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 pipe_base. 

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 

22"""Module defining connection classes for PipelineTask. 

23""" 

24 

25__all__ = ["PipelineTaskConnections", "InputQuantizedConnection", "OutputQuantizedConnection", 

26 "DeferredDatasetRef", "iterConnections"] 

27 

28from collections import UserDict, namedtuple 

29from types import SimpleNamespace 

30import typing 

31 

32import itertools 

33import string 

34 

35from . import config as configMod 

36from .connectionTypes import (InitInput, InitOutput, Input, PrerequisiteInput, 

37 Output, BaseConnection) 

38from lsst.daf.butler import DatasetRef, DatasetType, NamedKeyDict, Quantum 

39 

40if typing.TYPE_CHECKING: 40 ↛ 41line 40 didn't jump to line 41, because the condition on line 40 was never true

41 from .config import PipelineTaskConfig 

42 

43 

44class ScalarError(TypeError): 

45 """Exception raised when dataset type is configured as scalar 

46 but there are multiple data IDs in a Quantum for that dataset. 

47 """ 

48 

49 

50class PipelineTaskConnectionDict(UserDict): 

51 """This is a special dict class used by PipelineTaskConnectionMetaclass 

52 

53 This dict is used in PipelineTaskConnection class creation, as the 

54 dictionary that is initially used as __dict__. It exists to 

55 intercept connection fields declared in a PipelineTaskConnection, and 

56 what name is used to identify them. The names are then added to class 

57 level list according to the connection type of the class attribute. The 

58 names are also used as keys in a class level dictionary associated with 

59 the corresponding class attribute. This information is a duplicate of 

60 what exists in __dict__, but provides a simple place to lookup and 

61 iterate on only these variables. 

62 """ 

63 def __init__(self, *args, **kwargs): 

64 super().__init__(*args, **kwargs) 

65 # Initialize class level variables used to track any declared 

66 # class level variables that are instances of 

67 # connectionTypes.BaseConnection 

68 self.data['inputs'] = [] 

69 self.data['prerequisiteInputs'] = [] 

70 self.data['outputs'] = [] 

71 self.data['initInputs'] = [] 

72 self.data['initOutputs'] = [] 

73 self.data['allConnections'] = {} 

74 

75 def __setitem__(self, name, value): 

76 if isinstance(value, Input): 

77 self.data['inputs'].append(name) 

78 elif isinstance(value, PrerequisiteInput): 

79 self.data['prerequisiteInputs'].append(name) 

80 elif isinstance(value, Output): 

81 self.data['outputs'].append(name) 

82 elif isinstance(value, InitInput): 82 ↛ 83line 82 didn't jump to line 83, because the condition on line 82 was never true

83 self.data['initInputs'].append(name) 

84 elif isinstance(value, InitOutput): 84 ↛ 85line 84 didn't jump to line 85, because the condition on line 84 was never true

85 self.data['initOutputs'].append(name) 

86 # This should not be an elif, as it needs tested for 

87 # everything that inherits from BaseConnection 

88 if isinstance(value, BaseConnection): 

89 object.__setattr__(value, 'varName', name) 

90 self.data['allConnections'][name] = value 

91 # defer to the default behavior 

92 super().__setitem__(name, value) 

93 

94 

95class PipelineTaskConnectionsMetaclass(type): 

96 """Metaclass used in the declaration of PipelineTaskConnections classes 

97 """ 

98 def __prepare__(name, bases, **kwargs): # noqa: 805 

99 # Create an instance of our special dict to catch and track all 

100 # variables that are instances of connectionTypes.BaseConnection 

101 # Copy any existing connections from a parent class 

102 dct = PipelineTaskConnectionDict() 

103 for base in bases: 

104 if isinstance(base, PipelineTaskConnectionsMetaclass): 104 ↛ 103line 104 didn't jump to line 103, because the condition on line 104 was never false

105 for name, value in base.allConnections.items(): 105 ↛ 106line 105 didn't jump to line 106, because the loop on line 105 never started

