Coverage for python/lsst/obs/base/gen2to3/convertRepo.py : 32%

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 typing import Iterable, Optional, List, Dict
30from lsst.utils import doImport
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
48@dataclass
49class ConfiguredSkyMap:
50 """Struct containing information about a skymap that may appear in a Gen2
51 repository.
52 """
54 name: str
55 """Name of the skymap used in Gen3 data IDs.
56 """
58 sha1: bytes
59 """Hash computed by `BaseSkyMap.getSha1`.
60 """
62 instance: BaseSkyMap
63 """Name of the skymap used in Gen3 data IDs.
64 """
66 used: bool = False
67 """Whether this skymap has been found in at least one repository being
68 converted.
69 """
72@dataclass
73class Rerun:
74 """Specification for a Gen2 processing-output repository to convert.
75 """
77 path: str
78 """Absolute or relative (to the root repository) path to the Gen2
79 repository (`str`).
80 """
82 runName: str
83 """Name of the `~lsst.daf.butler.CollectionType.RUN` collection datasets
84 will be inserted into (`str`).
85 """
87 chainName: Optional[str]
88 """Name of a `~lsst.daf.butler.CollectionType.CHAINED` collection that will
89 combine this repository's datasets with those of its parent repositories
90 (`str`, optional).
91 """
93 parents: List[str]
94 """Collection names associated with parent repositories, used to define the
95 chained collection (`list` [ `str` ]).
97 Ignored if `chainName` is `None`. Runs used in the root repo are
98 automatically included.
99 """
102class ConvertRepoSkyMapConfig(Config):
103 """Sub-config used to hold the parameters of a SkyMap.
105 Notes
106 -----
107 This config only needs to exist because we can't put a
108 `~lsst.pex.config.RegistryField` directly inside a
109 `~lsst.pex.config.ConfigDictField`.
111 It needs to have its only field named "skyMap" for compatibility with the
112 configuration of `lsst.pipe.tasks.MakeSkyMapTask`, which we want so we can
113 use one config file in an obs package to configure both.
115 This name leads to unfortunate repetition with the field named
116 "skymap" that holds it - "skyMap[name].skyMap" - but that seems
117 unavoidable.
118 """
119 skyMap = skyMapRegistry.makeField(
120 doc="Type and parameters for the SkyMap itself.",
121 default="dodeca",
122 )
125class ConvertRepoConfig(Config):
126 raws = ConfigurableField(
127 "Configuration for subtask responsible for ingesting raws and adding "
128 "exposure dimension entries.",
129 target=RawIngestTask,
130 )
131 defineVisits = ConfigurableField(
132 "Configuration for the subtask responsible for defining visits from "
133 "exposures.",
134 target=DefineVisitsTask,
135 )
136 skyMaps = ConfigDictField(
137 "Mapping from Gen3 skymap name to the parameters used to construct a "
138 "BaseSkyMap instance. This will be used to associate names with "
139 "existing skymaps found in the Gen2 repo.",
140 keytype=str,
141 itemtype=ConvertRepoSkyMapConfig,
142 default={}
143 )
144 rootSkyMapName = Field(
145 "Name of a Gen3 skymap (an entry in ``self.skyMaps``) to assume for "
146 "datasets in the root repository when no SkyMap is found there. ",
147 dtype=str,
148 optional=True,
149 default=None,
150 )
151 runs = DictField(
152 "A mapping from dataset type name to the RUN collection they should "
153 "be inserted into. This must include all datasets that can be found "
154 "in the root repository; other repositories will use per-repository "
155 "runs.",
156 keytype=str,
157 itemtype=str,
158 default={
159 "deepCoadd_skyMap": "skymaps",
160 "brightObjectMask": "masks",
161 }
162 )
163 storageClasses = DictField(
164 "Mapping from dataset type name or Gen2 policy entry (e.g. 'python' "
165 "or 'persistable') to the Gen3 StorageClass name.",
166 keytype=str,
167 itemtype=str,
168 default={
169 "bias": "ExposureF",
170 "dark": "ExposureF",
171 "flat": "ExposureF",
172 "defects": "Defects",
173 "crosstalk": "CrosstalkCalib",
174 "BaseSkyMap": "SkyMap",
175 "BaseCatalog": "Catalog",
176 "BackgroundList": "Background",
177 "raw": "Exposure",
178 "MultilevelParquetTable": "DataFrame",
179 "ParquetTable": "DataFrame",
180 "SkyWcs": "Wcs",
181 }
182 )
183 formatterClasses = DictField(
184 "Mapping from dataset type name to formatter class. "
185 "By default these are derived from the formatters listed in the"
186 " Gen3 datastore configuration.",
187 keytype=str,
188 itemtype=str,
189 default={}
190 )
191 targetHandlerClasses = DictField(
