fix(histogram): fix numeric-like enum counting by using dict lookup

exaflow/algorithms/federated/statistics/histogram.py

@@ -85,7 +85,9 @@ def hist(
         )
 
     def _value_counts(self, series: pd.Series, categories) -> List[int]:
-        counts = series.value_counts()
+        # Use a plain dict to avoid pandas Series.get positional fallback on
+        # integer-like keys (e.g. key=0 on index ["1"] returning first element).
+        counts = series.value_counts().to_dict()
         resolved = []
         for cat in categories:
             count = counts.get(cat, 0)
@@ -94,7 +96,7 @@ def _value_counts(self, series: pd.Series, categories) -> List[int]:
             resolved.append(int(count))
         return resolved
 
-    def _value_counts_numeric_fallback(self, counts, cat: str) -> int:
+    def _value_counts_numeric_fallback(self, counts: Dict, cat: str) -> int:
         for caster in (float, int):
             try:
                 coerced = caster(cat)

tests/standalone_tests/federated_algorithms/statistics/test_histogram.py

@@ -296,3 +296,31 @@ def test_federated_algorithm_with_multiple_workers(self, case):
             y_levels=y_levels,
             min_row_count=1,
         )
+
+    @pytest.mark.parametrize("n_workers", [1, 3])
+    def test_numeric_like_enum_codes_are_counted_by_label(self, n_workers):
+        y_levels = ["0", "1", "9"]
+        df = pd.DataFrame(
+            {
+                "y": ["1", "1", "1", "1", "1", "1"],
+                "group": ["A", "A", "B", "B", "C", "C"],
+            }
+        )
+        metadata = {
+            "y": {
+                "is_categorical": True,
+                "enumerations": {"0": "No", "1": "Yes", "9": "Unknown"},
+            },
+            "group": {
+                "is_categorical": True,
+                "enumerations": {"A": "A", "B": "B", "C": "C"},
+            },
+        }
+        self.run_comparison(
+            X=df,
+            y=np.zeros((df.shape[0],), dtype=float),
+            n_workers=n_workers,
+            metadata=metadata,
+            y_levels=y_levels,
+            min_row_count=1,
+        )