106 dct[name] = value 

107 return dct 

108 

109 def __new__(cls, name, bases, dct, **kwargs): 

110 dimensionsValueError = TypeError("PipelineTaskConnections class must be created with a dimensions " 

111 "attribute which is an iterable of dimension names") 

112 

113 if name != 'PipelineTaskConnections': 

114 # Verify that dimensions are passed as a keyword in class 

115 # declaration 

116 if 'dimensions' not in kwargs: 116 ↛ 117line 116 didn't jump to line 117, because the condition on line 116 was never true

117 for base in bases: 

118 if hasattr(base, 'dimensions'): 

119 kwargs['dimensions'] = base.dimensions 

120 break 

121 if 'dimensions' not in kwargs: 

122 raise dimensionsValueError 

123 try: 

124 dct['dimensions'] = set(kwargs['dimensions']) 

125 except TypeError as exc: 

126 raise dimensionsValueError from exc 

127 # Lookup any python string templates that may have been used in the 

128 # declaration of the name field of a class connection attribute 

129 allTemplates = set() 

130 stringFormatter = string.Formatter() 

131 # Loop over all connections 

132 for obj in dct['allConnections'].values(): 

133 nameValue = obj.name 

134 # add all the parameters to the set of templates 

135 for param in stringFormatter.parse(nameValue): 

136 if param[1] is not None: 136 ↛ 137line 136 didn't jump to line 137, because the condition on line 136 was never true

137 allTemplates.add(param[1]) 

138 

139 # look up any template from base classes and merge them all 

140 # together 

141 mergeDict = {} 

142 for base in bases[::-1]: 

143 if hasattr(base, 'defaultTemplates'): 143 ↛ 144line 143 didn't jump to line 144, because the condition on line 143 was never true

144 mergeDict.update(base.defaultTemplates) 

145 if 'defaultTemplates' in kwargs: 145 ↛ 146line 145 didn't jump to line 146, because the condition on line 145 was never true

146 mergeDict.update(kwargs['defaultTemplates']) 

147 

148 if len(mergeDict) > 0: 148 ↛ 149line 148 didn't jump to line 149, because the condition on line 148 was never true

149 kwargs['defaultTemplates'] = mergeDict 

150 

151 # Verify that if templated strings were used, defaults were 

152 # supplied as an argument in the declaration of the connection 

153 # class 

154 if len(allTemplates) > 0 and 'defaultTemplates' not in kwargs: 154 ↛ 155line 154 didn't jump to line 155, because the condition on line 154 was never true

155 raise TypeError("PipelineTaskConnection class contains templated attribute names, but no " 

156 "defaut templates were provided, add a dictionary attribute named " 

157 "defaultTemplates which contains the mapping between template key and value") 

158 if len(allTemplates) > 0: 158 ↛ 160line 158 didn't jump to line 160, because the condition on line 158 was never true

159 # Verify all templates have a default, and throw if they do not 

160 defaultTemplateKeys = set(kwargs['defaultTemplates'].keys()) 

161 templateDifference = allTemplates.difference(defaultTemplateKeys) 

162 if templateDifference: 

163 raise TypeError(f"Default template keys were not provided for {templateDifference}") 

164 # Verify that templates do not share names with variable names 

165 # used for a connection, this is needed because of how 

166 # templates are specified in an associated config class. 

167 nameTemplateIntersection = allTemplates.intersection(set(dct['allConnections'].keys())) 

168 if len(nameTemplateIntersection) > 0: 

169 raise TypeError(f"Template parameters cannot share names with Class attributes" 

170 f" (conflicts are {nameTemplateIntersection}).") 

