Coverage for python/lsst/analysis/tools/interfaces/_analysisTools.py: 22%
176 statements
« prev ^ index » next coverage.py v7.5.0, created at 2024-04-24 04:10 -0700
« prev ^ index » next coverage.py v7.5.0, created at 2024-04-24 04:10 -0700
1# This file is part of analysis_tools.
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 <https://www.gnu.org/licenses/>.
22from __future__ import annotations
24__all__ = ("AnalysisTool",)
26from collections import ChainMap
27from collections.abc import Mapping
28from functools import wraps
29from operator import attrgetter
30from typing import Callable, Iterable, Protocol, runtime_checkable
32import lsst.pex.config as pexConfig
33from lsst.obs.base import Instrument
34from lsst.pex.config import Field, FieldValidationError, ListField
35from lsst.pex.config.configurableActions import ConfigurableActionField
36from lsst.pipe.base import Pipeline
37from lsst.verify import Measurement
39from ._actions import AnalysisAction, JointAction, JointResults, NoPlot, PlotAction
40from ._interfaces import KeyedData, KeyedDataSchema, KeyedResults, PlotTypes
41from ._stages import BasePrep, BaseProcess, BaseProduce
44@runtime_checkable
45class _HasOutputNames(Protocol):
46 def getOutputNames(self, config: pexConfig.Config | None = None) -> Iterable[str]: ... 46 ↛ exitline 46 didn't jump to line 46, because
49def _finalizeWrapper(
50 f: Callable[[AnalysisTool], None], cls: type[AnalysisTool]
51) -> Callable[[AnalysisTool], None]:
52 """Wrap a classes finalize function to ensure the base classes special
53 finalize method only fires after the most derived finalize method.
55 Parameters
56 ----------
57 f : `Callable`
58 Function that is being wrapped
59 cls : `type` of `AnalysisTool`
60 The class which is having its function wrapped
62 Returns
63 -------
64 function : `Callable`
65 The new function which wraps the old
66 """
68 @wraps(f)
69 def wrapper(self: AnalysisTool) -> None:
70 # call the wrapped finalize function
71 f(self)
72 # get the method resolution order for the self variable
73 mro = self.__class__.mro()
75 # Find which class in the mro that last defines a finalize method
76 # note that this is in the reverse order from the mro, because the
77 # last class in an inheritance stack is the first in the mro (aka you
78 # walk from the furthest child first.
79 #
80 # Also note that the most derived finalize method need not be the same
81 # as the type of self, as that might inherit from a parent somewhere
82 # between it and the furthest parent.
83 mostDerived: type | None = None
84 for klass in mro:
85 # inspect the classes dictionary to see if it specifically defines
86 # finalize. This is needed because normal lookup will go through
87 # the mro, but this needs to be restricted to each class.
88 if "finalize" in vars(klass):
89 mostDerived = klass
90 break
92 # Find what stage in the MRO walking process the recursive function
93 # call is in.
94 this = super(cls, self).__thisclass__
96 # If the current place in the MRO walking is also the class that
97 # defines the most derived instance of finalize, then call the base
98 # classes private finalize that must be called after everything else.
99 if mostDerived is not None and this == mostDerived:
100 self._baseFinalize()
102 return wrapper
105class AnalysisTool(AnalysisAction):
106 r"""A tool which which calculates a single type of analysis on input data,
107 though it may return more than one result.
109 Although `AnalysisTool`\ s are considered a single type of analysis, the
110 classes themselves can be thought of as a container. `AnalysisTool`\ s
111 are aggregations of `AnalysisAction`\ s to form prep, process, and
112 produce stages. These stages allow better reuse of individual
113 `AnalysisActions` and easier introspection in contexts such as a notebook
114 or interpreter.
116 An `AnalysisTool` can be thought of an an individual configuration that
117 specifies which `AnalysisAction` should run for each stage.
119 The stages themselves are also configurable, allowing control over various
120 aspects of the individual `AnalysisAction`\ s.
121 """
123 prep = ConfigurableActionField[AnalysisAction](doc="Action to run to prepare inputs", default=BasePrep)
124 process = ConfigurableActionField[AnalysisAction](
125 doc="Action to process data into intended form", default=BaseProcess
126 )
127 produce = ConfigurableActionField[AnalysisAction](
128 doc="Action to perform any finalization steps", default=BaseProduce
129 )
130 metric_tags = ListField[str](
131 doc="List of tags which will be associated with metric measurement(s)", default=[]
132 )
134 def __init_subclass__(cls: type[AnalysisTool], **kwargs):
135 super().__init_subclass__(**kwargs)
136 # Wrap all definitions of the finalize method in a special wrapper that
137 # ensures that the bases classes private finalize is called last.
