of a SkyMap.
106 Notes
107 -----
108 This config only needs to exist because we can't put a
109 `~lsst.pex.config.RegistryField` directly inside a
110 `~lsst.pex.config.ConfigDictField`.
112 It needs to have its only field named "skyMap" for compatibility with the
113 configuration of `lsst.pipe.tasks.MakeSkyMapTask`, which we want so we can
114 use one config file in an obs package to configure both.
116 This name leads to unfortunate repetition with the field named
117 "skymap" that holds it - "skyMap[name].skyMap" - but that seems
118 unavoidable.
119 """
120 skyMap = skyMapRegistry.makeField(
121 doc="Type and parameters for the SkyMap itself.",
122 default="dodeca",
123 )
126class ConvertRepoConfig(Config):
127 raws = ConfigurableField(
128 "Configuration for subtask responsible for ingesting raws and adding "
129 "exposure dimension entries.",
130 target=RawIngestTask,
131 )
132 defineVisits = ConfigurableField(
133 "Configuration for the subtask responsible for defining visits from "
134 "exposures.",
135 target=DefineVisitsTask,
136 )
137 skyMaps = ConfigDictField(
138 "Mapping from Gen3 skymap name to the parameters used to construct a "
139 "BaseSkyMap instance. This will be used to associate names with "
140 "existing skymaps found in the Gen2 repo.",
141 keytype=str,
142 itemtype=ConvertRepoSkyMapConfig,
143 default={}
144 )
145 rootSkyMapName = Field(
146 "Name of a Gen3 skymap (an entry in ``self.skyMaps``) to assume for "
147 "datasets in the root repository when no SkyMap is found there. ",
148 dtype=str,
149 optional=True,
150 default=None,
151 )
152 runs = DictField(
153 "A mapping from dataset type name to the RUN collection they should "
154 "be inserted into. This must include all datasets that can be found "
155 "in the root repository; other repositories will use per-repository "
156 "runs.",
157 keytype=str,
158 itemtype=str,
159 default={
160 "deepCoadd_skyMap": "skymaps",
161 }
162 )
163 runsForced = DictField(
164 "Like ``runs``, but is used even when the dataset is present in a "
165 "non-root repository (i.e. rerun), overriding the non-root "
166 "repository's main collection.",
167 keytype=str,
168 itemtype=str,
169 default={
170 "brightObjectMask": "masks",
171 }
172 )
173 storageClasses = DictField(
174 "Mapping from dataset type name or Gen2 policy entry (e.g. 'python' "
175 "or 'persistable') to the Gen3 StorageClass name.",
176 keytype=str,
177 itemtype=str,
178 default={
179 "bias": "ExposureF",
180 "dark": "ExposureF",
181 "flat": "ExposureF",
182 "defects": "Defects",
183 "crosstalk": "CrosstalkCalib",
184 "BaseSkyMap": "SkyMap",
185 "BaseCatalog": "Catalog",
186 "BackgroundList": "Background",
187 "raw": "Exposure",
188 "MultilevelParquetTable": "DataFrame",
189 "ParquetTable": "DataFrame",
190 "SkyWcs": "Wcs",
191 }
192 )
193 formatterClasses = DictField(
194 "Mapping from dataset type name to formatter class. "
195 "By default these are derived from the formatters listed in the"
196 " Gen3 datastore configuration.",
197 keytype=str,
198 itemtype=str,
199 default={}
200 )
201 targetHandlerClasses = DictField(
202 "Mapping from dataset type name to target handler class.",
203 keytype=str,
204 itemtype=str,
205 default={}
206 )
207 doRegisterInstrument = Field(
208 "If True (default), add dimension records for the Instrument and its "
209 "filters and detectors to the registry instead of assuming they are "
210 "already present.",
211 dtype=bool,
212 default=True,
213 )
214 doWriteCuratedCalibrations = Field(
215 "If True (default), ingest human-curated calibrations directly via "
216 "the Instrument interface. Note that these calibrations are never "
217 "converted from Gen2 repositories.",
218 dtype=bool,
219 default=True,
220 )
221 refCats = ListField(
222 "The names of reference catalogs (subdirectories under ref_cats) to "
223 "be converted",
224 dtype=str,
225 default=[]
226 )
227 fileIgnorePatterns = ListField(
228 "Filename globs that should be ignored instead of being treated as "
229 "datasets.",
230 dtype=str,
231 default=["README.txt", "*~?", "butler.yaml", "gen3.sqlite3",
232 "registry.sqlite3", "calibRegistry.sqlite3", "_mapper",
233 "_parent", "repositoryCfg.yaml"]
234 )
235 rawDatasetType = Field(
236 "Gen2 dataset type to use for raw data.",
237 dtype=str,
238 default="raw",
239 )
240 datasetIncludePatterns = ListField(
241 "Glob-style patterns for dataset type names that should be converted.",
242 dtype=str,
243 default=["*"]
244 )
245 datasetIgnorePatterns = ListField(
246 "Glob-style patterns for dataset type names that should not be "
247 "converted despite matching a pattern in datasetIncludePatterns.",
248 dtype=str,
249 default=[]
250 )
251 ccdKey = Field(
252 "Key used for the Gen2 equivalent of 'detector' in data IDs.",
253 dtype=str,
254 default="ccd",
255 )
256 relatedOnly = Field(
257 "If True (default), only convert datasets that are related to the "
258 "ingested visits. Ignored unless a list of visits is passed to "
259 "run().",
260 dtype=bool,
261 default=False,
262 )
264 @property
265 def transfer(self):
266 return self.raws.transfer
268 @transfer.setter
269 def transfer(self, value):
270 self.raws.transfer = value
272 def setDefaults(self):
273 self.transfer = None
275 # TODO: check that there are no collection overrides for curated
276 # calibrations, since we don't have a good way to utilize them.
