add more pytest on the completely_random_rotation; #59

tests/test_random_state_reproducibility.py

@@ -39,3 +39,34 @@ def test_random_state_reproducibility():
         assert np.sum(ensemble_df1['stixel_checklist_count'].values -
                 ensemble_df2['stixel_checklist_count'].values) != 0
 
+
+def test_random_state_reproducibility_completely_random_rotation_angle():
+    model = make_AdaSTEMRegressor()
+    model = model.set_params(random_state=42, completely_random_rotation=True)
+    model.split(X, verbosity=1)
+    ensemble_df1 = model.ensemble_df.copy()
+
+    model = model.set_params(random_state=990324, completely_random_rotation=True)
+    model.split(X, verbosity=1)
+    ensemble_df2 = model.ensemble_df.copy()
+
+    model = model.set_params(random_state=42, completely_random_rotation=True)
+    model.split(X, verbosity=1)
+    ensemble_df3 = model.ensemble_df.copy()
+
+    # 1 and 3 (with the same random state)
+    assert ensemble_df1.shape == ensemble_df3.shape
+    assert np.allclose(ensemble_df1['calibration_point_x_jitter'].values,
+                ensemble_df3['calibration_point_x_jitter'].values)
+    assert np.allclose(ensemble_df1['stixel_checklist_count'].values,
+                ensemble_df3['stixel_checklist_count'].values)
+
+    # 1 and 2 (with the different random state)
+    if ensemble_df1.shape != ensemble_df2.shape:
+        pass
+    else:
+        assert np.sum(ensemble_df1['calibration_point_x_jitter'].values -
+                ensemble_df2['calibration_point_x_jitter'].values) != 0
+        assert np.sum(ensemble_df1['stixel_checklist_count'].values -
+                ensemble_df2['stixel_checklist_count'].values) != 0
+