Coverage for sparkle/CLI/generate_report.py: 88%

1#!/usr/bin/env python3 

2"""Sparkle command to generate a report for an executed experiment.""" 

3 

4import sys 

5import argparse 

6from pathlib import Path, PurePath 

7 

8from sparkle.CLI.help import global_variables as gv 

9from sparkle.solver.ablation import AblationScenario 

10from sparkle.platform import generate_report_for_selection as sgfs 

11from sparkle.platform import \ 

12 generate_report_for_configuration as sgrfch 

13from sparkle.CLI.help import logging as sl 

14from sparkle.platform.settings_objects import Settings, SettingState 

15from sparkle.CLI.help import argparse_custom as ac 

16from sparkle.CLI.help.reporting_scenario import Scenario 

17from sparkle.platform import \ 

18 generate_report_for_parallel_portfolio as sgrfpph 

19from sparkle.solver import Solver 

20from sparkle.solver.validator import Validator 

21from sparkle.instance import Instance_Set 

22from sparkle.structures import PerformanceDataFrame, FeatureDataFrame 

23from sparkle.platform.output.configuration_output import ConfigurationOutput 

24from sparkle.platform.output.selection_output import SelectionOutput 

25from sparkle.platform.output.parallel_portfolio_output import ParallelPortfolioOutput 

26 

27from sparkle.platform import CommandName, COMMAND_DEPENDENCIES 

28from sparkle.CLI.initialise import check_for_initialise 

29from sparkle.CLI.help.nicknames import resolve_object_name 

30 

31 

32def parser_function() -> argparse.ArgumentParser: 

33 """Define the command line arguments.""" 

34 parser = argparse.ArgumentParser( 

35 description="Without any arguments a report for the most recent algorithm " 

36 "selection or algorithm configuration procedure is generated.", 

37 epilog="Note that if a test instance set is given, the training instance set " 

38 "must also be given.") 

39 # Configuration arguments 

40 parser.add_argument(*ac.SolverReportArgument.names, 

41 **ac.SolverReportArgument.kwargs) 

42 parser.add_argument(*ac.InstanceSetTrainReportArgument.names, 

43 **ac.InstanceSetTrainReportArgument.kwargs) 

44 parser.add_argument(*ac.InstanceSetTestReportArgument.names, 

45 **ac.InstanceSetTestReportArgument.kwargs) 

46 parser.add_argument(*ac.NoAblationReportArgument.names, 

47 **ac.NoAblationReportArgument.kwargs) 

48 # Selection arguments 

49 parser.add_argument(*ac.SelectionReportArgument.names, 

50 **ac.SelectionReportArgument.kwargs) 

51 parser.add_argument(*ac.TestCaseDirectoryArgument.names, 

52 **ac.TestCaseDirectoryArgument.kwargs) 

53 # Common arguments 

54 parser.add_argument(*ac.SparkleObjectiveArgument.names, 

55 **ac.SparkleObjectiveArgument.kwargs) 

56 parser.add_argument(*ac.SettingsFileArgument.names, 

57 **ac.SettingsFileArgument.kwargs) 

58 parser.add_argument(*ac.GenerateJSONArgument.names, 

59 **ac.GenerateJSONArgument.kwargs) 

60 return parser 

61 

62 

63def main(argv: list[str]) -> None: 

64 """Generate a report for an executed experiment.""" 

65 prev_settings = Settings(PurePath("Settings/latest.ini")) 

66 

67 # Log command call 

68 sl.log_command(sys.argv) 

69 

70 # Define command line arguments 

71 parser = parser_function() 

72 

73 # Process command line arguments 

74 args = parser.parse_args(argv) 

75 

76 check_for_initialise(COMMAND_DEPENDENCIES[CommandName.GENERATE_REPORT]) 

77 

78 # Do first, so other command line options can override settings from the file 

79 if ac.set_by_user(args, "settings_file"): 

80 gv.settings().read_settings_ini( 

81 args.settings_file, SettingState.CMD_LINE 

82 ) 

83 if args.objectives is not None: 

84 gv.settings().set_general_sparkle_objectives( 

85 args.objectives, SettingState.CMD_LINE) 

86 selection = args.selection 

87 test_case_dir = args.test_case_directory 

88 only_json = args.only_json 

89 

90 solver = resolve_object_name(args.solver, 

91 gv.solver_nickname_mapping, 

92 gv.settings().DEFAULT_solver_dir, Solver) 