138 if "finalize" in vars(cls):
139 cls.finalize = _finalizeWrapper(cls.finalize, cls)
141 dynamicOutputNames: bool | Field[bool] = False
142 """Determines whether to grant the ``getOutputNames`` method access to
143 config parameters.
144 """
146 parameterizedBand: bool | Field[bool] = True
147 """Specifies if an `AnalysisTool` may parameterize a band within any field
148 in any stage, or if the set of bands is already uniquely determined though
149 configuration. I.e. can this `AnalysisTool` be automatically looped over to
150 produce a result for multiple bands.
151 """
153 propagateData: bool | Field[bool] = False
154 """If this value is set to True, the input data `KeyedData` will be passed
155 to each stage in addition to the ``prep`` stage. Any keys created in a
156 stage with the same key that exists in the input ``data`` will shadow that
157 key/value.
158 """
160 def __call__(self, data: KeyedData, **kwargs) -> KeyedResults:
161 bands = kwargs.pop("bands", None)
162 if "plotInfo" in kwargs and kwargs.get("plotInfo") is not None:
163 if "plotName" not in kwargs["plotInfo"] or kwargs["plotInfo"]["plotName"] is None:
164 kwargs["plotInfo"]["plotName"] = self.identity
165 if not self.parameterizedBand or bands is None:
166 if "band" not in kwargs:
167 # Some tasks require a "band" key for naming. This shouldn't
168 # affect the results. DM-35813 should make this unnecessary.
169 kwargs["band"] = "analysisTools"
170 return self._call_single(data, **kwargs)
171 results: KeyedResults = {}
172 for band in bands:
173 kwargs["band"] = band
174 if "plotInfo" in kwargs:
175 kwargs["plotInfo"]["bands"] = band
176 subResult = self._call_single(data, **kwargs)
177 for key, value in subResult.items():
178 match value:
179 case PlotTypes():
180 results[f"{band}_{key}"] = value
181 case Measurement():
182 results[key] = value
183 return results
185 def _call_single(self, data: KeyedData, **kwargs) -> KeyedResults:
186 # create a shallow copy of kwargs
187 kwargs = dict(**kwargs)
188 kwargs["metric_tags"] = list(self.metric_tags or ())
189 prepped: KeyedData = self.prep(data, **kwargs) # type: ignore
190 if self.propagateData:
191 prepped = ChainMap(data, prepped)
192 processed: KeyedData = self.process(prepped, **kwargs) # type: ignore
193 if self.propagateData:
194 processed = ChainMap(data, processed)
195 finalized: (
196 Mapping[str, PlotTypes] | PlotTypes | Mapping[str, Measurement] | Measurement | JointResults
197 ) = self.produce(
198 processed, **kwargs
199 ) # type: ignore
200 return self._process_single_results(finalized)
202 def _getPlotType(self) -> str:
203 match self.produce:
204 case PlotAction():
205 return self.produce.getPlotType()
206 case JointAction(plot=NoPlot()):
207 pass
208 case JointAction(plot=plotter):
209 return plotter.getPlotType()
211 return ""
213 def _process_single_results(
214 self,
215 results: Mapping[str, PlotTypes] | PlotTypes | Mapping[str, Measurement] | Measurement | JointResults,
216 ) -> KeyedResults:
217 accumulation = {}
218 predicate = f"{self.identity}" if self.identity else ""
219 match results:
220 case Mapping():
221 suffix = self._getPlotType()
222 for key, value in results.items():
223 match value:
224 case Measurement():
225 iterable = (predicate, key)
226 case PlotTypes():
227 iterable = (predicate, key, suffix)
228 case _:
229 raise RuntimeError(f"Unexpected {key=}, {value=} from:\n{self=}")
230 refKey = "_".join(x for x in iterable if x)
231 accumulation[refKey] = value
232 case PlotTypes():
233 suffix = self._getPlotType()
234 refKey = "_".join(x for x in (predicate, suffix) if x)
235 accumulation[refKey] = results
236 case Measurement():
237 accumulation[f"{predicate}"] = results
238 case JointResults(plot=plotResults, metric=metricResults):
239 if plotResults is not None:
240 subResult = self._process_single_results(plotResults)
241 accumulation.update(subResult)
242 if metricResults is not None:
243 subResult = self._process_single_results(metricResults)
244 accumulation.update(subResult)
245 return accumulation
247 def getInputSchema(self) -> KeyedDataSchema:
248 return self.prep.getInputSchema()
250 def populatePrepFromProcess(self):
251 """Add additional inputs to the prep stage if supported.
253 If the configured prep action supports adding to it's input schema,
254 attempt to add the required inputs schema from the process stage to the
255 prep stage.
257 This method will be a no-op if the prep action does not support this
258 feature.
