Coverage for python/lsst/ctrl/mpexec/preExecInit.py: 19%
145 statements
« prev ^ index » next coverage.py v7.5.0, created at 2024-05-03 09:59 +0000
« prev ^ index » next coverage.py v7.5.0, created at 2024-05-03 09:59 +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, MissingDatasetTypeError
46from lsst.pipe.base.automatic_connection_constants import (
47 PACKAGES_INIT_OUTPUT_NAME,
48 PACKAGES_INIT_OUTPUT_STORAGE_CLASS,
49)
50from lsst.utils.packages import Packages
52if TYPE_CHECKING:
53 from lsst.daf.butler import Butler, LimitedButler
54 from lsst.pipe.base import QuantumGraph, TaskDef, TaskFactory
56_LOG = logging.getLogger(__name__)
59class MissingReferenceError(Exception):
60 """Exception raised when resolved reference is missing from graph."""
62 pass
65def _compare_packages(old_packages: Packages, new_packages: Packages) -> None:
66 """Compare two versions of Packages.
68 Parameters
69 ----------
70 old_packages : `Packages`
71 Previously recorded package versions.
72 new_packages : `Packages`
73 New set of package versions.
75 Raises
76 ------
77 TypeError
78 Raised if parameters are inconsistent.
79 """
80 diff = new_packages.difference(old_packages)
81 if diff:
82 versions_str = "; ".join(f"{pkg}: {diff[pkg][1]} vs {diff[pkg][0]}" for pkg in diff)
83 raise TypeError(f"Package versions mismatch: ({versions_str})")
84 else:
85 _LOG.debug("new packages are consistent with old")
88class PreExecInitBase(abc.ABC):
89 """Common part of the implementation of PreExecInit classes that does not
90 depend on Butler type.
92 Parameters
93 ----------
94 butler : `~lsst.daf.butler.LimitedButler`
95 Butler to use.
96 taskFactory : `lsst.pipe.base.TaskFactory`
97 Task factory.
98 extendRun : `bool`
99 Whether extend run parameter is in use.
100 """
102 def __init__(self, butler: LimitedButler, taskFactory: TaskFactory, extendRun: bool):
103 self.butler = butler
104 self.taskFactory = taskFactory
105 self.extendRun = extendRun
107 def initialize(
108 self,
109 graph: QuantumGraph,
110 saveInitOutputs: bool = True,
111 registerDatasetTypes: bool = False,
112 saveVersions: bool = True,
113 ) -> None:
114 """Perform all initialization steps.
116 Convenience method to execute all initialization steps. Instead of
117 calling this method and providing all options it is also possible to
118 call methods individually.
120 Parameters
121 ----------
122 graph : `~lsst.pipe.base.QuantumGraph`
123 Execution graph.
124 saveInitOutputs : `bool`, optional
125 If ``True`` (default) then save "init outputs", configurations,
126 and package versions to butler.
127 registerDatasetTypes : `bool`, optional
128 If ``True`` then register dataset types in registry, otherwise
129 they must be already registered.
130 saveVersions : `bool`, optional
131 If ``False`` then do not save package versions even if
132 ``saveInitOutputs`` is set to ``True``.
133 """
134 # register dataset types or check consistency
135 self.initializeDatasetTypes(graph, registerDatasetTypes)
137 # Save task initialization data or check that saved data
138 # is consistent with what tasks would save
139 if saveInitOutputs:
140 self.saveInitOutputs(graph)
141 self.saveConfigs(graph)
142 if saveVersions:
143 self.savePackageVersions(graph)
145 @abc.abstractmethod
146 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
147 """Save or check DatasetTypes output by the tasks in a graph.
149 Iterates over all DatasetTypes for all tasks in a graph and either
150 tries to add them to registry or compares them to existing ones.
152 Parameters
153 ----------
154 graph : `~lsst.pipe.base.QuantumGraph`
155 Execution graph.
156 registerDatasetTypes : `bool`, optional
157 If ``True`` then register dataset types in registry, otherwise
158 they must be already registered.
