Coverage for tests/test_simple_pipeline_executor.py: 25%

132 statements  

« prev     ^ index     » next       coverage.py v7.4.4, created at 2024-03-20 04:15 -0700

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# (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 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 <https://www.gnu.org/licenses/>. 

27 

28from __future__ import annotations 

29 

30import os 

31import shutil 

32import tempfile 

33import unittest 

34from typing import Any 

35 

36import lsst.daf.butler 

37import lsst.utils.tests 

38from lsst.ctrl.mpexec import SimplePipelineExecutor 

39from lsst.pipe.base import Struct, TaskDef, TaskMetadata, connectionTypes 

40from lsst.pipe.base.tests.no_dimensions import ( 

41 NoDimensionsTestConfig, 

42 NoDimensionsTestConnections, 

43 NoDimensionsTestTask, 

44) 

45from lsst.utils.introspection import get_full_type_name 

46 

47TESTDIR = os.path.abspath(os.path.dirname(__file__)) 

48 

49 

50class NoDimensionsTestConnections2(NoDimensionsTestConnections, dimensions=set()): 

51 """A connections class used for testing.""" 

52 

53 input = connectionTypes.Input( 

54 name="input", doc="some dict-y input data for testing", storageClass="TaskMetadataLike" 

55 ) 

56 

57 

58class NoDimensionsTestConfig2(NoDimensionsTestConfig, pipelineConnections=NoDimensionsTestConnections2): 

59 """A config used for testing.""" 

60 

61 

62class NoDimensionsMetadataTestConnections(NoDimensionsTestConnections, dimensions=set()): 

63 """Test connection class for metadata. 

64 

65 Deliberately choose a storage class that does not match the metadata 

66 default TaskMetadata storage class. 

67 """ 

68 

69 meta = connectionTypes.Input( 

70 name="a_metadata", doc="Metadata from previous task", storageClass="StructuredDataDict" 

71 ) 

72 

73 

74class NoDimensionsMetadataTestConfig( 

75 NoDimensionsTestConfig, pipelineConnections=NoDimensionsMetadataTestConnections 

76): 

77 """A config used for testing the metadata.""" 

78 

79 

80class NoDimensionsMetadataTestTask(NoDimensionsTestTask): 

81 """A simple pipeline task that can take a metadata as input.""" 

82 

83 ConfigClass = NoDimensionsMetadataTestConfig 

84 _DefaultName = "noDimensionsMetadataTest" 

85 

86 def run(self, input: dict[str, int], meta: dict[str, Any]) -> Struct: 

87 """Run the task, adding the configured key-value pair to the input 

88 argument and returning it as the output. 

89 

90 Parameters 

91 ---------- 

92 input : `dict` 

93 Dictionary to update and return. 

94 meta : `dict` 

95 Metadata to add. 

96 

97 Returns 

98 ------- 

99 result : `lsst.pipe.base.Struct` 

100 Struct with a single ``output`` attribute. 

101 """ 

102 self.log.info("Run metadata method given data of type: %s", get_full_type_name(input)) 

103 output = input.copy() 

104 output[self.config.key] = self.config.value 

105 

106 self.log.info("Received task metadata (%s): %s", get_full_type_name(meta), meta) 

107 

108 # Can change the return type via configuration. 

109 if "TaskMetadata" in self.config.outputSC: 

110 output = TaskMetadata.from_dict(output) 

111 elif type(output) == TaskMetadata: 

112 # Want the output to be a dict 

113 output = output.to_dict() 

114 self.log.info("Run method returns data of type: %s", get_full_type_name(output)) 

115 return Struct(output=output) 

116 

117 

118class SimplePipelineExecutorTests(lsst.utils.tests.TestCase): 

119 """Test the SimplePipelineExecutor API with a trivial task.""" 

120 

121 def setUp(self): 

122 self.path = tempfile.mkdtemp() 

123 # standalone parameter forces the returned config to also include 

124 # the information from the search paths. 