171 dct['defaultTemplates'] = kwargs.get('defaultTemplates', {}) 

172 

173 # Convert all the connection containers into frozensets so they cannot 

174 # be modified at the class scope 

175 for connectionName in ("inputs", "prerequisiteInputs", "outputs", "initInputs", "initOutputs"): 

176 dct[connectionName] = frozenset(dct[connectionName]) 

177 # our custom dict type must be turned into an actual dict to be used in 

178 # type.__new__ 

179 return super().__new__(cls, name, bases, dict(dct)) 

180 

181 def __init__(cls, name, bases, dct, **kwargs): 

182 # This overrides the default init to drop the kwargs argument. Python 

183 # metaclasses will have this argument set if any kwargs are passes at 

184 # class construction time, but should be consumed before calling 

185 # __init__ on the type metaclass. This is in accordance with python 

186 # documentation on metaclasses 

187 super().__init__(name, bases, dct) 

188 

189 

190class QuantizedConnection(SimpleNamespace): 

191 """A Namespace to map defined variable names of connections to their 

192 `lsst.daf.buter.DatasetRef`s 

193 

194 This class maps the names used to define a connection on a 

195 PipelineTaskConnectionsClass to the corresponding 

196 `lsst.daf.butler.DatasetRef`s provided by a `lsst.daf.butler.Quantum` 

197 instance. This will be a quantum of execution based on the graph created 

198 by examining all the connections defined on the 

199 `PipelineTaskConnectionsClass`. 

200 """ 

201 def __init__(self, **kwargs): 

202 # Create a variable to track what attributes are added. This is used 

203 # later when iterating over this QuantizedConnection instance 

204 object.__setattr__(self, "_attributes", set()) 

205 

206 def __setattr__(self, name: str, value: typing.Union[DatasetRef, typing.List[DatasetRef]]): 

207 # Capture the attribute name as it is added to this object 

208 self._attributes.add(name) 

209 super().__setattr__(name, value) 

210 

211 def __delattr__(self, name): 

212 object.__delattr__(self, name) 

213 self._attributes.remove(name) 

214 

215 def __iter__(self) -> typing.Generator[typing.Tuple[str, typing.Union[DatasetRef, 

216 typing.List[DatasetRef]]], None, None]: 

217 """Make an Iterator for this QuantizedConnection 

218 

219 Iterating over a QuantizedConnection will yield a tuple with the name 

220 of an attribute and the value associated with that name. This is 

221 similar to dict.items() but is on the namespace attributes rather than 

222 dict keys. 

223 """ 

224 yield from ((name, getattr(self, name)) for name in self._attributes) 

225 

226 def keys(self) -> typing.Generator[str, None, None]: 

227 """Returns an iterator over all the attributes added to a 

228 QuantizedConnection class 

229 """ 

230 yield from self._attributes 

231 

232 

233class InputQuantizedConnection(QuantizedConnection): 

234 pass 

235 

236 

237class OutputQuantizedConnection(QuantizedConnection): 

238 pass 

239 

240 

241class DeferredDatasetRef(namedtuple("DeferredDatasetRefBase", "datasetRef")): 

242 """Class which denotes that a datasetRef should be treated as deferred when 

243 interacting with the butler 

244 

245 Parameters 

246 ---------- 

247 datasetRef : `lsst.daf.butler.DatasetRef` 

248 The `lsst.daf.butler.DatasetRef` that will be eventually used to 

249 resolve a dataset 

250 """ 

251 __slots__ = () 

252 

253 

254class PipelineTaskConnections(metaclass=PipelineTaskConnectionsMetaclass): 

255 """PipelineTaskConnections is a class used to declare desired IO when a 

256 PipelineTask is run by an activator 

257 

258 Parameters 

259 ---------- 

260 config : `PipelineTaskConfig` 

261 A `PipelineTaskConfig` class instance whose class has been configured 

262 to use this `PipelineTaskConnectionsClass` 

263 

264 Notes 

265 ----- 

266 ``PipelineTaskConnection`` classes are created by declaring class 

267 attributes of types defined in `lsst.pipe.base.connectionTypes` and are 

268 listed as follows: 

269 

270 * ``InitInput`` - Defines connections in a quantum graph which are used as 

271 inputs to the ``__init__`` function of the `PipelineTask` corresponding 

272 to this class 

273 * ``InitOuput`` - Defines connections in a quantum graph which are to be 

274 persisted using a butler at the end of the ``__init__`` function of the 

275 `PipelineTask` corresponding to this class. The variable name used to 

276 define this connection should be the same as an attribute name on the 

277 `PipelineTask` instance. E.g. if an ``InitOutput`` is declared with 

278 the name ``outputSchema`` in a ``PipelineTaskConnections`` class, then 

279 a `PipelineTask` instance should have an attribute 

280 ``self.outputSchema`` defined. Its value is what will be saved by the 

281 activator framework. 

