Fix duplicated column in feature type detector. Fix docs

bluecast/eda/analyse.py

@@ -206,7 +206,7 @@ def univariate_plots(df: pd.DataFrame) -> None:
 
     Expects numeric columns only.
     """
-    for col in df.columns:
+    for col in df.columns.to_list():
         plt.figure(figsize=(8, 4))
 
         # Histogram
@@ -661,7 +661,7 @@ def plot_null_percentage(dataframe: pd.DataFrame) -> None:
 
 
 def check_unique_values(
-    df: pd.DataFrame, columns: List[Union[str, int, float]], threshold: float
+    df: pd.DataFrame, columns: List[Union[str, int, float]], threshold: float = 0.9
 ) -> List[Union[str, int, float]]:
     """
     Check if the columns have an amount of unique values that is almost the number of total rows (being above the defined threshold)

bluecast/preprocessing/feature_types.py

@@ -152,7 +152,8 @@ def identify_num_columns(self, df: pd.DataFrame) -> pd.DataFrame:
             for vartype in self.num_dtypes:
                 num_cols = df.select_dtypes(include=[vartype]).columns
                 for col in num_cols:
-                    num_col_list.append(col)
+                    if col not in num_col_list:
+                        num_col_list.append(col)
 
         for col in df.columns.to_list():
             max_length = df[col].astype(str).str.len().max()
@@ -161,14 +162,16 @@ def identify_num_columns(self, df: pd.DataFrame) -> pd.DataFrame:
                     if self.check_if_column_is_float_from_string(df[col]):
                         df[col] = df[col].astype(float)
                         self.detected_col_types[col] = "float"
-                        num_col_list.append(col)
+                        if col not in num_col_list:
+                            num_col_list.append(col)
                     elif (
                         self.check_if_column_is_int_from_string(df[col])
                         and df[col].nunique() > 2
                     ):
                         df[col] = df[col].astype("Int64")
                         self.detected_col_types[col] = "Int64"
-                        num_col_list.append(col)
+                        if col not in num_col_list:
+                            num_col_list.append(col)
                 except Exception:
                     pass
             self.num_columns = num_col_list
@@ -248,13 +251,15 @@ def cast_rest_columns_to_object(
             if col in self.cat_columns:
                 df[col] = df[col].astype(str)
                 self.detected_col_types[col] = "object"
-                cat_columns.append(col)
+                if col not in cat_columns:
+                    cat_columns.append(col)
             if col in self.num_columns:
                 pass
             else:
                 df[col] = df[col].astype(str)
                 self.detected_col_types[col] = "object"
-                cat_columns.append(col)
+                if col not in cat_columns:
+                    cat_columns.append(col)
         self.cat_columns = cat_columns
         return df
 

docs/source/EDA.md

@@ -58,7 +58,7 @@ from bluecast.eda.analyse import (
     plot_tsne,
     check_unique_values,
     plot_null_percentage,
-    mutual_info_to_target.
+    mutual_info_to_target,
     plot_pie_chart,
 )
 
@@ -69,7 +69,7 @@ feat_type_detector = FeatureTypeDetector()
 train_data = feat_type_detector.fit_transform_feature_types(train_data)
 
 # detect columns with a very high share of unique values
-many_unique_cols = check_unique_values(train_data, feat_type_detector.cat_columns)
+many_unique_cols = check_unique_values(train_data, train_data.columns.to_list())
 ```
 
 ## Pie chart