192 "Mapping from dataset type name to target handler class.",
193 keytype=str,
194 itemtype=str,
195 default={}
196 )
197 doRegisterInstrument = Field(
198 "If True (default), add dimension records for the Instrument and its "
199 "filters and detectors to the registry instead of assuming they are "
200 "already present.",
201 dtype=bool,
202 default=True,
203 )
204 doWriteCuratedCalibrations = Field(
205 "If True (default), ingest human-curated calibrations directly via "
206 "the Instrument interface. Note that these calibrations are never "
207 "converted from Gen2 repositories.",
208 dtype=bool,
209 default=True,
210 )
211 refCats = ListField(
212 "The names of reference catalogs (subdirectories under ref_cats) to "
213 "be converted",
214 dtype=str,
215 default=[]
216 )
217 fileIgnorePatterns = ListField(
218 "Filename globs that should be ignored instead of being treated as "
219 "datasets.",
220 dtype=str,
221 default=["README.txt", "*~?", "butler.yaml", "gen3.sqlite3",
222 "registry.sqlite3", "calibRegistry.sqlite3", "_mapper",
223 "_parent", "repositoryCfg.yaml"]
224 )
225 rawDatasetType = Field(
226 "Gen2 dataset type to use for raw data.",
227 dtype=str,
228 default="raw",
229 )
230 datasetIncludePatterns = ListField(
231 "Glob-style patterns for dataset type names that should be converted.",
232 dtype=str,
233 default=["*"]
234 )
235 datasetIgnorePatterns = ListField(
236 "Glob-style patterns for dataset type names that should not be "
237 "converted despite matching a pattern in datasetIncludePatterns.",
238 dtype=str,
239 default=[]
240 )
241 ccdKey = Field(
242 "Key used for the Gen2 equivalent of 'detector' in data IDs.",
243 dtype=str,
244 default="ccd",
245 )
246 relatedOnly = Field(
247 "If True (default), only convert datasets that are related to the "
248 "ingested visits. Ignored unless a list of visits is passed to "
249 "run().",
250 dtype=bool,
251 default=False,
252 )
253 curatedCalibrations = ListField(
254 "Dataset types that are handled by `Instrument.writeCuratedCalibrations()` "
255 "and thus should not be converted using the standard calibration "
256 "conversion system.",
257 dtype=str,
258 default=["camera",
259 "transmission_sensor",
260 "transmission_filter",
261 "transmission_optics",
262 "transmission_atmosphere",
263 "bfKernel"]
264 )
265 instrument = Field(
266 doc=("Fully-qualified Python name of the `Instrument` subclass for "
267 "all converted datasets."),
268 dtype=str,
269 optional=False,
270 default=None,
271 )
273 @property
274 def transfer(self):
275 return self.raws.transfer
277 @transfer.setter
278 def transfer(self, value):
279 self.raws.transfer = value
281 def setDefaults(self):
282 self.transfer = None
284 # TODO: check that there are no collection overrides for curated
285 # calibrations, since we don't have a good way to utilize them.
288class ConvertRepoTask(Task):
289 """A task that converts one or more related Gen2 data repositories to a
290 single Gen3 data repository (with multiple collections).
292 Parameters
293 ----------
294 config: `ConvertRepoConfig`
295 Configuration for this task.
296 butler3: `lsst.daf.butler.Butler`
297 A writeable Gen3 Butler instance that represents the data repository
298 that datasets will be ingested into. If the 'raw' dataset is
299 configured to be included in the conversion, ``butler3.run`` should be
300 set to the name of the collection raws should be ingested into, and
301 ``butler3.collections`` should include a calibration collection from
302 which the ``camera`` dataset can be loaded, unless a calibration repo
303 is converted and ``doWriteCuratedCalibrations`` is `True`.
304 **kwargs
305 Other keyword arguments are forwarded to the `Task` constructor.
307 Notes
308 -----
309 Most of the work of converting repositories is delegated to instances of
310 the `RepoConverter` hierarchy. The `ConvertRepoTask` instance itself holds
311 only state that is relevant for all Gen2 repositories being ingested, while
312 each `RepoConverter` instance holds only state relevant for the conversion
313 of a single Gen2 repository. Both the task and the `RepoConverter`
314 instances are single use; `ConvertRepoTask.run` and most `RepoConverter`
315 methods may only be called once on a particular instance.
316 """
318 ConfigClass = ConvertRepoConfig
320 _DefaultName = "convertRepo"
322 def __init__(self, config=None, *, butler3: Butler3, **kwargs):
323 config.validate() # Not a CmdlineTask nor PipelineTask, so have to validate the config here.
