Coverage for python/lsst/ctrl/mpexec/preExecInit.py: 18%
155 statements
« prev ^ index » next coverage.py v7.3.2, created at 2023-11-04 09:50 +0000
« prev ^ index » next coverage.py v7.3.2, created at 2023-11-04 09:50 +0000
1# This file is part of ctrl_mpexec.
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/>.
28from __future__ import annotations
30__all__ = ["PreExecInit"]
32# -------------------------------
33# Imports of standard modules --
34# -------------------------------
35import abc
36import logging
37from collections.abc import Iterable, Iterator
38from contextlib import contextmanager
39from typing import TYPE_CHECKING, Any
41# -----------------------------
42# Imports for other modules --
43# -----------------------------
44from lsst.daf.butler import DatasetRef, DatasetType
45from lsst.daf.butler.registry import ConflictingDefinitionError
46from lsst.pipe.base import PipelineDatasetTypes
47from lsst.utils.packages import Packages
49if TYPE_CHECKING:
50 from lsst.daf.butler import Butler, LimitedButler
51 from lsst.pipe.base import QuantumGraph, TaskDef, TaskFactory
53_LOG = logging.getLogger(__name__)
56class MissingReferenceError(Exception):
57 """Exception raised when resolved reference is missing from graph."""
59 pass
62def _compare_packages(old_packages: Packages, new_packages: Packages) -> None:
63 """Compare two versions of Packages.
65 Parameters
66 ----------
67 old_packages : `Packages`
68 Previously recorded package versions.
69 new_packages : `Packages`
70 New set of package versions.
72 Raises
73 ------
74 TypeError
75 Raised if parameters are inconsistent.
76 """
77 diff = new_packages.difference(old_packages)
78 if diff:
79 versions_str = "; ".join(f"{pkg}: {diff[pkg][1]} vs {diff[pkg][0]}" for pkg in diff)
80 raise TypeError(f"Package versions mismatch: ({versions_str})")
81 else:
82 _LOG.debug("new packages are consistent with old")
85class PreExecInitBase(abc.ABC):
86 """Common part of the implementation of PreExecInit classes that does not
87 depend on Butler type.
88 """
90 def __init__(self, butler: LimitedButler, taskFactory: TaskFactory, extendRun: bool):
91 self.butler = butler
92 self.taskFactory = taskFactory
93 self.extendRun = extendRun
95 def initialize(
96 self,
97 graph: QuantumGraph,
98 saveInitOutputs: bool = True,
99 registerDatasetTypes: bool = False,
100 saveVersions: bool = True,
101 ) -> None:
102 """Perform all initialization steps.
104 Convenience method to execute all initialization steps. Instead of
105 calling this method and providing all options it is also possible to
106 call methods individually.
108 Parameters
109 ----------
110 graph : `~lsst.pipe.base.QuantumGraph`
111 Execution graph.
112 saveInitOutputs : `bool`, optional
113 If ``True`` (default) then save "init outputs", configurations,
114 and package versions to butler.
115 registerDatasetTypes : `bool`, optional
116 If ``True`` then register dataset types in registry, otherwise
117 they must be already registered.
118 saveVersions : `bool`, optional
119 If ``False`` then do not save package versions even if
120 ``saveInitOutputs`` is set to ``True``.
121 """
122 # register dataset types or check consistency
123 self.initializeDatasetTypes(graph, registerDatasetTypes)
125 # Save task initialization data or check that saved data
126 # is consistent with what tasks would save
127 if saveInitOutputs:
128 self.saveInitOutputs(graph)
129 self.saveConfigs(graph)
130 if saveVersions:
131 self.savePackageVersions(graph)
133 @abc.abstractmethod
134 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
135 """Save or check DatasetTypes output by the tasks in a graph.
137 Iterates over all DatasetTypes for all tasks in a graph and either
138 tries to add them to registry or compares them to existing ones.
140 Parameters
141 ----------
142 graph : `~lsst.pipe.base.QuantumGraph`
143 Execution graph.
144 registerDatasetTypes : `bool`, optional
145 If ``True`` then register dataset types in registry, otherwise
146 they must be already registered.