125 config = lsst.daf.butler.Butler.makeRepo( 

126 self.path, standalone=True, searchPaths=[os.path.join(TESTDIR, "config")] 

127 ) 

128 self.butler = SimplePipelineExecutor.prep_butler(config, [], "fake") 

129 self.butler.registry.registerDatasetType( 

130 lsst.daf.butler.DatasetType( 

131 "input", 

132 dimensions=self.butler.dimensions.empty, 

133 storageClass="StructuredDataDict", 

134 ) 

135 ) 

136 self.butler.put({"zero": 0}, "input") 

137 

138 def tearDown(self): 

139 shutil.rmtree(self.path, ignore_errors=True) 

140 

141 def test_from_task_class(self): 

142 """Test executing a single quantum with an executor created by the 

143 `from_task_class` factory method, and the 

144 `SimplePipelineExecutor.as_generator` method. 

145 """ 

146 executor = SimplePipelineExecutor.from_task_class(NoDimensionsTestTask, butler=self.butler) 

147 (quantum,) = executor.as_generator(register_dataset_types=True) 

148 self.assertEqual(self.butler.get("output"), {"zero": 0, "one": 1}) 

149 

150 def _configure_pipeline(self, config_a_cls, config_b_cls, storageClass_a=None, storageClass_b=None): 

151 """Configure a pipeline with from_pipeline.""" 

152 config_a = config_a_cls() 

153 config_a.connections.output = "intermediate" 

154 if storageClass_a: 

155 config_a.outputSC = storageClass_a 

156 config_b = config_b_cls() 

157 config_b.connections.input = "intermediate" 

158 if storageClass_b: 

159 config_b.outputSC = storageClass_b 

160 config_b.key = "two" 

161 config_b.value = 2 

162 task_defs = [ 

163 TaskDef(label="a", taskClass=NoDimensionsTestTask, config=config_a), 

164 TaskDef(label="b", taskClass=NoDimensionsTestTask, config=config_b), 

165 ] 

166 executor = SimplePipelineExecutor.from_pipeline(task_defs, butler=self.butler) 

167 return executor 

168 

169 def _test_logs(self, log_output, input_type_a, output_type_a, input_type_b, output_type_b): 

170 """Check the expected input types received by tasks A and B. 

171 

172 Note that these are the types as seen from the perspective of the task, 

173 so they must be consistent with the task's connections, but may not be 

174 consistent with the registry dataset types. 

175 """ 

176 all_logs = "\n".join(log_output) 

177 self.assertIn(f"lsst.a:Run method given data of type: {input_type_a}", all_logs) 

178 self.assertIn(f"lsst.b:Run method given data of type: {input_type_b}", all_logs) 

179 self.assertIn(f"lsst.a:Run method returns data of type: {output_type_a}", all_logs) 

180 self.assertIn(f"lsst.b:Run method returns data of type: {output_type_b}", all_logs) 

181 

182 def test_from_pipeline(self): 

183 """Test executing a two quanta from different configurations of the 

184 same task, with an executor created by the `from_pipeline` factory 

185 method, and the `SimplePipelineExecutor.run` method. 

186 """ 

187 executor = self._configure_pipeline( 

188 NoDimensionsTestTask.ConfigClass, NoDimensionsTestTask.ConfigClass 

189 ) 

190 

191 with self.assertLogs("lsst", level="INFO") as cm: 

192 quanta = executor.run(register_dataset_types=True, save_versions=False) 

193 self._test_logs(cm.output, "dict", "dict", "dict", "dict") 

194 

195 self.assertEqual(len(quanta), 2) 

196 self.assertEqual(self.butler.get("intermediate"), {"zero": 0, "one": 1}) 

197 self.assertEqual(self.butler.get("output"), {"zero": 0, "one": 1, "two": 2}) 

198 

199 def test_from_pipeline_intermediates_differ(self): 

200 """Run pipeline but intermediates definition in registry differs.""" 

201 # Pre-define the "intermediate" storage class to be something that is 

202 # like a dict but is not a dict. This will fail unless storage 

203 # class conversion is supported in put and get. 