279class ConvertRepoTask(Task):
280 """A task that converts one or more related Gen2 data repositories to a
281 single Gen3 data repository (with multiple collections).
283 Parameters
284 ----------
285 config: `ConvertRepoConfig`
286 Configuration for this task.
287 butler3: `lsst.daf.butler.Butler`
288 A writeable Gen3 Butler instance that represents the data repository
289 that datasets will be ingested into. If the 'raw' dataset is
290 configured to be included in the conversion, ``butler3.run`` should be
291 set to the name of the collection raws should be ingested into, and
292 ``butler3.collections`` should include a calibration collection from
293 which the ``camera`` dataset can be loaded, unless a calibration repo
294 is converted and ``doWriteCuratedCalibrations`` is `True`.
295 instrument : `lsst.obs.base.Instrument`
296 The Gen3 instrument that should be used for this conversion.
297 **kwargs
298 Other keyword arguments are forwarded to the `Task` constructor.
300 Notes
301 -----
302 Most of the work of converting repositories is delegated to instances of
303 the `RepoConverter` hierarchy. The `ConvertRepoTask` instance itself holds
304 only state that is relevant for all Gen2 repositories being ingested, while
305 each `RepoConverter` instance holds only state relevant for the conversion
306 of a single Gen2 repository. Both the task and the `RepoConverter`
307 instances are single use; `ConvertRepoTask.run` and most `RepoConverter`
308 methods may only be called once on a particular instance.
309 """
311 ConfigClass = ConvertRepoConfig
313 _DefaultName = "convertRepo"
315 def __init__(self, config=None, *, butler3: Butler3, instrument: Instrument, **kwargs):
316 config.validate() # Not a CmdlineTask nor PipelineTask, so have to validate the config here.
317 super().__init__(config, **kwargs)
318 self.butler3 = butler3
319 self.registry = self.butler3.registry
320 self.universe = self.registry.dimensions
321 if self.isDatasetTypeIncluded("raw"):
322 self.makeSubtask("raws", butler=butler3)
323 self.makeSubtask("defineVisits", butler=butler3)
324 else:
325 self.raws = None
326 self.defineVisits = None
327 self.instrument = instrument
328 self._configuredSkyMapsBySha1 = {}
329 self._configuredSkyMapsByName = {}
330 for name, config in self.config.skyMaps.items():
331 instance = config.skyMap.apply()
332 self._populateSkyMapDicts(name, instance)
333 self._usedSkyPix = set()
334 self.translatorFactory = self.instrument.makeDataIdTranslatorFactory()
335 self.translatorFactory.log = self.log.getChild("translators")
337 def _reduce_kwargs(self):
338 # Add extra parameters to pickle
339 return dict(**super()._reduce_kwargs(), butler3=self.butler3, instrument=self.instrument)
341 def _populateSkyMapDicts(self, name, instance):
342 struct = ConfiguredSkyMap(name=name, sha1=instance.getSha1(), instance=instance)
343 self._configuredSkyMapsBySha1[struct.sha1] = struct
344 self._configuredSkyMapsByName[struct.name] = struct
346 def isDatasetTypeIncluded(self, datasetTypeName: str):
347 """Return `True` if configuration indicates that the given dataset type
348 should be converted.
350 This method is intended to be called primarily by the
351 `RepoConverter` instances used interally by the task.
