Fix problem bad return value

daisypy/optim/problem.py

@@ -93,6 +93,7 @@ def __init__(
         self.debug = debug
 
     def __call__(self, parameter_values):
+        # TODO: Rewrite to accept a dict of parameters. This is too brittle
         """Run Daisy with the given parameters and evaluate the objective. The return value depends
         on
 
@@ -129,5 +130,5 @@ def _run(self, output_directory, named_parameters):
         sim_result = self.runner(dai_file, output_directory)
         if sim_result.returncode != 0:
             print(sim_result)
-            return np.nan
+            return { self.objective_fn.name : np.nan }
         return self.objective_fn(output_directory)

tests/mockup.py

@@ -0,0 +1,41 @@
+# pylint: disable=too-few-public-methods
+from subprocess import CompletedProcess
+from daisypy.optim.file_generator import FileGenerator
+
+class MockFileGenerator(FileGenerator):
+    '''Mock file generator that always generates the paths it was constructed with'''
+    def __init__(self, paths):
+        self.paths = paths
+
+    def __call__(self, output_directory, params, tagged=True):
+        return self.paths
+
+
+class MockRunner:
+    '''Mock runner always returning a CompletedProcess with a specified returncode'''
+    def __init__(self, args=None, returncode=0):
+        self.args = args if args is not None else []
+        self.returncode = returncode
+
+    def __call__(self, dai_file, output_directory):
+        return CompletedProcess(self.args, self.returncode)
+
+class MockObjective:
+    '''Mock objective always returning a specific value'''
+    def __init__(self, name='mock', value=0):
+        self.name = name
+        self.value = value
+
+    def __call__(self, daisy_output_directory):
+        return { self.name : self.value }
+
+
+class MockProblem:
+    '''Problem case for test purposes. Will evaluate an objective by forwarding parameters'''
+    def __init__(self, parameters, objective_fn):
+        self.parameters = parameters
+        self.objective_fn = objective_fn
+
+    def __call__(self, parameter_values):
+        named_parameters = { p.name : value for p, value in zip(self.parameters, parameter_values) }
+        return { 'mock' : self.objective_fn(**named_parameters) }

tests/test_cma_optimizer.py

@@ -5,7 +5,7 @@
     DefaultLogger,
     DaisyCMAOptimizer,
 )
-from .mock_problem import MockProblem
+from .mockup import MockProblem
 from .test_objectives import beale_function
 
 def test_cma_optimizer():

tests/test_problem.py

@@ -0,0 +1,49 @@
+import numpy as np
+from daisypy.optim import DaisyOptimizationProblem, ContinuousParameter, MultiObjective
+from .mockup import (MockRunner, MockFileGenerator, MockObjective)
+
+
+def test_runner_succeds(tmp_path):
+    '''Test that the return value is as expected when the runner succeds'''
+    file_generator = MockFileGenerator({'dai' : ''})
+    runner = MockRunner()
+    parameters = { 'dai' : [ContinuousParameter('p', 0, (-1, 1))] }
+    out_dir = tmp_path
+    objective = MockObjective('mock', 123)
+
+    problem = DaisyOptimizationProblem(
+        runner, file_generator, objective, parameters, out_dir
+    )
+    result = problem([-1])
+    assert result['mock'] == objective.value
+
+
+def test_runner_fails(tmp_path):
+    '''Test that the return value is nan when the runner fails'''
+    file_generator = MockFileGenerator({'dai' : ''})
+    runner = MockRunner(returncode=1)
+    parameters = { 'dai' : [ContinuousParameter('p', 0, (-1, 1))] }
+    out_dir = tmp_path
+    objective = MockObjective('mock', 123)
+
+    problem = DaisyOptimizationProblem(
+        runner, file_generator, objective, parameters, out_dir
+    )
+    result = problem([-1])
+    assert np.isnan(result['mock'])
+
+def test_multi_objective(tmp_path):
+    '''Test that the return value is as expected for multiple objectives'''
+    file_generator = MockFileGenerator({'dai' : ''})
+    runner = MockRunner()
+    parameters = { 'dai' : [ContinuousParameter('p', 0, (-1, 1))] }
+    out_dir = tmp_path
+    objectives = [ MockObjective(f'mock-{i}', i*123) for i in range(3) ]
+    objective = MultiObjective('multi', objectives)
+
+    problem = DaisyOptimizationProblem(
+        runner, file_generator, objective, parameters, out_dir
+    )
+    result = problem([0])
+    for obj in objectives:
+        assert result[obj.name] == obj.value

tests/test_sequential_optimizer.py

@@ -6,7 +6,7 @@
     DefaultLogger,
     DaisySequentialOptimizer,
 )
-from .mock_problem import MockProblem
+from .mockup import MockProblem
 
 class Objective:
     # pylint: disable=too-few-public-methods