324 super().__init__(config, **kwargs)
325 self.butler3 = butler3
326 self.registry = self.butler3.registry
327 self.universe = self.registry.dimensions
328 if self.isDatasetTypeIncluded("raw"):
329 self.makeSubtask("raws", butler=butler3)
330 self.makeSubtask("defineVisits", butler=butler3)
331 else:
332 self.raws = None
333 self.defineVisits = None
334 self.instrument = doImport(self.config.instrument)()
335 self._configuredSkyMapsBySha1 = {}
336 self._configuredSkyMapsByName = {}
337 for name, config in self.config.skyMaps.items():
338 instance = config.skyMap.apply()
339 self._populateSkyMapDicts(name, instance)
340 self._usedSkyPix = set()
341 self.translatorFactory = self.instrument.makeDataIdTranslatorFactory()
342 self.translatorFactory.log = self.log.getChild("translators")
344 def _populateSkyMapDicts(self, name, instance):
345 struct = ConfiguredSkyMap(name=name, sha1=instance.getSha1(), instance=instance)
346 self._configuredSkyMapsBySha1[struct.sha1] = struct
347 self._configuredSkyMapsByName[struct.name] = struct
349 def isDatasetTypeIncluded(self, datasetTypeName: str):
350 """Return `True` if configuration indicates that the given dataset type
351 should be converted.
353 This method is intended to be called primarily by the
354 `RepoConverter` instances used interally by the task.
356 Parameters
357 ----------
358 datasetTypeName: str
359 Name of the dataset type.
361 Returns
362 -------
363 included : `bool`
364 Whether the dataset should be included in the conversion.
365 """
366 return (
367 any(fnmatch.fnmatchcase(datasetTypeName, pattern)
368 for pattern in self.config.datasetIncludePatterns)
369 and not any(fnmatch.fnmatchcase(datasetTypeName, pattern)
370 for pattern in self.config.datasetIgnorePatterns)
371 )
373 def useSkyMap(self, skyMap: BaseSkyMap, skyMapName: str) -> str:
374 """Indicate that a repository uses the given SkyMap.
376 This method is intended to be called primarily by the
377 `RepoConverter` instances used interally by the task.
379 Parameters
380 ----------
381 skyMap : `lsst.skymap.BaseSkyMap`
382 SkyMap instance being used, typically retrieved from a Gen2
383 data repository.
384 skyMapName : `str`
385 The name of the gen2 skymap, for error reporting.
387 Returns
388 -------
389 name : `str`
390 The name of the skymap in Gen3 data IDs.
392 Raises
393 ------
394 LookupError
395 Raised if the specified skymap cannot be found.
396 """
397 sha1 = skyMap.getSha1()
398 if sha1 not in self._configuredSkyMapsBySha1:
399 self._populateSkyMapDicts(skyMapName, skyMap)
400 try:
401 struct = self._configuredSkyMapsBySha1[sha1]
402 except KeyError as err:
403 msg = f"SkyMap '{skyMapName}' with sha1={sha1} not included in configuration."
404 raise LookupError(msg) from err
405 struct.used = True
406 return struct.name
408 def registerUsedSkyMaps(self, subset: Optional[ConversionSubset]):
409 """Register all skymaps that have been marked as used.
411 This method is intended to be called primarily by the
412 `RepoConverter` instances used interally by the task.
414 Parameters
415 ----------
416 subset : `ConversionSubset`, optional
417 Object that will be used to filter converted datasets by data ID.
418 If given, it will be updated with the tracts of this skymap that
419 overlap the visits in the subset.
420 """
421 for struct in self._configuredSkyMapsBySha1.values():
422 if struct.used:
423 struct.instance.register(struct.name, self.registry)
424 if subset is not None and self.config.relatedOnly:
425 subset.addSkyMap(self.registry, struct.name)
427 def useSkyPix(self, dimension: SkyPixDimension):
428 """Indicate that a repository uses the given SkyPix dimension.
430 This method is intended to be called primarily by the
431 `RepoConverter` instances used interally by the task.
433 Parameters
434 ----------
435 dimension : `lsst.daf.butler.SkyPixDimension`
436 Dimension represening a pixelization of the sky.
437 """
438 self._usedSkyPix.add(dimension)
440 def registerUsedSkyPix(self, subset: Optional[ConversionSubset]):
441 """Register all skymaps that have been marked as used.
443 This method is intended to be called primarily by the
444 `RepoConverter` instances used interally by the task.
446 Parameters
447 ----------
448 subset : `ConversionSubset`, optional
449 Object that will be used to filter converted datasets by data ID.
450 If given, it will be updated with the pixelization IDs that
451 overlap the visits in the subset.
452 """
453 if subset is not None and self.config.relatedOnly:
454 for dimension in self._usedSkyPix:
455 subset.addSkyPix(self.registry, dimension)
457 def run(self, root: str, *,
458 calibs: Dict[str, str] = None,
459 reruns: List[Rerun],
460 visits: Optional[Iterable[int]] = None):
461 """Convert a group of related data repositories.