160 Raises
161 ------
162 ValueError
163 Raised if existing DatasetType is different from DatasetType
164 in a graph.
165 KeyError
166 Raised if ``registerDatasetTypes`` is ``False`` and DatasetType
167 does not exist in registry.
168 """
169 raise NotImplementedError()
171 def saveInitOutputs(self, graph: QuantumGraph) -> None:
172 """Write any datasets produced by initializing tasks in a graph.
174 Parameters
175 ----------
176 graph : `~lsst.pipe.base.QuantumGraph`
177 Execution graph.
179 Raises
180 ------
181 TypeError
182 Raised if the type of existing object in butler is different from
183 new data.
184 """
185 _LOG.debug("Will save InitOutputs for all tasks")
186 for taskDef in self._task_iter(graph):
187 init_input_refs = graph.initInputRefs(taskDef) or []
188 task = self.taskFactory.makeTask(
189 graph.pipeline_graph.tasks[taskDef.label], self.butler, init_input_refs
190 )
191 for name in taskDef.connections.initOutputs:
192 attribute = getattr(taskDef.connections, name)
193 init_output_refs = graph.initOutputRefs(taskDef) or []
194 init_output_ref, obj_from_store = self._find_dataset(init_output_refs, attribute.name)
195 if init_output_ref is None:
196 raise ValueError(f"Cannot find dataset reference for init output {name} in a graph")
197 init_output_var = getattr(task, name)
199 if obj_from_store is not None:
200 _LOG.debug(
201 "Retrieving InitOutputs for task=%s key=%s dsTypeName=%s", task, name, attribute.name
202 )
203 obj_from_store = self.butler.get(init_output_ref)
204 # Types are supposed to be identical.
205 # TODO: Check that object contents is identical too.
206 if type(obj_from_store) is not type(init_output_var):
207 raise TypeError(
208 f"Stored initOutput object type {type(obj_from_store)} "
209 "is different from task-generated type "
210 f"{type(init_output_var)} for task {taskDef}"
211 )
212 else:
213 _LOG.debug("Saving InitOutputs for task=%s key=%s", taskDef.label, name)
214 # This can still raise if there is a concurrent write.
215 self.butler.put(init_output_var, init_output_ref)
217 def saveConfigs(self, graph: QuantumGraph) -> None:
218 """Write configurations for pipeline tasks to butler or check that
219 existing configurations are equal to the new ones.
221 Parameters
222 ----------
223 graph : `~lsst.pipe.base.QuantumGraph`
224 Execution graph.
226 Raises
227 ------
228 TypeError
229 Raised if existing object in butler is different from new data.
230 Exception
231 Raised if ``extendRun`` is `False` and datasets already exists.
232 Content of a butler collection should not be changed if exception
233 is raised.
234 """
236 def logConfigMismatch(msg: str) -> None:
237 """Log messages about configuration mismatch.
239 Parameters
240 ----------
241 msg : `str`
242 Log message to use.
243 """
244 _LOG.fatal("Comparing configuration: %s", msg)
246 _LOG.debug("Will save Configs for all tasks")
247 # start transaction to rollback any changes on exceptions
248 with self.transaction():
249 for taskDef in self._task_iter(graph):
250 # Config dataset ref is stored in task init outputs, but it
251 # may be also be missing.
252 task_output_refs = graph.initOutputRefs(taskDef)
253 if task_output_refs is None:
254 continue
256 config_ref, old_config = self._find_dataset(task_output_refs, taskDef.configDatasetName)
257 if config_ref is None:
258 continue
260 if old_config is not None:
261 if not taskDef.config.compare(old_config, shortcut=False, output=logConfigMismatch):
262 raise TypeError(
263 f"Config does not match existing task config {taskDef.configDatasetName!r} in "
264 "butler; tasks configurations must be consistent within the same run collection"
265 )
266 else:
267 _LOG.debug(
268 "Saving Config for task=%s dataset type=%s", taskDef.label, taskDef.configDatasetName
269 )
270 self.butler.put(taskDef.config, config_ref)
272 def savePackageVersions(self, graph: QuantumGraph) -> None:
273 """Write versions of software packages to butler.
