fix: resolve pytest, pandas, sklearn warnings and rust compiler dead code warnings

Author: Secbone

src/merge_generic.rs

@@ -72,11 +72,13 @@ fn should_break(
 }
 
 /// Helper function to fill NaN values for floating point types
+#[allow(dead_code)]
 fn fill_nan_f64(arr: &Array1<f64>, fill_value: f64) -> Array1<f64> {
     arr.mapv(|x| if x.is_nan() { fill_value } else { x })
 }
 
 /// Helper function to fill NaN values for integer types (no-op)
+#[allow(dead_code)]
 fn fill_nan_int<T: Num + Copy>(arr: &Array1<T>, _fill_value: T) -> Array1<T> {
     arr.clone() // Integers don't have NaN, so just return a copy
 }

toad/preprocessing/process.py

@@ -129,7 +129,10 @@ def process(self, data):
                     if col == _ALL_SYMBOL_:
                         col = None
 
-                    r = g.apply(f, col = col)
+                    try:
+                        r = g.apply(f, col = col, include_groups = False)
+                    except TypeError:
+                        r = g.apply(f, col = col)
 
                 if isinstance(r, pd.Series):
                     r = pd.DataFrame(r)

toad/stats.py

@@ -421,8 +421,8 @@ def feature_bin_stats(df_bin,feature,target):
     table['total'] = table['good'] + table['bad']
     table['badrate'] = table['bad'] / table['total']
     table['prop'] = table['total'] / table['total'].sum()
-    table['y_prop'] = table['good'] / table['good'].sum()
-    table['n_prop'] = table['bad'] / table['bad'].sum()
+    table['y_prop'] = table['bad'] / table['bad'].sum()
+    table['n_prop'] = table['good'] / table['good'].sum()
     table['woe'] = table.apply(lambda x : WOE(x['y_prop'], x['n_prop']),axis=1)
     table['iv'] = table.apply(lambda x : (x['y_prop'] - x['n_prop']) * WOE(x['y_prop'], x['n_prop']), axis=1)
     return table

toad/stats_test.py

@@ -62,8 +62,8 @@ def test_quality():
     assert result.loc['feature', 'unique'] == 500
 
 def test_quality_iv_only():
-    result = quality(df, 'target', iv_only = True)
-    assert np.isnan(result.loc['feature', 'gini'])
+    result = quality(df, 'target', indicators = ['iv'])
+    assert 'gini' not in result.columns
 
 def test_quality_with_merge():
     result = quality(df, 'target', n_bins = 5, method = 'chi')

toad/transform.py

@@ -385,7 +385,7 @@ def fit_(self, X, y, **kwargs):
         gbdt = GradientBoostingClassifier(**kwargs)
         gbdt.fit(X, y)
 
-        X = gbdt.apply(X)
+        X = gbdt.apply(X.values if hasattr(X, 'values') else X)
         X = X.reshape(-1, X.shape[1])
 
         onehot = OneHotEncoder().fit(X)
@@ -405,7 +405,7 @@ def transform_(self, rules, X):
         Returns:
             array-like
         """
-        X = rules['gbdt'].apply(X)
+        X = rules['gbdt'].apply(X.values if hasattr(X, 'values') else X)
         X = X.reshape(-1, X.shape[1])
         res = rules['onehot'].transform(X).toarray()
         return res

toad/utils/progress/progress_test.py

@@ -2,7 +2,7 @@
 from .progress import Progress
 
 
-class TestIterator:
+class MockIterator:
     def __init__(self, size):
         self._size = size
 
@@ -22,7 +22,7 @@ def test_progress_size():
     assert p.size == 9527
 
 def test_iterator():
-    ti = TestIterator(100)
+    ti = MockIterator(100)
     p = Progress(ti)
     for i in p:
         sleep(0.01)