Use explicit exceptions for rake validation

Author: neuralsorcerer

balance/weighting_methods/rake.py

@@ -206,30 +206,30 @@ def rake(
         result["weight"].tolist()
         # [1.0, 1.0]
     """
-    assert (
-        "weight" not in sample_df.columns.values
-    ), "weight shouldn't be a name for covariate in the sample data"
-    assert (
-        "weight" not in target_df.columns.values
-    ), "weight shouldn't be a name for covariate in the target data"
+    if "weight" in sample_df.columns.values:
+        raise ValueError("weight shouldn't be a name for covariate in the sample data")
+    if "weight" in target_df.columns.values:
+        raise ValueError("weight shouldn't be a name for covariate in the target data")
 
     # TODO: move the input checks into separate funnction for rake, ipw, poststratify
-    assert isinstance(sample_df, pd.DataFrame), "sample_df must be a pandas DataFrame"
-    assert isinstance(target_df, pd.DataFrame), "target_df must be a pandas DataFrame"
-    assert isinstance(
-        sample_weights, pd.Series
-    ), "sample_weights must be a pandas Series"
-    assert isinstance(
-        target_weights, pd.Series
-    ), "target_weights must be a pandas Series"
-    assert sample_df.shape[0] == sample_weights.shape[0], (
-        "sample_weights must be the same length as sample_df"
-        f"{sample_df.shape[0]}, {sample_weights.shape[0]}"
-    )
-    assert target_df.shape[0] == target_weights.shape[0], (
-        "target_weights must be the same length as target_df"
-        f"{target_df.shape[0]}, {target_weights.shape[0]}"
-    )
+    if not isinstance(sample_df, pd.DataFrame):
+        raise TypeError("sample_df must be a pandas DataFrame")
+    if not isinstance(target_df, pd.DataFrame):
+        raise TypeError("target_df must be a pandas DataFrame")
+    if not isinstance(sample_weights, pd.Series):
+        raise TypeError("sample_weights must be a pandas Series")
+    if not isinstance(target_weights, pd.Series):
+        raise TypeError("target_weights must be a pandas Series")
+    if sample_df.shape[0] != sample_weights.shape[0]:
+        raise ValueError(
+            "sample_weights must be the same length as sample_df"
+            f"{sample_df.shape[0]}, {sample_weights.shape[0]}"
+        )
+    if target_df.shape[0] != target_weights.shape[0]:
+        raise ValueError(
+            "target_weights must be the same length as target_df"
+            f"{target_df.shape[0]}, {target_weights.shape[0]}"
+        )
     if not isinstance(store_fit_metadata, bool):
         raise TypeError("`store_fit_metadata` must be a bool.")
     if store_fit_metadata and transformations == "default":
@@ -337,10 +337,11 @@ def rake(
             )
         if len(target_over_set):
             if len(alphabetized_variables) == 1:
-                missing_mask = target_df[variable].isin(target_over_set)
-                missing_level_target_weight = target_weights.loc[
-                    target_df.index[missing_mask]
-                ].sum()
+                missing_mask = target_df[variable].isin(target_over_set).to_numpy()
+                missing_indices = target_df.index[missing_mask]
+                missing_level_target_weight = float(
+                    target_weights.loc[missing_indices].sum()
+                )
                 if missing_level_target_weight > 0:
                     raise ValueError(
                         "Single-variable rake requires that all target levels are "

tests/test_rake.py

@@ -67,7 +67,7 @@ def _assert_rake_raises_with_message(
             **kwargs: Additional arguments to pass to rake()
         """
         self.assertRaisesRegex(
-            AssertionError,
+            Exception,
             expected_message,
             rake,
             sample_df,