463 Parameters
464 ----------
465 root : `str`
466 Complete path to the root Gen2 data repository. This should be
467 a data repository that includes a Gen2 registry and any raw files
468 and/or reference catalogs.
469 calibs : `dict`
470 Dictionary mapping calibration repository path to the
471 `~lsst.daf.butler.CollectionType.RUN` collection that converted
472 datasets within it should be inserted into.
473 reruns : `list` of `Rerun`
474 Specifications for rerun (processing output) collections to
475 convert.
476 visits : iterable of `int`, optional
477 The integer IDs of visits to convert. If not provided, all visits
478 in the Gen2 root repository will be converted.
479 """
480 if calibs is None:
481 calibs = {}
482 if visits is not None:
483 subset = ConversionSubset(instrument=self.instrument.getName(), visits=frozenset(visits))
484 else:
485 if self.config.relatedOnly:
486 self.log.warn("config.relatedOnly is True but all visits are being ingested; "
487 "no filtering will be done.")
488 subset = None
490 # Make converters for all Gen2 repos.
491 converters = []
492 rootConverter = RootRepoConverter(task=self, root=root, subset=subset)
493 converters.append(rootConverter)
494 for calibRoot, run in calibs.items():
495 if not os.path.isabs(calibRoot):
496 calibRoot = os.path.join(rootConverter.root, calibRoot)
497 converter = CalibRepoConverter(task=self, root=calibRoot, run=run,
498 mapper=rootConverter.mapper,
499 subset=rootConverter.subset)
500 converters.append(converter)
501 for spec in reruns:
502 runRoot = spec.path
503 if not os.path.isabs(runRoot):
504 runRoot = os.path.join(rootConverter.root, runRoot)
505 converter = StandardRepoConverter(task=self, root=runRoot, run=spec.runName,
506 subset=rootConverter.subset)
507 converters.append(converter)
509 # Register the instrument if we're configured to do so.
510 if self.config.doRegisterInstrument:
511 # Allow registration to fail on the assumption that this means
512 # we are reusing a butler
513 try:
514 self.instrument.register(self.registry)
515 except Exception:
516 pass
518 # Run raw ingest (does nothing if we weren't configured to convert the
519 # 'raw' dataset type).
520 rootConverter.runRawIngest()
522 # Write curated calibrations to all calibration repositories.
523 # Add new collections to the list of collections the butler was
524 # initialized to pass to DefineVisitsTask, to deal with the (likely)
525 # case the only 'camera' dataset in the repo will be one we're adding
526 # here.
527 if self.config.doWriteCuratedCalibrations:
528 for run in calibs.values():
529 butler3 = Butler3(butler=self.butler3, run=run)
530 self.instrument.writeCuratedCalibrations(butler3)
532 # Define visits (also does nothing if we weren't configurd to convert
533 # the 'raw' dataset type).
534 rootConverter.runDefineVisits()
536 # Walk Gen2 repos to find datasets convert.
537 for converter in converters:
538 converter.prep()
540 # Insert dimensions needed by any converters. In practice this is just
541 # calibration_labels right now, because exposures and visits (and
542 # things related to them) are handled by RawIngestTask and
543 # DefineVisitsTask earlier and skymaps are handled later.
544 #
545 # Note that we do not try to filter dimensions down to just those
546 # related to the given visits, even if config.relatedOnly is True; we
547 # need them in the Gen3 repo in order to be able to know which datasets
548 # to convert, because Gen2 alone doesn't know enough about the
549 # relationships between data IDs.
550 for converter in converters:
551 converter.insertDimensionData()
553 # Insert dimensions that are potentially shared by all Gen2
554 # repositories (and are hence managed directly by the Task, rather
555 # than a converter instance).
556 # This also finishes setting up the (shared) converter.subsets object
557 # that is used to filter data IDs for config.relatedOnly.
558 self.registerUsedSkyMaps(rootConverter.subset)
559 self.registerUsedSkyPix(rootConverter.subset)
561 # Look for datasets, generally by scanning the filesystem.
562 # This requires dimensions to have already been inserted so we can use
563 # dimension information to identify related datasets.
564 for converter in converters:
565 converter.findDatasets()
567 # Expand data IDs.
568 for converter in converters:
569 converter.expandDataIds()
571 # Actually ingest datasets.
572 for converter in converters:
573 converter.ingest()
575 # Add chained collections for reruns.
576 for spec in reruns:
577 if spec.chainName is not None:
578 self.butler3.registry.registerCollection(spec.chainName, type=CollectionType.CHAINED)
579 chain = [spec.runName]
580 chain.extend(spec.parents)
581 chain.extend(rootConverter.getCollectionChain())
582 self.log.info("Defining %s from chain %s.", spec.chainName, chain)
583 self.butler3.registry.setCollectionChain(spec.chainName, chain)