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

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

# This file is part of verify. 

# 

# Developed for the LSST Data Management System. 

# This product includes software developed by the LSST Project 

# (https://www.lsst.org). 

# See the COPYRIGHT file at the top-level directory of this distribution 

# for details of code ownership. 

# 

# This program is free software: you can redistribute it and/or modify 

# it under the terms of the GNU General Public License as published by 

# the Free Software Foundation, either version 3 of the License, or 

# (at your option) any later version. 

# 

# This program is distributed in the hope that it will be useful, 

# but WITHOUT ANY WARRANTY; without even the implied warranty of 

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

# GNU General Public License for more details. 

# 

# You should have received a copy of the GNU General Public License 

# along with this program. If not, see <https://www.gnu.org/licenses/>. 

 

 

__all__ = ["MetricComputationError", "MetricTask", "MetricConfig", 

"MetricConnections"] 

 

 

import abc 

import traceback 

 

import lsst.pipe.base as pipeBase 

from lsst.pipe.base import connectionTypes 

 

from lsst.verify import Name 

 

 

class MetricComputationError(RuntimeError): 

"""This class represents unresolvable errors in computing a metric. 

 

`lsst.verify.tasks.MetricTask` raises ``MetricComputationError`` 

instead of other data- or processing-related exceptions to let code that 

calls a mix of data processing and metric tasks distinguish between 

the two. Therefore, most ``MetricComputationError`` instances should be 

chained to another exception representing the underlying problem. 

""" 

pass 

 

 

class MetricConnections(pipeBase.PipelineTaskConnections, 

defaultTemplates={"package": None, "metric": None}, 

dimensions={"instrument", "visit", "detector"}, 

): 

"""An abstract connections class defining a metric output. 

 

This class assumes detector-level metrics, which is the most common case. 

Subclasses can redeclare ``measurement`` and ``dimensions`` to override 

this assumption. 

 

Notes 

----- 

``MetricConnections`` defines the following dataset templates: 

``package`` 

Name of the metric's namespace. By 

:ref:`verify_metrics <verify-metrics-package>` convention, this is 

the name of the package the metric is most closely 

associated with. 

``metric`` 

Name of the metric, excluding any namespace. 

""" 

measurement = connectionTypes.Output( 

name="metricvalue_{package}_{metric}", 

doc="The metric value computed by this task.", 

storageClass="MetricValue", 

dimensions={"instrument", "visit", "detector"}, 

) 

 

 

class MetricConfig(pipeBase.PipelineTaskConfig, 

pipelineConnections=MetricConnections): 

 

def validate(self): 

super().validate() 

 

if "." in self.connections.package: 

raise ValueError(f"package name {self.connections.package} must " 

"not contain periods") 

if "." in self.connections.metric: 

raise ValueError(f"metric name {self.connections.metric} must " 

"not contain periods; use connections.package " 

"instead") 

 

@property 

def metricName(self): 

"""The metric calculated by a `MetricTask` with this config 

(`lsst.verify.Name`, read-only). 

""" 

return Name(package=self.connections.package, 

metric=self.connections.metric) 

 

 

class MetricTask(pipeBase.PipelineTask, metaclass=abc.ABCMeta): 

"""A base class for tasks that compute one metric from input datasets. 

 

Parameters 

---------- 

*args 

**kwargs 

Constructor parameters are the same as for 

`lsst.pipe.base.PipelineTask`. 

 

Notes 

----- 

In general, both the ``MetricTask``'s metric and its input data are 

configurable. Metrics may be associated with a data ID at any level of 

granularity, including repository-wide. 

 

Like `lsst.pipe.base.PipelineTask`, this class should be customized by 

overriding `run` and by providing a `lsst.pipe.base.connectionTypes.Input` 

for each parameter of `run`. For requirements that are specific to 

``MetricTask``, see `run`. 

""" 

 

ConfigClass = MetricConfig 

 

def __init__(self, **kwargs): 

super().__init__(**kwargs) 

 

@abc.abstractmethod 

def run(self, **kwargs): 

"""Run the MetricTask on in-memory data. 

 

Parameters 

---------- 

**kwargs 

Keyword arguments matching the inputs given in the class config; 

see `lsst.pipe.base.PipelineTask.run` for more details. 

 

Returns 

------- 

struct : `lsst.pipe.base.Struct` 

A `~lsst.pipe.base.Struct` containing at least the 

following component: 

 

- ``measurement``: the value of the metric 

(`lsst.verify.Measurement` or `None`). This method is not 

responsible for adding mandatory metadata (e.g., the data ID); 

this is handled by the caller. 

 

Raises 

------ 

lsst.verify.tasks.MetricComputationError 

Raised if an algorithmic or system error prevents calculation 

of the metric. Examples include corrupted input data or 

unavoidable exceptions raised by analysis code. The 

`~lsst.verify.tasks.MetricComputationError` should be chained to a 

more specific exception describing the root cause. 

 

Not having enough data for a metric to be applicable is not an 

error, and should not trigger this exception. 

 

Notes 

----- 

All input data must be treated as optional. This maximizes the 

``MetricTask``'s usefulness for incomplete pipeline runs or runs with 

optional processing steps. If a metric cannot be calculated because 

the necessary inputs are missing, the ``MetricTask`` must return `None` 

in place of the measurement. 

""" 

 

def runQuantum(self, butlerQC, inputRefs, outputRefs): 

"""Do Butler I/O to provide in-memory objects for run. 

 

This specialization of runQuantum performs error-handling specific to 

MetricTasks. Most or all of this functionality may be moved to 

activators in the future. 