93 instance_set_train = resolve_object_name( 

94 args.instance_set_train, 

95 gv.file_storage_data_mapping[gv.instances_nickname_path], 

96 gv.settings().DEFAULT_instance_dir, Instance_Set) 

97 instance_set_test = resolve_object_name( 

98 args.instance_set_train, 

99 gv.file_storage_data_mapping[gv.instances_nickname_path], 

100 gv.settings().DEFAULT_instance_dir, Instance_Set) 

101 

102 Settings.check_settings_changes(gv.settings(), prev_settings) 

103 # If no arguments are set get the latest scenario 

104 if not selection and test_case_dir is None and solver is None: 

105 scenario = gv.latest_scenario().get_latest_scenario() 

106 if scenario == Scenario.SELECTION: 

107 selection = True 

108 test_case_dir = gv.latest_scenario().get_selection_test_case_directory() 

109 elif scenario == Scenario.CONFIGURATION: 

110 solver = gv.latest_scenario().get_config_solver() 

111 instance_set_train = gv.latest_scenario().get_config_instance_set_train() 

112 instance_set_test = gv.latest_scenario().get_config_instance_set_test() 

113 elif scenario == Scenario.PARALLEL_PORTFOLIO: 

114 parallel_portfolio_path = gv.latest_scenario().get_parallel_portfolio_path() 

115 pap_instance_set =\ 

116 gv.latest_scenario().get_parallel_portfolio_instance_set() 

117 

118 flag_instance_set_train = instance_set_train is not None 

119 flag_instance_set_test = instance_set_test is not None 

120 

121 # Reporting for algorithm selection 

122 if selection or test_case_dir is not None: 

123 objective = gv.settings().get_general_sparkle_objectives()[0] 

124 if not objective.time: 

125 print("ERROR: The selection report is not implemented for " 

126 " non-runtime objectives!") 

127 sys.exit(-1) 

128 selection_scenario = gv.latest_scenario().get_selection_scenario_path() 

129 actual_portfolio_selector_path = selection_scenario / "portfolio_selector" 

130 if not actual_portfolio_selector_path.is_file(): 

131 print("Before generating a Sparkle report, please first construct the " 

132 "Sparkle portfolio selector. Not generating a Sparkle report, stopping" 

133 " execution!") 

134 sys.exit(-1) 

135 

136 print("Generating report for selection...") 

137 train_data = PerformanceDataFrame(gv.settings().DEFAULT_performance_data_path) 

138 feature_data = FeatureDataFrame(gv.settings().DEFAULT_feature_data_path) 

139 test_data = None 

140 test_case_path = Path(test_case_dir) if test_case_dir is not None else None 

141 if test_case_dir is not None and (test_case_path 

142 / "performance_data.csv").exists(): 

143 test_data = PerformanceDataFrame(test_case_path / "performance_data.csv") 

144 # Create machine readable selection output 

145 instance_dirs = set(Path(instance).parent for instance in train_data.instances) 

146 instance_sets = [] 

147 for dir in instance_dirs: 

148 instance_sets.append(Instance_Set(dir)) 

149 test_set = None if test_case_dir is None else Instance_Set(Path(test_case_dir)) 

150 cutoff_time = gv.settings().get_general_target_cutoff_time() 

151 output = gv.settings().DEFAULT_selection_output_analysis 

152 selection_output = SelectionOutput( 

153 selection_scenario, train_data, feature_data, 

154 instance_sets, test_set, objective, cutoff_time, 

155 output) 

156 selection_output.write_output() 

157 print("Machine readable output is placed at: ", selection_output.output) 

158 

159 if not only_json: 

160 sgfs.generate_report_selection( 

161 gv.settings().DEFAULT_selection_output_analysis, 

162 gv.settings().DEFAULT_latex_source, 

163 "template-Sparkle-for-selection.tex", 

164 gv.settings().DEFAULT_latex_bib, 

165 gv.settings().DEFAULT_extractor_dir, 

166 selection_scenario, 

167 feature_data, 

168 train_data, 

169 objective, 

170 gv.settings().get_general_extractor_cutoff_time(), 

171 gv.settings().get_general_target_cutoff_time(), 

172 test_data 

173 ) 

174 if test_case_dir is None: 

175 print("Report generated ...") 

176 else: 

177 print("Report for test generated ...") 

