Handle empty filtered aggregate results

 Return 0 for sum() on empty filtered views.
 Return NaN for mean() and std() on empty filtered views.
 Keep min()/max() raising because they have no identity.
 Add tests for empty filtered aggregate behavior.

Author: FrancescAlted

src/blosc2/ctable.py

@@ -1317,6 +1317,8 @@ def _sum_lazy_fastpath(self, acc_dtype, where=None):
     def sum(self, dtype=None, *, where=None):
         """Sum of all live, non-null values.
 
+        Returns zero for an empty column or filtered view.
+
         Supported dtypes: bool, int, uint, float, complex.
         Bool values are counted as 0 / 1.
         Null sentinel values are skipped.
@@ -1351,8 +1353,6 @@ def sum(self, dtype=None, *, where=None):
         """
         self._require_kind("biufc", "sum")
         where = self._normalize_sum_where(where)
-        if where is None:
-            self._require_nonempty("sum")
         # Use a wide accumulator to reduce overflow risk
         acc_dtype = np.dtype(dtype).type if dtype is not None else None
         if acc_dtype is None:
@@ -1478,7 +1478,9 @@ def mean(self, *, where=None) -> float:
         """
         self._require_kind("biuf", "mean")
         where = self._normalize_sum_where(where)
-        if where is None:
+        if where is None and len(self) == 0:
+            if self._table.base is not None:
+                return float("nan")
             self._require_nonempty("mean")
         fast = self._lazy_aggregate_fastpath("mean", where=where)
         if fast is not NotImplemented:
@@ -1512,7 +1514,9 @@ def std(self, ddof: int = 0, *, where=None) -> float:
         """
         self._require_kind("biuf", "std")
         where = self._normalize_sum_where(where)
-        if where is None:
+        if where is None and len(self) == 0:
+            if self._table.base is not None:
+                return float("nan")
             self._require_nonempty("std")
         fast = self._lazy_aggregate_fastpath("std", where=where, ddof=ddof)
         if fast is not NotImplemented:

tests/ctable/test_column.py

@@ -374,10 +374,14 @@ def test_sum_skips_deleted_rows():
     assert t["id"].sum() == sum(range(1, 20))
 
 
-def test_sum_empty_raises():
+def test_sum_empty_returns_zero():
     t = CTable(Row)
-    with pytest.raises(ValueError, match="empty"):
-        t["id"].sum()
+    assert t["id"].sum() == 0
+
+
+def test_sum_empty_filtered_view_returns_zero():
+    t = CTable(Row, new_data=DATA20)
+    assert t[t.id < 0]["id"].sum() == 0
 
 
 def test_sum_wrong_type_raises():
@@ -466,6 +470,11 @@ def test_mean_empty_raises():
         t["id"].mean()
 
 
+def test_mean_empty_filtered_view_is_nan():
+    t = CTable(Row, new_data=DATA20)
+    assert np.isnan(t[t.id < 0]["id"].mean())
+
+
 # -------------------------------------------------------------------
 # Aggregates: std
 # -------------------------------------------------------------------
@@ -499,6 +508,11 @@ def test_std_empty_raises():
         t["id"].std()
 
 
+def test_std_empty_filtered_view_is_nan():
+    t = CTable(Row, new_data=DATA20)
+    assert np.isnan(t[t.id < 0]["id"].std())
+
+
 # -------------------------------------------------------------------
 # Aggregates: any / all
 # -------------------------------------------------------------------