259 """
260 self.prep.addInputSchema(self.process.getInputSchema())
262 def getOutputNames(self, config: pexConfig.Config | None = None) -> Iterable[str]:
263 """Return the names of the plots produced by this analysis tool.
265 If there is a `PlotAction` defined in the produce action, these names
266 will either come from the `PlotAction` if it defines a
267 ``getOutputNames`` method (likely if it returns a mapping of figures),
268 or a default value is used and a single figure is assumed.
270 Parameters
271 ----------
272 config : `lsst.pex.config.Config`, optional
273 Configuration of the task. This is only used if the output naming
274 needs to be config-aware.
276 Returns
277 -------
278 result : `tuple` of `str`
279 Names for each plot produced by this action.
280 """
281 match self.produce:
282 case JointAction(plot=NoPlot()):
283 return tuple()
284 case _HasOutputNames():
285 outNames = tuple(self.produce.getOutputNames(config=config))
286 case _:
287 raise ValueError(f"Unsupported Action type {type(self.produce)} for getting output names")
289 results = []
290 suffix = self._getPlotType()
291 if self.parameterizedBand:
292 prefix = "_".join(x for x in ("{band}", self.identity) if x)
293 else:
294 prefix = f"{self.identity}" if self.identity else ""
296 if outNames:
297 for name in outNames:
298 results.append("_".join(x for x in (prefix, name, suffix) if x))
299 else:
300 results.append("_".join(x for x in (prefix, suffix) if x))
301 return results
303 @classmethod
304 def fromPipeline(
305 cls,
306 pipeline: str | Pipeline,
307 name: str,
308 fullpath: bool = False,
309 instrument: Instrument | str | None = None,
310 ) -> AnalysisTool | None:
311 """Construct an `AnalysisTool` from a definition written in a
312 `~lsst.pipe.base.Pipeline`.
314 Parameters
315 ----------
316 pipeline : `str` or `~lsst.pipe.base.Pipeline`
317 The pipeline to load the `AnalysisTool` from.
318 name : `str`
319 The name of the analysis tool to run. This can either be just the
320 name assigned to the tool, or an absolute name in a config
321 hierarchy.
322 fullpath : `bool`
323 Determines if the name is interpreted as an absolute path in a
324 config hierarchy, or is relative to an `AnalysisTool` ``atools``
325 `~lsst.pex.config.configurableActions.ConfigurableActionStructField`
326 .
327 instrument : `~lsst.daf.butler.instrument.Instrument` or `str` or\
328 `None`
329 Either a derived class object of a `lsst.daf.butler.instrument` or
330 a string corresponding to a fully qualified
331 `lsst.daf.butler.instrument` name or None if no instrument needs
332 specified or if the pipeline contains the instrument in it.
333 Defaults to None.
335 Returns
336 -------
337 tool : `AnalysisTool`
338 The loaded `AnalysisTool` as configured in the pipeline.
340 Raises
341 ------
342 ValueError
343 Raised if the config field specified does not point to an
344 `AnalysisTool`.
345 Raised if an instrument is specified and it conflicts with the
346 pipelines instrument.
347 """
348 if not isinstance(pipeline, Pipeline):
349 pipeline = Pipeline.fromFile(pipeline)
350 # If the caller specified an instrument, verify it does not conflict
351 # with the pipelines instrument, and add it to the pipeline
352 if instrument is not None:
353 if (pipeInstrument := pipeline.getInstrument()) and pipeInstrument != instrument:
354 raise ValueError(
355 f"The supplied instrument {instrument} conflicts with the pipelines instrument "
356 f"{pipeInstrument}."
357 )
358 else:
359 pipeline.addInstrument(instrument)
360 try:
361 pipelineGraph = pipeline.to_graph()
362 except (FieldValidationError, ValueError) as err:
363 raise ValueError(
364 "There was an error instantiating the pipeline, do you need to specify an instrument?"
365 ) from err
366 if not fullpath:
367 name = f"atools.{name}"
368 for task in pipelineGraph.tasks.values():
369 config = task.config
370 try:
371 attr = attrgetter(name)(config)
372 except AttributeError:
373 continue
374 if not isinstance(attr, AnalysisTool):
375 raise ValueError("The requested name did not refer to an analysisTool")
376 return attr
377 return None
379 def finalize(self) -> None:
380 """Run any finalization code that depends on configuration being
381 complete.
382 """
383 pass
385 def _baseFinalize(self) -> None:
386 self.populatePrepFromProcess()
388 def freeze(self):
389 if not self.__dict__.get("_finalizeRun"):
390 self.finalize()
391 self.__dict__["_finalizeRun"] = True
392 super().freeze()
395# explicitly wrap the finalize of the base class
396AnalysisTool.finalize = _finalizeWrapper(AnalysisTool.finalize, AnalysisTool)