178 

179 elif gv.latest_scenario().get_latest_scenario() == Scenario.PARALLEL_PORTFOLIO: 

180 # Reporting for parallel portfolio 

181 # Machine readable Output 

182 cutoff_time = gv.settings().get_general_target_cutoff_time() 

183 objective = gv.settings().get_general_sparkle_objectives()[0] 

184 output = gv.settings().DEFAULT_parallel_portfolio_output_analysis 

185 parallel_portfolio_output = ParallelPortfolioOutput(parallel_portfolio_path, 

186 pap_instance_set, 

187 objective, 

188 output) 

189 parallel_portfolio_output.write_output() 

190 print("Machine readable output is placed at: ", parallel_portfolio_output.output) 

191 

192 if not only_json: 

193 sgrfpph.generate_report_parallel_portfolio( 

194 parallel_portfolio_path, 

195 gv.settings().DEFAULT_parallel_portfolio_output_analysis, 

196 gv.settings().DEFAULT_latex_source, 

197 gv.settings().DEFAULT_latex_bib, 

198 gv.settings().get_general_sparkle_objectives()[0], 

199 gv.settings().get_general_target_cutoff_time(), 

200 pap_instance_set) 

201 print("Parallel portfolio report generated ...") 

202 else: 

203 # Reporting for algorithm configuration 

204 if solver is None: 

205 print("Error! No Solver found for configuration report generation.") 

206 sys.exit(-1) 

207 

208 # If only the testing set is given return an error 

209 if not flag_instance_set_train and flag_instance_set_test: 

210 print("Argument Error! Only a testing set was provided, please also " 

211 "provide a training set") 

212 print(f"Usage: {sys.argv[0]} --solver <solver> [--instance-set-train " 

213 "<instance-set-train>] [--instance-set-test <instance-set-test>]") 

214 sys.exit(-1) 

215 # Generate a report depending on which instance sets are provided 

216 if flag_instance_set_train or flag_instance_set_test: 

217 # Check if there are result to generate a report from 

218 validator = Validator(gv.settings().DEFAULT_validation_output) 

219 train_res = validator.get_validation_results( 

220 solver, instance_set_train) 

221 if instance_set_test is not None: 

222 test_res = validator.get_validation_results(solver, 

223 instance_set_test) 

224 if len(train_res) == 0 or (instance_set_test is not None 

225 and len(test_res) == 0): 

226 print("Error: Results not found for the given solver and instance set(s)" 

227 ' combination. Make sure the "configure_solver" and "validate_' 

228 'configured_vs_default" commands were correctly executed. ') 

229 sys.exit(-1) 

230 else: 

231 print("Error: No results from validate_configured_vs_default found that " 

232 "can be used in the report!") 

233 sys.exit(-1) 

234 # Extract config scenario data for report, but this should be read from the 

235 # scenario file instead as we can't know wether features were used or not now 

236 configurator = gv.settings().get_general_sparkle_configurator() 

237 config_scenario = gv.latest_scenario().get_configuration_scenario( 

238 configurator.scenario_class) 

239 ablation_scenario = None 

240 if args.flag_ablation: 

241 ablation_scenario = AblationScenario( 

242 solver, instance_set_train, instance_set_test, 

243 gv.settings().DEFAULT_ablation_output) 

244 

245 # Create machine readable output 

246 output = gv.settings().DEFAULT_configuration_output_analysis 

247 config_output = ConfigurationOutput(config_scenario.directory, 

248 configurator, 

249 config_scenario, 

250 instance_set_test, 

251 output) 

252 config_output.write_output() 

253 print("Machine readable output is placed at: ", config_output.output) 

254 

255 if not only_json: 

256 sgrfch.generate_report_for_configuration( 

257 solver, 

258 gv.settings().get_general_sparkle_configurator(), 

259 Validator(gv.settings().DEFAULT_validation_output), 

260 gv.settings().DEFAULT_extractor_dir, 

261 gv.settings().DEFAULT_configuration_output_analysis, 

262 gv.settings().DEFAULT_latex_source, 

263 gv.settings().DEFAULT_latex_bib, 

264 gv.settings().get_general_extractor_cutoff_time(), 

265 config_scenario, 

266 instance_set_test, 

267 ablation=ablation_scenario 

268 ) 

269 

270 # Write used settings to file 

271 gv.settings().write_used_settings() 

272 sys.exit(0) 

273 

274 

275if __name__ == "__main__": 

276 main(sys.argv[1:])