353 Parameters
354 ----------
355 datasetTypeName: str
356 Name of the dataset type.
358 Returns
359 -------
360 included : `bool`
361 Whether the dataset should be included in the conversion.
362 """
363 return (
364 any(fnmatch.fnmatchcase(datasetTypeName, pattern)
365 for pattern in self.config.datasetIncludePatterns)
366 and not any(fnmatch.fnmatchcase(datasetTypeName, pattern)
367 for pattern in self.config.datasetIgnorePatterns)
368 )
370 def useSkyMap(self, skyMap: BaseSkyMap, skyMapName: str) -> str:
371 """Indicate that a repository uses the given SkyMap.
373 This method is intended to be called primarily by the
374 `RepoConverter` instances used interally by the task.
376 Parameters
377 ----------
378 skyMap : `lsst.skymap.BaseSkyMap`
379 SkyMap instance being used, typically retrieved from a Gen2
380 data repository.
381 skyMapName : `str`
382 The name of the gen2 skymap, for error reporting.
384 Returns
385 -------
386 name : `str`
387 The name of the skymap in Gen3 data IDs.
389 Raises
390 ------
391 LookupError
392 Raised if the specified skymap cannot be found.
393 """
394 sha1 = skyMap.getSha1()
395 if sha1 not in self._configuredSkyMapsBySha1:
396 self._populateSkyMapDicts(skyMapName, skyMap)
397 try:
398 struct = self._configuredSkyMapsBySha1[sha1]
399 except KeyError as err:
400 msg = f"SkyMap '{skyMapName}' with sha1={sha1} not included in configuration."
401 raise LookupError(msg) from err
402 struct.used = True
403 return struct.name
405 def registerUsedSkyMaps(self, subset: Optional[ConversionSubset]):
406 """Register all skymaps that have been marked as used.
408 This method is intended to be called primarily by the
409 `RepoConverter` instances used interally by the task.
411 Parameters
412 ----------
413 subset : `ConversionSubset`, optional
414 Object that will be used to filter converted datasets by data ID.
415 If given, it will be updated with the tracts of this skymap that
416 overlap the visits in the subset.
417 """
418 for struct in self._configuredSkyMapsBySha1.values():
419 if struct.used:
420 struct.instance.register(struct.name, self.registry)
421 if subset is not None and self.config.relatedOnly:
422 subset.addSkyMap(self.registry, struct.name)
424 def useSkyPix(self, dimension: SkyPixDimension):
425 """Indicate that a repository uses the given SkyPix dimension.
427 This method is intended to be called primarily by the
428 `RepoConverter` instances used interally by the task.
430 Parameters
431 ----------
432 dimension : `lsst.daf.butler.SkyPixDimension`
433 Dimension represening a pixelization of the sky.
434 """
435 self._usedSkyPix.add(dimension)
437 def registerUsedSkyPix(self, subset: Optional[ConversionSubset]):
438 """Register all skymaps that have been marked as used.
440 This method is intended to be called primarily by the
441 `RepoConverter` instances used interally by the task.
443 Parameters
444 ----------
445 subset : `ConversionSubset`, optional
446 Object that will be used to filter converted datasets by data ID.
447 If given, it will be updated with the pixelization IDs that
448 overlap the visits in the subset.
449 """
450 if subset is not None and self.config.relatedOnly:
451 for dimension in self._usedSkyPix:
452 subset.addSkyPix(self.registry, dimension)
454 def run(self, root: str, *,
455 calibs: Dict[str, str] = None,
456 reruns: List[Rerun],
457 visits: Optional[Iterable[int]] = None,
458 pool: Optional[Pool] = None,
459 processes: int = 1):
460 """Convert a group of related data repositories.
462 Parameters
463 ----------
464 root : `str`
465 Complete path to the root Gen2 data repository. This should be
466 a data repository that includes a Gen2 registry and any raw files
467 and/or reference catalogs.
468 calibs : `dict`
469 Dictionary mapping calibration repository path to the
470 `~lsst.daf.butler.CollectionType.CALIBRATION` collection that
471 converted datasets within it should be certified into.
472 reruns : `list` of `Rerun`
473 Specifications for rerun (processing output) collections to
474 convert.
475 visits : iterable of `int`, optional
476 The integer IDs of visits to convert. If not provided, all visits
477 in the Gen2 root repository will be converted.
478 pool : `multiprocessing.Pool`, optional
479 If not `None`, a process pool with which to parallelize some
480 operations.
481 processes : `int`, optional
482 The number of processes to use for conversion.
483 """
484 if pool is None and processes > 1:
485 pool = Pool(processes)
486 if calibs is None:
487 calibs = {}
488 if visits is not None:
489 subset = ConversionSubset(instrument=self.instrument.getName(), visits=frozenset(visits))
490 else:
491 if self.config.relatedOnly:
492 self.log.warn("config.relatedOnly is True but all visits are being ingested; "
493 "no filtering will be done.")