148 Raises
149 ------
150 ValueError
151 Raised if existing DatasetType is different from DatasetType
152 in a graph.
153 KeyError
154 Raised if ``registerDatasetTypes`` is ``False`` and DatasetType
155 does not exist in registry.
156 """
157 raise NotImplementedError()
159 def saveInitOutputs(self, graph: QuantumGraph) -> None:
160 """Write any datasets produced by initializing tasks in a graph.
162 Parameters
163 ----------
164 graph : `~lsst.pipe.base.QuantumGraph`
165 Execution graph.
167 Raises
168 ------
169 TypeError
170 Raised if the type of existing object in butler is different from
171 new data.
172 """
173 _LOG.debug("Will save InitOutputs for all tasks")
174 for taskDef in self._task_iter(graph):
175 init_input_refs = graph.initInputRefs(taskDef) or []
176 task = self.taskFactory.makeTask(taskDef, self.butler, init_input_refs)
177 for name in taskDef.connections.initOutputs:
178 attribute = getattr(taskDef.connections, name)
179 init_output_refs = graph.initOutputRefs(taskDef) or []
180 init_output_ref, obj_from_store = self._find_dataset(init_output_refs, attribute.name)
181 if init_output_ref is None:
182 raise ValueError(f"Cannot find dataset reference for init output {name} in a graph")
183 init_output_var = getattr(task, name)
185 if obj_from_store is not None:
186 _LOG.debug(
187 "Retrieving InitOutputs for task=%s key=%s dsTypeName=%s", task, name, attribute.name
188 )
189 obj_from_store = self.butler.get(init_output_ref)
190 # Types are supposed to be identical.
191 # TODO: Check that object contents is identical too.
192 if type(obj_from_store) is not type(init_output_var):
193 raise TypeError(
194 f"Stored initOutput object type {type(obj_from_store)} "
195 "is different from task-generated type "
196 f"{type(init_output_var)} for task {taskDef}"
197 )
198 else:
199 _LOG.debug("Saving InitOutputs for task=%s key=%s", taskDef.label, name)
200 # This can still raise if there is a concurrent write.
201 self.butler.put(init_output_var, init_output_ref)
203 def saveConfigs(self, graph: QuantumGraph) -> None:
204 """Write configurations for pipeline tasks to butler or check that
205 existing configurations are equal to the new ones.
207 Parameters
208 ----------
209 graph : `~lsst.pipe.base.QuantumGraph`
210 Execution graph.
212 Raises
213 ------
214 TypeError
215 Raised if existing object in butler is different from new data.
216 Exception
217 Raised if ``extendRun`` is `False` and datasets already exists.
218 Content of a butler collection should not be changed if exception
219 is raised.
220 """
222 def logConfigMismatch(msg: str) -> None:
223 """Log messages about configuration mismatch."""
224 _LOG.fatal("Comparing configuration: %s", msg)
226 _LOG.debug("Will save Configs for all tasks")
227 # start transaction to rollback any changes on exceptions
228 with self.transaction():
229 for taskDef in self._task_iter(graph):
230 # Config dataset ref is stored in task init outputs, but it
231 # may be also be missing.
232 task_output_refs = graph.initOutputRefs(taskDef)
233 if task_output_refs is None:
234 continue
236 config_ref, old_config = self._find_dataset(task_output_refs, taskDef.configDatasetName)
237 if config_ref is None:
238 continue
240 if old_config is not None:
241 if not taskDef.config.compare(old_config, shortcut=False, output=logConfigMismatch):
242 raise TypeError(
243 f"Config does not match existing task config {taskDef.configDatasetName!r} in "
244 "butler; tasks configurations must be consistent within the same run collection"
245 )
246 else:
247 _LOG.debug(
248 "Saving Config for task=%s dataset type=%s", taskDef.label, taskDef.configDatasetName
249 )
250 self.butler.put(taskDef.config, config_ref)
252 def savePackageVersions(self, graph: QuantumGraph) -> None:
253 """Write versions of software packages to butler.
