change tests and add new ones

tests/test_datasets/test_dataset.py

@@ -1495,7 +1495,7 @@ def test_to_dataset_not_modify_dataframe():
 @pytest.mark.parametrize("start_idx,end_idx", [(1, None), (None, 1), (1, 2), (1, -1)])
 def test_tsdataset_idx_slice(tsdf_with_exog, start_idx, end_idx):
     ts_slice = tsdf_with_exog.tsdataset_idx_slice(start_idx=start_idx, end_idx=end_idx)
-    assert ts_slice.known_future == tsdf_with_exog.known_future
+    assert ts_slice._known_future == tsdf_with_exog._known_future
     assert ts_slice.regressors == tsdf_with_exog.regressors
     pd.testing.assert_frame_equal(ts_slice._df, tsdf_with_exog._df.iloc[start_idx:end_idx])
     pd.testing.assert_frame_equal(ts_slice._df_exog, tsdf_with_exog._df_exog)
@@ -1944,7 +1944,7 @@ def test_create_from_misaligned_without_exog(df_name, freq, original_timestamp_n
     expected_df_exog = TSDataset.to_dataset(timestamp_df)
     pd.testing.assert_frame_equal(ts._df_exog, expected_df_exog)
 
-    assert original_timestamp_name in ts.known_future
+    assert original_timestamp_name in ts._known_future
     assert ts.freq is None
 
 
@@ -1996,7 +1996,7 @@ def test_create_from_misaligned_with_exog(
     pd.testing.assert_frame_equal(ts._df_exog, expected_df_exog)
 
     expected_known_future = sorted(set(known_future).union([original_timestamp_name]))
-    assert ts.known_future == expected_known_future
+    assert ts._known_future == expected_known_future
 
     assert ts.freq is None
 
@@ -2046,7 +2046,7 @@ def test_create_from_misaligned_with_exog_all(
     expected_df_exog = TSDataset.to_dataset(expected_df_exog)
     pd.testing.assert_frame_equal(ts._df_exog, expected_df_exog)
 
-    assert ts.known_future == expected_known_future
+    assert ts._known_future == expected_known_future
     assert ts.freq is None
 
 
@@ -2111,3 +2111,18 @@ def test_features(ts_name, expected_features, request):
     ts = request.getfixturevalue(ts_name)
     features = ts.features
     assert sorted(features) == sorted(expected_features)
+
+
+def test_error_set_read_only_known_future(df_and_regressors):
+    df, df_exog, regressors = df_and_regressors
+    ts = TSDataset(df=df, freq="D", df_exog=df_exog, known_future=regressors)
+    with pytest.raises(AttributeError, match="can't set attribute"):
+        ts.known_future = ["exog_1"]
+
+
+def test_not_equal_updated_known_futures(df_and_regressors):
+    df, df_exog, regressors = df_and_regressors
+    ts = TSDataset(df=df, freq="D", df_exog=df_exog, known_future=regressors)
+    new_known_future = ts.known_future
+    new_known_future.append("new_regressor")
+    assert ts.known_future != new_known_future

tests/test_models/test_inference/test_forecast.py

@@ -1164,7 +1164,7 @@ def _test_forecast_mixed_in_out_sample(ts, model, transforms, num_skip_points=50
         future_ts = ts.make_future(future_steps=future_prediction_size)
         future_df = future_ts.to_pandas().loc[:, pd.IndexSlice[:, "target"]]
         df_full = pd.concat((df, future_df))
-        forecast_full_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts.known_future)
+        forecast_full_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts._known_future)
         forecast_full_ts.transform(transforms)
         forecast_full_ts._df = forecast_full_ts._df.iloc[(num_skip_points - model.context_size) :]
         full_prediction_size = len(forecast_full_ts.timestamps) - model.context_size

tests/test_models/test_inference/test_predict.py

@@ -834,15 +834,15 @@ def _test_predict_mixed_in_out_sample(ts, model, transforms, num_skip_points=50,
 
         # predicting mixed in-sample and out-sample
         df_full = pd.concat((train_df, future_df))
-        forecast_full_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts.known_future)
+        forecast_full_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts._known_future)
         forecast_full_ts.transform(transforms)
         forecast_full_ts._df = forecast_full_ts._df.iloc[(num_skip_points - model.context_size) :]
         full_prediction_size = len(forecast_full_ts.timestamps) - model.context_size
         forecast_full_ts = make_predict(model=model, ts=forecast_full_ts, prediction_size=full_prediction_size)
 