494 subset = None
496 # Make converters for all Gen2 repos.
497 converters = []
498 rootConverter = RootRepoConverter(task=self, root=root, subset=subset, instrument=self.instrument)
499 converters.append(rootConverter)
500 for calibRoot, collection in calibs.items():
501 if not os.path.isabs(calibRoot):
502 calibRoot = os.path.join(rootConverter.root, calibRoot)
503 converter = CalibRepoConverter(task=self, root=calibRoot, collection=collection,
504 instrument=self.instrument,
505 mapper=rootConverter.mapper,
506 subset=rootConverter.subset)
507 converters.append(converter)
508 rerunConverters = {}
509 for spec in reruns:
510 runRoot = spec.path
511 if not os.path.isabs(runRoot):
512 runRoot = os.path.join(rootConverter.root, runRoot)
513 converter = StandardRepoConverter(task=self, root=runRoot, run=spec.runName,
514 instrument=self.instrument, subset=rootConverter.subset)
515 converters.append(converter)
516 rerunConverters[spec.runName] = converter
518 # Register the instrument if we're configured to do so.
519 if self.config.doRegisterInstrument:
520 self.instrument.register(self.registry)
522 # Run raw ingest (does nothing if we weren't configured to convert the
523 # 'raw' dataset type).
524 rootConverter.runRawIngest(pool=pool)
526 # Write curated calibrations to all calibration runs and
527 # also in the default collection.
528 # Add new collections to the list of collections the butler was
529 # initialized to pass to DefineVisitsTask, to deal with the (likely)
530 # case the only 'camera' dataset in the repo will be one we're adding
531 # here.
532 if self.config.doWriteCuratedCalibrations:
533 butler3 = Butler3(butler=self.butler3)
534 # Write curated calibrations to any new calibration collections we
535 # created by converting a Gen2 calibration repo.
536 calibCollections = set()
537 for collection in calibs.values():
538 self.instrument.writeCuratedCalibrations(butler3, collection=collection)
539 calibCollections.add(collection)
540 # Ensure that we have the curated calibrations even if there
541 # is no calibration conversion. It's possible that the default
542 # calib collection will have been specified (in fact the
543 # butler convert script enforces that behavior for now) so
544 # we check for the default situation
545 # Assume we know the default rather than letting
546 # writeCuratedCalibrations default itself
547 defaultCalibCollection = self.instrument.makeCollectionName("calib")
548 if defaultCalibCollection not in calibCollections:
549 self.instrument.writeCuratedCalibrations(butler3, collection=defaultCalibCollection)
551 # Define visits (also does nothing if we weren't configurd to convert
552 # the 'raw' dataset type).
553 rootConverter.runDefineVisits(pool=pool)
555 # Walk Gen2 repos to find datasets convert.
556 for converter in converters:
557 converter.prep()
559 # Insert dimensions that are potentially shared by all Gen2
560 # repositories (and are hence managed directly by the Task, rather
561 # than a converter instance).
562 # This also finishes setting up the (shared) converter.subsets object
563 # that is used to filter data IDs for config.relatedOnly.
564 self.registerUsedSkyMaps(rootConverter.subset)
565 self.registerUsedSkyPix(rootConverter.subset)
567 # Look for datasets, generally by scanning the filesystem.
568 # This requires dimensions to have already been inserted so we can use
569 # dimension information to identify related datasets.
570 for converter in converters:
571 converter.findDatasets()
573 # Expand data IDs.
574 for converter in converters:
575 converter.expandDataIds()
577 # Actually ingest datasets.
578 for converter in converters:
579 converter.ingest()
581 # Perform any post-ingest processing.
582 for converter in converters:
583 converter.finish()
585 # Add chained collections for reruns.
586 for spec in reruns:
587 if spec.chainName is not None:
588 self.butler3.registry.registerCollection(spec.chainName, type=CollectionType.CHAINED)
589 chain = [spec.runName]
590 chain.extend(rerunConverters[spec.runName].getCollectionChain())
591 for parent in spec.parents:
592 chain.append(parent)
593 parentConverter = rerunConverters.get(parent)
594 if parentConverter is not None:
595 chain.extend(parentConverter.getCollectionChain())
596 chain.extend(rootConverter.getCollectionChain())
597 self.log.info("Defining %s from chain %s.", spec.chainName, chain)
598 self.butler3.registry.setCollectionChain(spec.chainName, chain)