282 * ``PrerequisiteInput`` - An input connection type that defines a 

283 `lsst.daf.butler.DatasetType` that must be present at execution time, 

284 but that will not be used during the course of creating the quantum 

285 graph to be executed. These most often are things produced outside the 

286 processing pipeline, such as reference catalogs. 

287 * ``Input`` - Input `lsst.daf.butler.DatasetType` objects that will be used 

288 in the ``run`` method of a `PipelineTask`. The name used to declare 

289 class attribute must match a function argument name in the ``run`` 

290 method of a `PipelineTask`. E.g. If the ``PipelineTaskConnections`` 

291 defines an ``Input`` with the name ``calexp``, then the corresponding 

292 signature should be ``PipelineTask.run(calexp, ...)`` 

293 * ``Output`` - A `lsst.daf.butler.DatasetType` that will be produced by an 

294 execution of a `PipelineTask`. The name used to declare the connection 

295 must correspond to an attribute of a `Struct` that is returned by a 

296 `PipelineTask` ``run`` method. E.g. if an output connection is 

297 defined with the name ``measCat``, then the corresponding 

298 ``PipelineTask.run`` method must return ``Struct(measCat=X,..)`` where 

299 X matches the ``storageClass`` type defined on the output connection. 

300 

301 The process of declaring a ``PipelineTaskConnection`` class involves 

302 parameters passed in the declaration statement. 

303 

304 The first parameter is ``dimensions`` which is an iterable of strings which 

305 defines the unit of processing the run method of a corresponding 

306 `PipelineTask` will operate on. These dimensions must match dimensions that 

307 exist in the butler registry which will be used in executing the 

308 corresponding `PipelineTask`. 

309 

310 The second parameter is labeled ``defaultTemplates`` and is conditionally 

311 optional. The name attributes of connections can be specified as python 

312 format strings, with named format arguments. If any of the name parameters 

313 on connections defined in a `PipelineTaskConnections` class contain a 

314 template, then a default template value must be specified in the 

315 ``defaultTemplates`` argument. This is done by passing a dictionary with 

316 keys corresponding to a template identifier, and values corresponding to 

317 the value to use as a default when formatting the string. For example if 

318 ``ConnectionClass.calexp.name = '{input}Coadd_calexp'`` then 

319 ``defaultTemplates`` = {'input': 'deep'}. 

320 

321 Once a `PipelineTaskConnections` class is created, it is used in the 

322 creation of a `PipelineTaskConfig`. This is further documented in the 

323 documentation of `PipelineTaskConfig`. For the purposes of this 

324 documentation, the relevant information is that the config class allows 

325 configuration of connection names by users when running a pipeline. 

326 

327 Instances of a `PipelineTaskConnections` class are used by the pipeline 

328 task execution framework to introspect what a corresponding `PipelineTask` 

329 will require, and what it will produce. 

330 

331 Examples 

332 -------- 

333 >>> from lsst.pipe.base import connectionTypes as cT 

334 >>> from lsst.pipe.base import PipelineTaskConnections 

335 >>> from lsst.pipe.base import PipelineTaskConfig 

336 >>> class ExampleConnections(PipelineTaskConnections, 

337 ... dimensions=("A", "B"), 

338 ... defaultTemplates={"foo": "Example"}): 

339 ... inputConnection = cT.Input(doc="Example input", 

340 ... dimensions=("A", "B"), 

341 ... storageClass=Exposure, 

342 ... name="{foo}Dataset") 