255 Parameters
256 ----------
257 graph : `~lsst.pipe.base.QuantumGraph`
258 Execution graph.
260 Raises
261 ------
262 TypeError
263 Raised if existing object in butler is incompatible with new data.
264 """
265 packages = Packages.fromSystem()
266 _LOG.debug("want to save packages: %s", packages)
268 # start transaction to rollback any changes on exceptions
269 with self.transaction():
270 # Packages dataset ref is stored in graph's global init outputs,
271 # but it may be also be missing.
273 packages_ref, old_packages = self._find_dataset(
274 graph.globalInitOutputRefs(), PipelineDatasetTypes.packagesDatasetName
275 )
276 if packages_ref is None:
277 return
279 if old_packages is not None:
280 # Note that because we can only detect python modules that have
281 # been imported, the stored list of products may be more or
282 # less complete than what we have now. What's important is
283 # that the products that are in common have the same version.
284 _compare_packages(old_packages, packages)
285 # Update the old set of packages in case we have more packages
286 # that haven't been persisted.
287 extra = packages.extra(old_packages)
288 if extra:
289 _LOG.debug("extra packages: %s", extra)
290 old_packages.update(packages)
291 # have to remove existing dataset first, butler has no
292 # replace option.
293 self.butler.pruneDatasets([packages_ref], unstore=True, purge=True)
294 self.butler.put(old_packages, packages_ref)
295 else:
296 self.butler.put(packages, packages_ref)
298 def _find_dataset(
299 self, refs: Iterable[DatasetRef], dataset_type: str
300 ) -> tuple[DatasetRef | None, Any | None]:
301 """Find a ref with a given dataset type name in a list of references
302 and try to retrieve its data from butler.
304 Parameters
305 ----------
306 refs : `~collections.abc.Iterable` [ `~lsst.daf.butler.DatasetRef` ]
307 References to check for matching dataset type.
308 dataset_type : `str`
309 Name of a dataset type to look for.
311 Returns
312 -------
313 ref : `~lsst.daf.butler.DatasetRef` or `None`
314 Dataset reference or `None` if there is no matching dataset type.
315 data : `Any`
316 An existing object extracted from butler, `None` if ``ref`` is
317 `None` or if there is no existing object for that reference.
318 """
319 ref: DatasetRef | None = None
320 for ref in refs:
321 if ref.datasetType.name == dataset_type:
322 break
323 else:
324 return None, None
326 try:
327 data = self.butler.get(ref)
328 if data is not None and not self.extendRun:
329 # It must not exist unless we are extending run.
330 raise ConflictingDefinitionError(f"Dataset {ref} already exists in butler")
331 except (LookupError, FileNotFoundError):
332 data = None
333 return ref, data
335 def _task_iter(self, graph: QuantumGraph) -> Iterator[TaskDef]:
336 """Iterate over TaskDefs in a graph, return only tasks that have one or
337 more associated quanta.
338 """
339 for taskDef in graph.iterTaskGraph():
340 if graph.getNumberOfQuantaForTask(taskDef) > 0:
341 yield taskDef
343 @contextmanager
344 def transaction(self) -> Iterator[None]:
345 """Context manager for transaction.
347 Default implementation has no transaction support.
348 """
349 yield
352class PreExecInit(PreExecInitBase):
353 """Initialization of registry for QuantumGraph execution.
355 This class encapsulates all necessary operations that have to be performed
356 on butler and registry to prepare them for QuantumGraph execution.
358 Parameters
359 ----------
360 butler : `~lsst.daf.butler.Butler`
361 Data butler instance.
362 taskFactory : `~lsst.pipe.base.TaskFactory`
363 Task factory.
364 extendRun : `bool`, optional
365 If `True` then do not try to overwrite any datasets that might exist
366 in ``butler.run``; instead compare them when appropriate/possible. If
367 `False`, then any existing conflicting dataset will cause a butler
368 exception to be raised.