         # predicting only in sample
         forecast_in_sample_ts = TSDataset(
-            df=train_df, df_exog=train_ts._df_exog, freq=ts.freq, known_future=ts.known_future
+            df=train_df, df_exog=train_ts._df_exog, freq=ts.freq, known_future=ts._known_future
         )
         forecast_in_sample_ts.transform(transforms)
         to_skip = num_skip_points - model.context_size
@@ -853,7 +853,7 @@ def _test_predict_mixed_in_out_sample(ts, model, transforms, num_skip_points=50,
         )
 
         # predicting only out sample
-        forecast_out_sample_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts.known_future)
+        forecast_out_sample_ts = TSDataset(df=df_full, df_exog=ts._df_exog, freq=ts.freq, known_future=ts._known_future)
         forecast_out_sample_ts.transform(transforms)
         to_remain = model.context_size + future_prediction_size
         forecast_out_sample_ts._df = forecast_out_sample_ts._df.iloc[-to_remain:]

tests/test_reconciliation/test_base.py

@@ -116,7 +116,7 @@ def test_reconcile(hierarchical_ts, source_level, target_level, mapping_matrix,
 
     assert obtained_ts.freq == expected_ts.freq
     assert obtained_ts.current_df_level == expected_ts.current_df_level
-    assert obtained_ts.known_future == expected_ts.known_future
+    assert obtained_ts._known_future == expected_ts._known_future
     assert obtained_ts.regressors == expected_ts.regressors
     pd.testing.assert_frame_equal(obtained_ts._df, expected_ts._df)
     pd.testing.assert_frame_equal(obtained_ts._df_exog, expected_ts._df_exog)
@@ -138,7 +138,7 @@ def test_reconcile_with_target_components(
 
     assert obtained_ts.freq == expected_ts.freq
     assert obtained_ts.current_df_level == expected_ts.current_df_level
-    assert obtained_ts.known_future == expected_ts.known_future
+    assert obtained_ts._known_future == expected_ts._known_future
     assert obtained_ts.regressors == expected_ts.regressors
     pd.testing.assert_frame_equal(obtained_ts.get_target_components(), expected_ts.get_target_components())
     pd.testing.assert_frame_equal(obtained_ts._df, expected_ts._df)

tests/test_transforms/test_timestamp/test_fourier_transform.py

@@ -343,7 +343,7 @@ def test_transform_values_with_shift(shift, in_column, ts_name, period, mod, req
     ts = request.getfixturevalue(ts_name)
 
     ts_1 = ts
-    ts_2 = TSDataset(df=ts._raw_df.iloc[shift:], df_exog=ts._df_exog, freq=ts.freq, known_future=ts.known_future)
+    ts_2 = TSDataset(df=ts._raw_df.iloc[shift:], df_exog=ts._df_exog, freq=ts.freq, known_future=ts._known_future)
     transform = FourierTransform(period=period, mods=[mod], out_column="regressor_fourier", in_column=in_column)
 
     transform.fit(ts)

tests/utils.py

@@ -30,7 +30,7 @@ def select_segments_subset(ts: TSDataset, segments: List[str]) -> TSDataset:
     df_exog = ts._df_exog
     if df_exog is not None:
         df_exog = df_exog.loc[:, pd.IndexSlice[segments, :]].copy()
-    known_future = ts.known_future
+    known_future = ts._known_future
     freq = ts.freq
     subset_ts = TSDataset(df=df, df_exog=df_exog, known_future=known_future, freq=freq)
     return subset_ts
@@ -49,7 +49,7 @@ def convert_ts_to_int_timestamp(ts: TSDataset, shift=0):
     ts = TSDataset(
         df=df,
         df_exog=df_exog,
-        known_future=ts.known_future,
+        known_future=ts._known_future,
         freq=None,
         hierarchical_structure=ts.hierarchical_structure,
     )
@@ -72,7 +72,7 @@ def convert_ts_index_to_freq(ts: TSDataset, freq: str, shift=0):
     ts = TSDataset(
         df=df,
         df_exog=df_exog,
-        known_future=ts.known_future,
+        known_future=ts._known_future,
         freq=freq,
         hierarchical_structure=ts.hierarchical_structure,
     )