275 Parameters
276 ----------
277 graph : `~lsst.pipe.base.QuantumGraph`
278 Execution graph.
280 Raises
281 ------
282 TypeError
283 Raised if existing object in butler is incompatible with new data.
284 """
285 packages = Packages.fromSystem()
286 _LOG.debug("want to save packages: %s", packages)
288 # start transaction to rollback any changes on exceptions
289 with self.transaction():
290 # Packages dataset ref is stored in graph's global init outputs,
291 # but it may be also be missing.
293 packages_ref, old_packages = self._find_dataset(
294 graph.globalInitOutputRefs(), PACKAGES_INIT_OUTPUT_NAME
295 )
296 if packages_ref is None:
297 return
299 if old_packages is not None:
300 # Note that because we can only detect python modules that have
301 # been imported, the stored list of products may be more or
302 # less complete than what we have now. What's important is
303 # that the products that are in common have the same version.
304 _compare_packages(old_packages, packages)
305 # Update the old set of packages in case we have more packages
306 # that haven't been persisted.
307 extra = packages.extra(old_packages)
308 if extra:
309 _LOG.debug("extra packages: %s", extra)
310 old_packages.update(packages)
311 # have to remove existing dataset first, butler has no
312 # replace option.
313 self.butler.pruneDatasets([packages_ref], unstore=True, purge=True)
314 self.butler.put(old_packages, packages_ref)
315 else:
316 self.butler.put(packages, packages_ref)
318 def _find_dataset(
319 self, refs: Iterable[DatasetRef], dataset_type: str
320 ) -> tuple[DatasetRef | None, Any | None]:
321 """Find a ref with a given dataset type name in a list of references
322 and try to retrieve its data from butler.
324 Parameters
325 ----------
326 refs : `~collections.abc.Iterable` [ `~lsst.daf.butler.DatasetRef` ]
327 References to check for matching dataset type.
328 dataset_type : `str`
329 Name of a dataset type to look for.
331 Returns
332 -------
333 ref : `~lsst.daf.butler.DatasetRef` or `None`
334 Dataset reference or `None` if there is no matching dataset type.
335 data : `Any`
336 An existing object extracted from butler, `None` if ``ref`` is
337 `None` or if there is no existing object for that reference.
338 """
339 ref: DatasetRef | None = None
340 for ref in refs:
341 if ref.datasetType.name == dataset_type:
342 break
343 else:
344 return None, None
346 try:
347 data = self.butler.get(ref)
348 if data is not None and not self.extendRun:
349 # It must not exist unless we are extending run.
350 raise ConflictingDefinitionError(f"Dataset {ref} already exists in butler")
351 except (LookupError, FileNotFoundError):
352 data = None
353 return ref, data
355 def _task_iter(self, graph: QuantumGraph) -> Iterator[TaskDef]:
356 """Iterate over TaskDefs in a graph, return only tasks that have one or
357 more associated quanta.
358 """
359 for taskDef in graph.iterTaskGraph():
360 if graph.getNumberOfQuantaForTask(taskDef) > 0:
361 yield taskDef
363 @contextmanager
364 def transaction(self) -> Iterator[None]:
365 """Context manager for transaction.
367 Default implementation has no transaction support.
369 Yields
370 ------
371 `None`
372 No transaction support.
373 """
374 yield
377class PreExecInit(PreExecInitBase):
378 """Initialization of registry for QuantumGraph execution.
380 This class encapsulates all necessary operations that have to be performed
381 on butler and registry to prepare them for QuantumGraph execution.
383 Parameters
384 ----------
385 butler : `~lsst.daf.butler.Butler`
386 Data butler instance.
387 taskFactory : `~lsst.pipe.base.TaskFactory`
388 Task factory.
389 extendRun : `bool`, optional
390 If `True` then do not try to overwrite any datasets that might exist
391 in ``butler.run``; instead compare them when appropriate/possible. If
392 `False`, then any existing conflicting dataset will cause a butler
393 exception to be raised.