369 """
371 def __init__(self, butler: Butler, taskFactory: TaskFactory, extendRun: bool = False):
372 super().__init__(butler, taskFactory, extendRun)
373 self.full_butler = butler
374 if self.extendRun and self.full_butler.run is None:
375 raise RuntimeError(
376 "Cannot perform extendRun logic unless butler is initialized "
377 "with a default output RUN collection."
378 )
380 @contextmanager
381 def transaction(self) -> Iterator[None]:
382 # dosctring inherited
383 with self.full_butler.transaction():
384 yield
386 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
387 # docstring inherited
388 pipeline = graph.taskGraph
389 pipelineDatasetTypes = PipelineDatasetTypes.fromPipeline(
390 pipeline, registry=self.full_butler.registry, include_configs=True, include_packages=True
391 )
393 for datasetTypes, is_input in (
394 (pipelineDatasetTypes.initIntermediates, True),
395 (pipelineDatasetTypes.initOutputs, False),
396 (pipelineDatasetTypes.intermediates, True),
397 (pipelineDatasetTypes.outputs, False),
398 ):
399 self._register_output_dataset_types(registerDatasetTypes, datasetTypes, is_input)
401 def _register_output_dataset_types(
402 self, registerDatasetTypes: bool, datasetTypes: Iterable[DatasetType], is_input: bool
403 ) -> None:
404 def _check_compatibility(datasetType: DatasetType, expected: DatasetType, is_input: bool) -> bool:
405 # These are output dataset types so check for compatibility on put.
406 is_compatible = expected.is_compatible_with(datasetType)
408 if is_input:
409 # This dataset type is also used for input so must be
410 # compatible on get as ell.
411 is_compatible = is_compatible and datasetType.is_compatible_with(expected)
413 if is_compatible:
414 _LOG.debug(
415 "The dataset type configurations differ (%s from task != %s from registry) "
416 "but the storage classes are compatible. Can continue.",
417 datasetType,
418 expected,
419 )
420 return is_compatible
422 missing_datasetTypes = set()
423 for datasetType in datasetTypes:
424 # Only composites are registered, no components, and by this point
425 # the composite should already exist.
426 if registerDatasetTypes and not datasetType.isComponent():
427 _LOG.debug("Registering DatasetType %s with registry", datasetType)
428 # this is a no-op if it already exists and is consistent,
429 # and it raises if it is inconsistent.
430 try:
431 self.full_butler.registry.registerDatasetType(datasetType)
432 except ConflictingDefinitionError:
433 if not _check_compatibility(
434 datasetType, self.full_butler.get_dataset_type(datasetType.name), is_input
435 ):
436 raise
437 else:
438 _LOG.debug("Checking DatasetType %s against registry", datasetType)
439 try:
440 expected = self.full_butler.get_dataset_type(datasetType.name)
441 except KeyError:
442 # Likely means that --register-dataset-types is forgotten.
443 missing_datasetTypes.add(datasetType.name)
444 continue
445 if expected != datasetType:
446 if not _check_compatibility(datasetType, expected, is_input):
447 raise ValueError(
448 f"DatasetType configuration does not match Registry: {datasetType} != {expected}"
449 )
451 if missing_datasetTypes:
452 plural = "s" if len(missing_datasetTypes) != 1 else ""
453 raise KeyError(
454 f"Missing dataset type definition{plural}: {', '.join(missing_datasetTypes)}. "
455 "Dataset types have to be registered with either `butler register-dataset-type` or "
456 "passing `--register-dataset-types` option to `pipetask run`."
457 )
460class PreExecInitLimited(PreExecInitBase):
461 """Initialization of registry for QuantumGraph execution.
463 This class works with LimitedButler and expects that all references in
464 QuantumGraph are resolved.
466 Parameters
467 ----------
468 butler : `~lsst.daf.butler.LimitedButler`
469 Limited data butler instance.
470 taskFactory : `~lsst.pipe.base.TaskFactory`
471 Task factory.
472 """
474 def __init__(self, butler: LimitedButler, taskFactory: TaskFactory):
475 super().__init__(butler, taskFactory, False)
477 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
478 # docstring inherited
479 # With LimitedButler we never create or check dataset types.
480 pass