def test_solver_honors_set_time_step(self): # Given dt = 0.1 tf = 1.0 pfreq = 1 solver = Solver(integrator=self.integrator, tf=tf, dt=dt, adaptive_timestep=False) solver.set_print_freq(pfreq) solver.acceleration_evals = [self.a_eval] solver.particles = [] record = [] def _mock_dump_output(): # Record the time at which the solver dumped anything record.append(solver.t) solver.dump_output = mock.Mock(side_effect=_mock_dump_output) # When solver.set_time_step(0.2) solver.solve(show_progress=False) # Then expected = np.asarray([0.0, 0.2, 0.4, 0.6, 0.8, 1.0]) error_message = "Expected %s, got %s" % (expected, record) self.assertEqual(len(expected), len(record), error_message) self.assertTrue( np.max(np.abs(expected - record)) < 1e-12, error_message)
def test_solver_dumps_output_given_output_at_times_landing_exactly(self): # Given dt = 0.1 self.integrator.compute_time_step.return_value = dt tf = 3.0 pfreq = 10 output_at_times = np.concatenate( (np.arange(0, 1.21, 0.2), np.arange(1.25, 1.51, 0.05))) solver = Solver(integrator=self.integrator, tf=tf, dt=dt, output_at_times=output_at_times) solver.set_print_freq(pfreq) solver.acceleration_evals = [self.a_eval] solver.particles = [] # When record = [] record_dt = [] def _mock_dump_output(): # Record the time at which the solver dumped anything record.append(solver.t) # This smells but ... sd = solver._get_solver_data() record_dt.append(sd['dt']) solver.dump_output = mock.Mock(side_effect=_mock_dump_output) solver.solve(show_progress=False) # Then expected = np.concatenate( (np.arange(0, 1.21, 0.2), np.arange(1.25, 1.49, 0.05), [1.5, 1.7, 2.7, 3.0])) error_message = "Expected %s, got %s" % (expected, record) self.assertEqual(len(expected), len(record), error_message) self.assertTrue( np.max(np.abs(expected - record)) < 1e-12, error_message) self.assertEqual(33, solver.count) # The final timestep should not be a tiny one due to roundoff. self.assertTrue(solver.dt > 0.1 * 0.25) npt.assert_array_almost_equal([0.1] * len(record_dt), record_dt, decimal=12)
def test_solver_dumps_output_given_output_at_times(self): # Given dt = 0.1 self.integrator.compute_time_step.return_value = dt tf = 10.05 pfreq = 5 output_at_times = [0.3, 0.35] solver = Solver(integrator=self.integrator, tf=tf, dt=dt, output_at_times=output_at_times) solver.set_print_freq(pfreq) solver.acceleration_eval = self.a_eval solver.particles = [] # When record = [] record_dt = [] def _mock_dump_output(): # Record the time at which the solver dumped anything record.append(solver.t) # This smells but ... sd = solver._get_solver_data() record_dt.append(sd['dt']) solver.dump_output = mock.Mock(side_effect=_mock_dump_output) solver.solve(show_progress=False) # Then expected = np.asarray([0.0, 0.3, 0.35] + np.arange(0.45, 10.1, 0.5).tolist() + [10.05]) error_message = "Expected %s, got %s" % (expected, record) self.assertEqual(len(expected), len(record), error_message) self.assertTrue( np.max(np.abs(expected - record)) < 1e-12, error_message) self.assertEqual(101, solver.count) # The final timestep should not be a tiny one due to roundoff. self.assertTrue(solver.dt > 0.1 * 0.25) npt.assert_array_almost_equal([0.1] * len(record_dt), record_dt, decimal=12)