343 ... outputConnection = cT.Output(doc="Example output", 

344 ... dimensions=("A", "B"), 

345 ... storageClass=Exposure, 

346 ... name="{foo}output") 

347 >>> class ExampleConfig(PipelineTaskConfig, 

348 ... pipelineConnections=ExampleConnections): 

349 ... pass 

350 >>> config = ExampleConfig() 

351 >>> config.connections.foo = Modified 

352 >>> config.connections.outputConnection = "TotallyDifferent" 

353 >>> connections = ExampleConnections(config=config) 

354 >>> assert(connections.inputConnection.name == "ModifiedDataset") 

355 >>> assert(connections.outputConnection.name == "TotallyDifferent") 

356 """ 

357 

358 def __init__(self, *, config: 'PipelineTaskConfig' = None): 

359 self.inputs = set(self.inputs) 

360 self.prerequisiteInputs = set(self.prerequisiteInputs) 

361 self.outputs = set(self.outputs) 

362 self.initInputs = set(self.initInputs) 

363 self.initOutputs = set(self.initOutputs) 

364 

365 if config is None or not isinstance(config, configMod.PipelineTaskConfig): 

366 raise ValueError("PipelineTaskConnections must be instantiated with" 

367 " a PipelineTaskConfig instance") 

368 self.config = config 

369 # Extract the template names that were defined in the config instance 

370 # by looping over the keys of the defaultTemplates dict specified at 

371 # class declaration time 

372 templateValues = {name: getattr(config.connections, name) for name in getattr(self, 

373 'defaultTemplates').keys()} 

374 # Extract the configured value corresponding to each connection 

375 # variable. I.e. for each connection identifier, populate a override 

376 # for the connection.name attribute 

377 self._nameOverrides = {name: getattr(config.connections, name).format(**templateValues) 

378 for name in self.allConnections.keys()} 

379 

380 # connections.name corresponds to a dataset type name, create a reverse 

381 # mapping that goes from dataset type name to attribute identifier name 

382 # (variable name) on the connection class 

383 self._typeNameToVarName = {v: k for k, v in self._nameOverrides.items()} 

384 

385 def buildDatasetRefs(self, quantum: Quantum) -> typing.Tuple[InputQuantizedConnection, 

386 OutputQuantizedConnection]: 

387 """Builds QuantizedConnections corresponding to input Quantum 

388 

389 Parameters 

390 ---------- 

391 quantum : `lsst.daf.butler.Quantum` 

392 Quantum object which defines the inputs and outputs for a given 

393 unit of processing 

394 

395 Returns 

396 ------- 

397 retVal : `tuple` of (`InputQuantizedConnection`, 

398 `OutputQuantizedConnection`) Namespaces mapping attribute names 

399 (identifiers of connections) to butler references defined in the 

400 input `lsst.daf.butler.Quantum` 

401 """ 

402 inputDatasetRefs = InputQuantizedConnection() 

403 outputDatasetRefs = OutputQuantizedConnection() 

404 # operate on a reference object and an interable of names of class 

405 # connection attributes 

406 for refs, names in zip((inputDatasetRefs, outputDatasetRefs), 

407 (itertools.chain(self.inputs, self.prerequisiteInputs), self.outputs)): 

408 # get a name of a class connection attribute 

409 for attributeName in names: 

410 # get the attribute identified by name 

411 attribute = getattr(self, attributeName) 

412 # Branch if the attribute dataset type is an input 

413 if attribute.name in quantum.predictedInputs: 

414 # Get the DatasetRefs 

415 quantumInputRefs = quantum.predictedInputs[attribute.name] 

416 # if the dataset is marked to load deferred, wrap it in a 

417 # DeferredDatasetRef 

418 if attribute.deferLoad: 

419 quantumInputRefs = [DeferredDatasetRef(datasetRef=ref) for ref in quantumInputRefs] 

420 # Unpack arguments that are not marked multiples (list of 

421 # length one) 

422 if not attribute.multiple: 

423 if len(quantumInputRefs) > 1: 

424 raise ScalarError( 

425 f"Received multiple datasets " 

426 f"{', '.join(str(r.dataId) for r in quantumInputRefs)} " 

427 f"for scalar connection {attributeName} " 

428 f"({quantumInputRefs[0].datasetType.name}) " 

429 f"of quantum for {quantum.taskName} with data ID {quantum.dataId}." 