394 """
396 def __init__(self, butler: Butler, taskFactory: TaskFactory, extendRun: bool = False):
397 super().__init__(butler, taskFactory, extendRun)
398 self.full_butler = butler
399 if self.extendRun and self.full_butler.run is None:
400 raise RuntimeError(
401 "Cannot perform extendRun logic unless butler is initialized "
402 "with a default output RUN collection."
403 )
405 @contextmanager
406 def transaction(self) -> Iterator[None]:
407 # dosctring inherited
408 with self.full_butler.transaction():
409 yield
411 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
412 # docstring inherited
413 missing_dataset_types: set[str] = set()
414 dataset_types = [node.dataset_type for node in graph.pipeline_graph.dataset_types.values()]
415 dataset_types.append(
416 DatasetType(
417 PACKAGES_INIT_OUTPUT_NAME, self.butler.dimensions.empty, PACKAGES_INIT_OUTPUT_STORAGE_CLASS
418 )
419 )
420 for dataset_type in dataset_types:
421 # Resolving the PipelineGraph when building the QuantumGraph should
422 # have already guaranteed that this is the registry dataset type
423 # and that all references to it use compatible storage classes,
424 # so we don't need another check for compatibility here; if the
425 # dataset type doesn't match the registry that's already a problem.
426 if registerDatasetTypes:
427 _LOG.debug("Registering DatasetType %s with registry", dataset_type.name)
428 try:
429 self.full_butler.registry.registerDatasetType(dataset_type)
430 except ConflictingDefinitionError:
431 expected = self.full_butler.registry.getDatasetType(dataset_type.name)
432 raise ConflictingDefinitionError(
433 f"DatasetType definition in registry has changed since the QuantumGraph was built: "
434 f"{dataset_type} (graph) != {expected} (registry)."
435 )
436 else:
437 _LOG.debug("Checking DatasetType %s against registry", dataset_type.name)
438 try:
439 expected = self.full_butler.registry.getDatasetType(dataset_type.name)
440 except MissingDatasetTypeError:
441 # Likely means that --register-dataset-types is forgotten,
442 # but we could also get here if there is a prerequisite
443 # input that is optional and none were found in this repo;
444 # that is not an error. And we don't bother to check if
445 # they are optional here, since the fact that we were able
446 # to make the QG says that they were, since there couldn't
447 # have been any datasets if the dataset types weren't
448 # registered.
449 if not graph.pipeline_graph.dataset_types[dataset_type.name].is_prerequisite:
450 missing_dataset_types.add(dataset_type.name)
451 continue
452 if expected != dataset_type:
453 raise ConflictingDefinitionError(
454 f"DatasetType definition in registry has changed since the QuantumGraph was built: "
455 f"{dataset_type} (graph) != {expected} (registry)."
456 )
457 if missing_dataset_types:
458 plural = "s" if len(missing_dataset_types) != 1 else ""
459 raise MissingDatasetTypeError(
460 f"Missing dataset type definition{plural}: {', '.join(missing_dataset_types)}. "
461 "Dataset types have to be registered with either `butler register-dataset-type` or "
462 "passing `--register-dataset-types` option to `pipetask run`."
463 )
466class PreExecInitLimited(PreExecInitBase):
467 """Initialization of registry for QuantumGraph execution.
469 This class works with LimitedButler and expects that all references in
470 QuantumGraph are resolved.
472 Parameters
473 ----------
474 butler : `~lsst.daf.butler.LimitedButler`
475 Limited data butler instance.
476 taskFactory : `~lsst.pipe.base.TaskFactory`
477 Task factory.
478 """
480 def __init__(self, butler: LimitedButler, taskFactory: TaskFactory):
481 super().__init__(butler, taskFactory, False)
483 def initializeDatasetTypes(self, graph: QuantumGraph, registerDatasetTypes: bool = False) -> None:
484 # docstring inherited
485 # With LimitedButler we never create or check dataset types.
486 pass