""" 

try: 

inputs = butlerQC.get(inputRefs) 

outputs = self.run(**inputs) 

if outputs.measurement is not None: 

butlerQC.put(outputs, outputRefs) 

else: 

self.log.debugf("Skipping measurement of {!r} on {} " 

"as not applicable.", self, inputRefs) 

except MetricComputationError: 

# Apparently lsst.log doesn't have built-in exception support? 

self.log.errorf( 

"Measurement of {!r} failed on {}->{}\n{}", 

self, inputRefs, outputRefs, traceback.format_exc()) 

 

def adaptArgsAndRun(self, inputData, inputDataIds, outputDataId): 

"""A wrapper around `run` used by 

`~lsst.verify.gen2tasks.MetricsControllerTask`. 

 

Task developers should not need to call or override this method. 

 

Parameters 

---------- 

inputData : `dict` from `str` to any 

Dictionary whose keys are the names of input parameters and values 

are Python-domain data objects (or lists of objects) retrieved 

from data butler. Input objects may be `None` to represent 

missing data. 

inputDataIds : `dict` from `str` to `list` of dataId 

Dictionary whose keys are the names of input parameters and values 

are data IDs (or lists of data IDs) that the task consumes for 

corresponding dataset type. Data IDs are guaranteed to match data 

objects in ``inputData``. 

outputDataId : `dict` from `str` to dataId 

Dictionary containing a single key, ``"measurement"``, which maps 

to a single data ID for the measurement. The data ID must have the 

same granularity as the metric. 

 

Returns 

------- 

struct : `lsst.pipe.base.Struct` 

A `~lsst.pipe.base.Struct` containing at least the 

following component: 

 

- ``measurement``: the value of the metric, computed from 

``inputData`` (`lsst.verify.Measurement` or `None`). The 

measurement is guaranteed to contain not only the value of the 

metric, but also any mandatory supplementary information. 

 

Raises 

------ 

lsst.verify.tasks.MetricComputationError 

Raised if an algorithmic or system error prevents calculation 

of the metric. Examples include corrupted input data or 

unavoidable exceptions raised by analysis code. The 

`~lsst.verify.tasks.MetricComputationError` should be chained to a 

more specific exception describing the root cause. 

 

Not having enough data for a metric to be applicable is not an 

error, and should not trigger this exception. 

 

Notes 

----- 

This implementation calls `run` on the contents of ``inputData``, 

followed by calling `addStandardMetadata` on the result before 

returning it. 

 

Examples 

-------- 

Consider a metric that characterizes PSF variations across the entire 

field of view, given processed images. Then, if `run` has the 

signature ``run(images)``: 

 

.. code-block:: py 

 

inputData = {'images': [image1, image2, ...]} 

inputDataIds = {'images': [{'visit': 42, 'ccd': 1}, 

{'visit': 42, 'ccd': 2}, 

...]} 

outputDataId = {'measurement': {'visit': 42}} 

result = task.adaptArgsAndRun( 

inputData, inputDataIds, outputDataId) 

""" 

result = self.run(**inputData) 

if result.measurement is not None: 

self.addStandardMetadata(result.measurement, 

outputDataId["measurement"]) 

return result 

 

@classmethod 

def getInputDatasetTypes(cls, config): 

"""Return input dataset types for this task. 

 

Parameters 

---------- 

config : ``cls.ConfigClass`` 

Configuration for this task. 

 

Returns 

------- 

datasets : `dict` from `str` to `str` 

Dictionary where the key is the name of the input dataset (must 

match a parameter to `run`) and the value is the name of its 

Butler dataset type. 

 

Notes 

----- 

The default implementation extracts a 

`~lsst.pipe.base.PipelineTaskConnections` object from ``config``. 

""" 

# Get connections from config for backward-compatibility 

connections = config.connections.ConnectionsClass(config=config) 

return {name: getattr(connections, name).name 

for name in connections.inputs} 

 

@classmethod 

def areInputDatasetsScalar(cls, config): 

"""Return input dataset multiplicity. 

 

Parameters 

---------- 

config : ``cls.ConfigClass`` 

Configuration for this task. 

 

Returns 

------- 

datasets : `Dict` [`str`, `bool`] 

Dictionary where the key is the name of the input dataset (must 

match a parameter to `run`) and the value is `True` if `run` takes 

only one object and `False` if it takes a list. 

 

Notes 

----- 

The default implementation extracts a 

`~lsst.pipe.base.PipelineTaskConnections` object from ``config``. 

""" 

connections = config.connections.ConnectionsClass(config=config) 

return {name: not getattr(connections, name).multiple 

for name in connections.inputs} 

 

def addStandardMetadata(self, measurement, outputDataId): 

"""Add data ID-specific metadata required for all metrics. 

 

This method currently does not add any metadata, but may do so 

in the future. 

 

Parameters 

---------- 

measurement : `lsst.verify.Measurement` 

The `~lsst.verify.Measurement` that the metadata are added to. 

outputDataId : ``dataId`` 

The data ID to which the measurement applies, at the appropriate 

level of granularity. 

 

Notes 

----- 

This method should not be overridden by subclasses. 

 

This method is not responsible for shared metadata like the execution 

environment (which should be added by this ``MetricTask``'s caller), 

nor for metadata specific to a particular metric (which should be 

added when the metric is calculated). 

 

.. warning:: 

This method's signature will change whenever additional data needs 

to be provided. This is a deliberate restriction to ensure that all 

subclasses pass in the new data as well. 

""" 

pass