204 self.butler.registry.registerDatasetType( 

205 lsst.daf.butler.DatasetType( 

206 "intermediate", 

207 dimensions=self.butler.dimensions.empty, 

208 storageClass="TaskMetadataLike", 

209 ) 

210 ) 

211 executor = self._configure_pipeline( 

212 NoDimensionsTestTask.ConfigClass, 

213 NoDimensionsTestTask.ConfigClass, 

214 storageClass_b="TaskMetadataLike", 

215 ) 

216 with self.assertLogs("lsst", level="INFO") as cm: 

217 quanta = executor.run(register_dataset_types=True, save_versions=False) 

218 # A dict is given to task a without change. 

219 # A returns a dict because it has not been told to do anything else. 

220 # That does not match the storage class so it will be converted 

221 # on put. 

222 # b is given a dict, because that's what its connection asks for. 

223 # b returns a TaskMetadata because that's how we configured it, and 

224 # since its output wasn't registered in advance, it will have been 

225 # registered as TaskMetadata and will now be received as TaskMetadata. 

226 self._test_logs(cm.output, "dict", "dict", "dict", "lsst.pipe.base.TaskMetadata") 

227 

228 self.assertEqual(len(quanta), 2) 

229 self.assertEqual(self.butler.get("intermediate"), TaskMetadata.from_dict({"zero": 0, "one": 1})) 

230 self.assertEqual(self.butler.get("output"), TaskMetadata.from_dict({"zero": 0, "one": 1, "two": 2})) 

231 

232 def test_from_pipeline_output_differ(self): 

233 """Run pipeline but output definition in registry differs.""" 

234 # Pre-define the "output" storage class to be something that is 

235 # like a dict but is not a dict. This will fail unless storage 

236 # class conversion is supported in put and get. 

237 self.butler.registry.registerDatasetType( 

238 lsst.daf.butler.DatasetType( 

239 "output", 

240 dimensions=self.butler.dimensions.empty, 

241 storageClass="TaskMetadataLike", 

242 ) 

243 ) 

244 executor = self._configure_pipeline( 

245 NoDimensionsTestTask.ConfigClass, 

246 NoDimensionsTestTask.ConfigClass, 

247 storageClass_a="TaskMetadataLike", 

248 ) 

249 with self.assertLogs("lsst", level="INFO") as cm: 

250 quanta = executor.run(register_dataset_types=True, save_versions=False) 

251 # a has been told to return a TaskMetadata but this will convert to 

252 # dict on read by b. 

253 # b returns a dict and that is converted to TaskMetadata on put. 

254 self._test_logs(cm.output, "dict", "lsst.pipe.base.TaskMetadata", "dict", "dict") 

255 

256 self.assertEqual(len(quanta), 2) 

257 self.assertEqual(self.butler.get("intermediate"), TaskMetadata.from_dict({"zero": 0, "one": 1})) 

258 self.assertEqual(self.butler.get("output"), TaskMetadata.from_dict({"zero": 0, "one": 1, "two": 2})) 

259 

260 def test_from_pipeline_input_differ(self): 

261 """Run pipeline but input definition in registry differs.""" 

262 # This config declares that the pipeline takes a TaskMetadata 

263 # as input but registry already thinks it has a StructureDataDict. 

264 executor = self._configure_pipeline(NoDimensionsTestConfig2, NoDimensionsTestTask.ConfigClass) 

265 

266 with self.assertLogs("lsst", level="INFO") as cm: 

267 quanta = executor.run(register_dataset_types=True, save_versions=False) 

268 self._test_logs(cm.output, "lsst.pipe.base.TaskMetadata", "dict", "dict", "dict") 

269 

270 self.assertEqual(len(quanta), 2) 

271 self.assertEqual(self.butler.get("intermediate"), {"zero": 0, "one": 1}) 

272 self.assertEqual(self.butler.get("output"), {"zero": 0, "one": 1, "two": 2}) 

273 

274 def test_from_pipeline_inconsistent_dataset_types(self): 

275 """Generate the QG (by initializing the executor), then register the 

276 dataset type with a different storage class than the QG should have 

277 predicted, to make sure execution fails as it should. 