430 ) 

431 if len(quantumInputRefs) == 0: 

432 continue 

433 quantumInputRefs = quantumInputRefs[0] 

434 # Add to the QuantizedConnection identifier 

435 setattr(refs, attributeName, quantumInputRefs) 

436 # Branch if the attribute dataset type is an output 

437 elif attribute.name in quantum.outputs: 

438 value = quantum.outputs[attribute.name] 

439 # Unpack arguments that are not marked multiples (list of 

440 # length one) 

441 if not attribute.multiple: 

442 value = value[0] 

443 # Add to the QuantizedConnection identifier 

444 setattr(refs, attributeName, value) 

445 # Specified attribute is not in inputs or outputs dont know how 

446 # to handle, throw 

447 else: 

448 raise ValueError(f"Attribute with name {attributeName} has no counterpoint " 

449 "in input quantum") 

450 return inputDatasetRefs, outputDatasetRefs 

451 

452 def adjustQuantum(self, datasetRefMap: NamedKeyDict[DatasetType, typing.Set[DatasetRef]] 

453 ) -> NamedKeyDict[DatasetType, typing.Set[DatasetRef]]: 

454 """Override to make adjustments to `lsst.daf.butler.DatasetRef` objects 

455 in the `lsst.daf.butler.core.Quantum` during the graph generation stage 

456 of the activator. 

457 

458 The base class implementation simply checks that input connections with 

459 ``multiple`` set to `False` have no more than one dataset. 

460 

461 Parameters 

462 ---------- 

463 datasetRefMap : `NamedKeyDict` 

464 Mapping from dataset type to a `set` of 

465 `lsst.daf.butler.DatasetRef` objects 

466 

467 Returns 

468 ------- 

469 datasetRefMap : `NamedKeyDict` 

470 Modified mapping of input with possibly adjusted 

471 `lsst.daf.butler.DatasetRef` objects. 

472 

473 Raises 

474 ------ 

475 ScalarError 

476 Raised if any `Input` or `PrerequisiteInput` connection has 

477 ``multiple`` set to `False`, but multiple datasets. 

478 Exception 

479 Overrides of this function have the option of raising an Exception 

480 if a field in the input does not satisfy a need for a corresponding 

481 pipelineTask, i.e. no reference catalogs are found. 

482 """ 

483 for connection in itertools.chain(iterConnections(self, "inputs"), 

484 iterConnections(self, "prerequisiteInputs")): 

485 refs = datasetRefMap[connection.name] 

486 if not connection.multiple and len(refs) > 1: 

487 raise ScalarError( 

488 f"Found multiple datasets {', '.join(str(r.dataId) for r in refs)} " 

489 f"for scalar connection {connection.name} ({refs[0].datasetType.name})." 

490 ) 

491 return datasetRefMap 

492 

493 

494def iterConnections(connections: PipelineTaskConnections, connectionType: str) -> typing.Generator: 

495 """Creates an iterator over the selected connections type which yields 

496 all the defined connections of that type. 

497 

498 Parameters 

499 ---------- 

500 connections: `PipelineTaskConnections` 

501 An instance of a `PipelineTaskConnections` object that will be iterated 

502 over. 

503 connectionType: `str` 

504 The type of connections to iterate over, valid values are inputs, 

505 outputs, prerequisiteInputs, initInputs, initOutputs. 

506 

507 Yields 

508 ------- 

509 connection: `BaseConnection` 

510 A connection defined on the input connections object of the type 

511 supplied. The yielded value Will be an derived type of 

512 `BaseConnection`. 

513 """ 

514 for name in getattr(connections, connectionType): 

515 yield getattr(connections, name)