Replace setting columns with join in SegmentEncoder

CHANGELOG.md

@@ -14,6 +14,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Make test_ts optional in plot_forecast ([#321](https://github.com/tinkoff-ai/etna/pull/321))
 - Speed up inference for multisegment regression models ([#333](https://github.com/tinkoff-ai/etna/pull/333))
 - Speed up Pipeline._get_backtest_forecasts ([#336](https://github.com/tinkoff-ai/etna/pull/336))
+- Speed up SegmentEncoderTransform ([#331](https://github.com/tinkoff-ai/etna/pull/331))
 
 ## [1.3.3] - 2021-11-24
 ### Added

etna/transforms/segment_encoder.py

@@ -54,11 +54,7 @@ def transform(self, df: pd.DataFrame) -> pd.DataFrame:
             index=df.index,
         )
         encoded_df = encoded_df.astype("category")
-
-        for segment in set(df.columns.get_level_values("segment")):
-            df.loc[self.idx[:], self.idx[segment, "regressor_segment_code"]] = encoded_df.loc[
-                self.idx[:], self.idx[segment, "regressor_segment_code"]
-            ]
+        df = df.join(encoded_df)
         df = df.sort_index(axis=1)
         return df
 

tests/test_transforms/test_segment_encoder_transform.py

@@ -26,13 +26,20 @@ def dummy_df() -> pd.DataFrame:
     return df
 
 
-def test_dummy(dummy_df):
+def test_segment_encoder_transform(dummy_df):
     transform = SegmentEncoderTransform()
     transformed_df = transform.fit_transform(dummy_df)
     assert (
         len(transformed_df.loc[:, pd.IndexSlice[:, "regressor_segment_code"]].columns) == 2
     ), "Number of columns not the same as segments"
     assert len(dummy_df) == len(transformed_df), "Row missing"
+    codes = set()
+    for segment in dummy_df.columns.get_level_values("segment").unique():
+        column = transformed_df.loc[:, pd.IndexSlice[segment, "regressor_segment_code"]]
+        assert column.dtype == "category", "Column type is not category"
+        assert np.all(column == column.iloc[0]), "Values are not the same for the whole column"
+        codes.add(column.iloc[0])
+    assert codes == {0, 1}, "Codes are not 0 and 1"
 
 
 @pytest.fixture