278 """ 

279 executor = self._configure_pipeline( 

280 NoDimensionsTestTask.ConfigClass, NoDimensionsTestTask.ConfigClass 

281 ) 

282 

283 # Incompatible output dataset type. 

284 self.butler.registry.registerDatasetType( 

285 lsst.daf.butler.DatasetType( 

286 "output", 

287 dimensions=self.butler.dimensions.empty, 

288 storageClass="StructuredDataList", 

289 ) 

290 ) 

291 

292 with self.assertRaisesRegex( 

293 ValueError, "StructuredDataDict.*inconsistent with registry definition.*StructuredDataList" 

294 ): 

295 executor.run(register_dataset_types=True, save_versions=False) 

296 

297 def test_from_pipeline_metadata(self): 

298 """Test two tasks where the output uses metadata from input.""" 

299 # Must configure a special pipeline for this test. 

300 config_a = NoDimensionsTestTask.ConfigClass() 

301 config_a.connections.output = "intermediate" 

302 config_b = NoDimensionsMetadataTestTask.ConfigClass() 

303 config_b.connections.input = "intermediate" 

304 config_b.key = "two" 

305 config_b.value = 2 

306 task_defs = [ 

307 TaskDef(label="a", taskClass=NoDimensionsTestTask, config=config_a), 

308 TaskDef(label="b", taskClass=NoDimensionsMetadataTestTask, config=config_b), 

309 ] 

310 executor = SimplePipelineExecutor.from_pipeline(task_defs, butler=self.butler) 

311 

312 with self.assertLogs("test_simple_pipeline_executor", level="INFO") as cm: 

313 quanta = executor.run(register_dataset_types=True, save_versions=False) 

314 self.assertIn(f"Received task metadata ({get_full_type_name(dict)})", "".join(cm.output)) 

315 

316 self.assertEqual(len(quanta), 2) 

317 self.assertEqual(self.butler.get("intermediate"), {"zero": 0, "one": 1}) 

318 self.assertEqual(self.butler.get("output"), {"zero": 0, "one": 1, "two": 2}) 

319 

320 def test_from_pipeline_file(self): 

321 """Test executing a two quanta from different configurations of the 

322 same task, with an executor created by the `from_pipeline_filename` 

323 factory method, and the `SimplePipelineExecutor.run` method. 

324 """ 

325 filename = os.path.join(self.path, "pipeline.yaml") 

326 with open(filename, "w") as f: 

327 f.write( 

328 """ 

329 description: test 

330 tasks: 

331 a: 

332 class: "lsst.pipe.base.tests.no_dimensions.NoDimensionsTestTask" 

333 config: 

334 connections.output: "intermediate" 

335 b: 

336 class: "lsst.pipe.base.tests.no_dimensions.NoDimensionsTestTask" 

337 config: 

338 connections.input: "intermediate" 

339 key: "two" 

340 value: 2 

341 """ 

342 ) 

343 executor = SimplePipelineExecutor.from_pipeline_filename(filename, butler=self.butler) 

344 quanta = executor.run(register_dataset_types=True, save_versions=False) 

345 self.assertEqual(len(quanta), 2) 

346 self.assertEqual(self.butler.get("intermediate"), {"zero": 0, "one": 1}) 

347 self.assertEqual(self.butler.get("output"), {"zero": 0, "one": 1, "two": 2}) 

348 

349 

350class MemoryTester(lsst.utils.tests.MemoryTestCase): 

351 """Generic tests for file leaks.""" 

352 

353 

354def setup_module(module): 

355 """Set up the module for pytest. 

356 

357 Parameters 

358 ---------- 

359 module : `~types.ModuleType` 

360 Module to set up. 

361 """ 

362 lsst.utils.tests.init() 

363 

364 

365if __name__ == "__main__": 

366 lsst.utils.tests.init() 

367 unittest.main()