Merge pull request #59 from UBC-MDS/column_scaler_improvements

Column scaler improvements

src/datamop/column_scaler.py

@@ -1,4 +1,4 @@
-# Formula in this function is adapted from Scikit Learn Documentation
+# Formula in this function is adapted from Scikit Learn documentation
 # https://scikit-learn.org/1.5/modules/generated/sklearn.preprocessing.MinMaxScaler.html
 # https://scikit-learn.org/1.6/modules/generated/sklearn.preprocessing.StandardScaler.html
 
@@ -33,8 +33,11 @@ def column_scaler(data, column, method="minmax", new_min=0, new_max=1, inplace=T
     Returns
     --------
     pandas.DataFrame
-        A copy of the DataFrame with the scaled column replacing the original column if `inplace` is set to `True`.
-        If `inplace` is set to `False`, the copy of DataFrame is returned with the new scaled column added, keeping the original column.
+        A copy of the DataFrame with the scaled column 
+        replacing the original column if `inplace` is set to `True`.
+        If `inplace` is set to `False`, 
+        the copy of DataFrame is returned with the new scaled column added, 
+        keeping the original column.
 
     Raises
     ------
@@ -47,7 +50,6 @@ def column_scaler(data, column, method="minmax", new_min=0, new_max=1, inplace=T
         If the `method` is not `minmax` or `standard`.
         If the `new_min` value is greater or equal to the `new_max` when using `minmax` method.
     
-
     Examples
     --------
     >>> import pandas as pd
@@ -58,15 +60,17 @@ def column_scaler(data, column, method="minmax", new_min=0, new_max=1, inplace=T
             0.0
             0.5
             1.0
-
     """
     # Check input is pd.DataFrame
     if not isinstance(data, pd.DataFrame):
         raise TypeError("Input must be a pandas DataFrame.")
 
     # Empty df warning
     if data.empty:
-        warnings.warn("Empty DataFrame detected. Empty DataFrame will be returned.", UserWarning)
+        warnings.warn(
+            "Empty DataFrame detected. Empty DataFrame will be returned.", 
+            UserWarning
+            )
         return data.copy()
 
     # Error handling
@@ -79,12 +83,28 @@ def column_scaler(data, column, method="minmax", new_min=0, new_max=1, inplace=T
 
     # Edge case warning
     if data[column].isna().any():
-        warnings.warn("NaN value detected in column '{column}'. They will be unchanged", UserWarning)
+        warnings.warn(
+            f"NaN value detected in column '{column}'. They will be unchanged", 
+            UserWarning
+            )
 
     if data[column].nunique() == 1:
-        warnings.warn("Single-value column detected. All values will be scaled to the midpoint of the `new_min` and `new_max`.", UserWarning)
-        midpoint = (new_min + new_max) / 2
-        scaled_column = pd.Series([midpoint] * len(data), index=data.index)
+        if method == "minmax":
+            warnings.warn(
+                "Single-value column detected. "
+                "All values will be scaled to the midpoint of the `new_min` and `new_max`.",
+                  UserWarning
+                  )
+            midpoint = (new_min + new_max) / 2
+            scaled_column = pd.Series([midpoint] * len(data), index=data.index)
+
+        elif method == "standard":
+            warnings.warn(
+                "Standard deviation is zero. "
+                "All values are set to 0 to prevent division by zero.",
+                UserWarning
+                ) 
+            scaled_column = pd.Series([0] * len(data), index=data.index) 
 
     # Scale the column
     else:
@@ -94,7 +114,12 @@ def column_scaler(data, column, method="minmax", new_min=0, new_max=1, inplace=T
                 raise ValueError("`new_min` cannot be greater than `new_max`.")
             min_value = data[column].min()
             max_value = data[column].max()
-            scaled_column = ((data[column] - min_value) / (max_value - min_value)) * (new_max - new_min) + new_min
+            scaled_column = (
+                ((data[column] - min_value) / (max_value - min_value)) 
+            * (new_max - new_min) 
+            + new_min
+            )
+
         # standard scaling
         elif method == "standard":
             mean_value = data[column].mean() 

tests/test_column_scaler.py

@@ -9,6 +9,11 @@ def one_column_df():
     """Return DataFrame with one column of numeric values. Used for testing."""
     return pd.DataFrame({"price": [25, 50, 75]})
 
+@pytest.fixture
+def one_column_df_float():
+    """Return DataFrame with one column of floating values. Used for testing."""
+    return pd.DataFrame({"price": [25.0, 50.0, 75.0]})
+
 @pytest.fixture
 def single_val_df():
     """Return DataFrame with one column with single repeated value. Used for testing."""
@@ -26,11 +31,17 @@ def non_numeric_df():
 
 # Expected use case tests
 def test_minmax_scaling_default(one_column_df):
-    """Test min-max scaling with default new_min=0 and new_max=1."""
+    """Test min-max scaling with default new_min=0 and new_max=1. Use float values."""
     scaled_df = column_scaler(one_column_df, column="price", method="minmax")
     expected = [0.0, 0.5, 1.0]
     assert scaled_df["price"].tolist() == expected
 
+def test_minmax_scaling_default_float(one_column_df_float):
+    """Test min-max scaling with default new_min=0 and new_max=1."""
+    scaled_df = column_scaler(one_column_df_float, column="price", method="minmax")
+    expected = [0.0, 0.5, 1.0]
+    assert scaled_df["price"].tolist() == expected
+
 def test_minmax_scaling_custom(one_column_df):
     """Test min-max scaling with custom new_min=10 and new_max=20."""
     scaled_df = column_scaler(one_column_df, column="price", method="minmax", new_min=10, new_max=20)
@@ -58,6 +69,15 @@ def test_single_value_column_minmax(single_val_df):
     expected = [15.0, 15.0, 15.0]
     assert scaled_df["price"].tolist() == expected
 
+def test_single_value_column_standard(single_val_df):
+    """Test standard scaling with column with single repeated values to prevent division by zero."""
+    with pytest.warns(UserWarning, 
+                      match="Standard deviation is zero"):
+        scaled_df = column_scaler(single_val_df, column="price", method="standard")
+
+    expected = [0, 0, 0]
+    assert scaled_df["price"].tolist() == expected
+
 def test_empty_dataframe(empty_df):
     """Test scaling on empty DataFrame."""
     with pytest.warns(UserWarning, match="Empty DataFrame detected"):
@@ -73,6 +93,7 @@ def test_column_with_nan():
     expected = [0.0, np.nan, 1.0]
     assert np.allclose(scaled_df["price"], expected, equal_nan=True)
 
+
 # Erroneous case tests
 
 def test_non_numeric_column(non_numeric_df):