TST: fix groupby-empty xfails (pandas-dev#44092)

Author: jbrockmendel

pandas/core/groupby/generic.py

@@ -1536,6 +1536,7 @@ def func(df):
             result = [index[i] if i >= 0 else np.nan for i in indices]
             return df._constructor_sliced(result, index=res.index)
 
+        func.__name__ = "idxmax"
         return self._python_apply_general(func, self._obj_with_exclusions)
 
     @Appender(DataFrame.idxmin.__doc__)
@@ -1557,6 +1558,7 @@ def func(df):
             result = [index[i] if i >= 0 else np.nan for i in indices]
             return df._constructor_sliced(result, index=res.index)
 
+        func.__name__ = "idxmin"
         return self._python_apply_general(func, self._obj_with_exclusions)
 
     boxplot = boxplot_frame_groupby

pandas/core/groupby/ops.py

@@ -753,6 +753,18 @@ def apply(
                 mutated = True
             result_values.append(res)
 
+        # getattr pattern for __name__ is needed for functools.partial objects
+        if len(group_keys) == 0 and getattr(f, "__name__", None) not in [
+            "idxmin",
+            "idxmax",
+            "nanargmin",
+            "nanargmax",
+        ]:
+            # If group_keys is empty, then no function calls have been made,
+            #  so we will not have raised even if this is an invalid dtype.
+            #  So do one dummy call here to raise appropriate TypeError.
+            f(data.iloc[:0])
+
         return result_values, mutated
 
     @cache_readonly

pandas/tests/groupby/test_groupby.py

@@ -24,6 +24,11 @@
     to_datetime,
 )
 import pandas._testing as tm
+from pandas.core.arrays import (
+    BooleanArray,
+    FloatingArray,
+    IntegerArray,
+)
 from pandas.core.base import SpecificationError
 import pandas.core.common as com
 
@@ -1822,17 +1827,23 @@ def test_pivot_table_values_key_error():
 )
 @pytest.mark.filterwarnings("ignore:Dropping invalid columns:FutureWarning")
 @pytest.mark.filterwarnings("ignore:.*Select only valid:FutureWarning")
-def test_empty_groupby(columns, keys, values, method, op, request):
+def test_empty_groupby(columns, keys, values, method, op, request, using_array_manager):
     # GH8093 & GH26411
     override_dtype = None
 
     if (
         isinstance(values, Categorical)
         and not isinstance(columns, list)
-        and op in ["sum", "prod"]
+        and op in ["sum", "prod", "skew", "mad"]
     ):
         # handled below GH#41291
-        pass
+
+        if using_array_manager and op == "mad":
+            right_msg = "Cannot interpret 'CategoricalDtype.* as a data type"
+            msg = "Regex pattern \"'Categorical' does not implement.*" + right_msg
+            mark = pytest.mark.xfail(raises=AssertionError, match=msg)
+            request.node.add_marker(mark)
+
     elif (
         isinstance(values, Categorical)
         and len(keys) == 1
@@ -1851,11 +1862,7 @@ def test_empty_groupby(columns, keys, values, method, op, request):
             raises=TypeError, match="'Categorical' does not implement"
         )
         request.node.add_marker(mark)
-    elif (
-        isinstance(values, Categorical)
-        and len(keys) == 1
-        and op in ["mad", "min", "max", "sum", "prod", "skew"]
-    ):
+    elif isinstance(values, Categorical) and len(keys) == 1 and op in ["sum", "prod"]:
         mark = pytest.mark.xfail(
             raises=AssertionError, match="(DataFrame|Series) are different"
         )
@@ -1869,7 +1876,30 @@ def test_empty_groupby(columns, keys, values, method, op, request):
             raises=AssertionError, match="(DataFrame|Series) are different"
         )
         request.node.add_marker(mark)
-    elif isinstance(values, pd.core.arrays.BooleanArray) and op in ["sum", "prod"]:
+    elif (
+        isinstance(values, (IntegerArray, FloatingArray))
+        and op == "mad"
+        and isinstance(columns, list)
+    ):
+        mark = pytest.mark.xfail(
+            raises=TypeError, match="can only perform ops with numeric values"
+        )
+        request.node.add_marker(mark)
+
+    elif (
+        op == "mad"
+        and not isinstance(columns, list)
+        and isinstance(values, pd.DatetimeIndex)
+        and values.tz is not None
+        and using_array_manager
+    ):
+        mark = pytest.mark.xfail(
+            raises=TypeError,
+            match=r"Cannot interpret 'datetime64\[ns, US/Eastern\]' as a data type",
+        )
+        request.node.add_marker(mark)
+
+    elif isinstance(values, BooleanArray) and op in ["sum", "prod"]:
         # We expect to get Int64 back for these
         override_dtype = "Int64"
 
@@ -1895,19 +1925,29 @@ def get_result():
 
     if columns == "C":
         # i.e. SeriesGroupBy
-        if op in ["prod", "sum"]:
+        if op in ["prod", "sum", "skew"]:
             # ops that require more than just ordered-ness
             if df.dtypes[0].kind == "M":
                 # GH#41291
                 # datetime64 -> prod and sum are invalid
-                msg = "datetime64 type does not support"
+                if op == "skew":
+                    msg = "'DatetimeArray' does not implement reduction 'skew'"
+                else:
+                    msg = "datetime64 type does not support"
                 with pytest.raises(TypeError, match=msg):
                     get_result()
 
                 return
-            elif isinstance(values, Categorical):
+        if op in ["prod", "sum", "skew", "mad"]:
+            if isinstance(values, Categorical):
                 # GH#41291
-                msg = "category type does not support"
+                if op == "mad":
+                    # mad calls mean, which Categorical doesn't implement
+                    msg = "'Categorical' does not implement reduction 'mean'"
+                elif op == "skew":
+                    msg = f"'Categorical' does not implement reduction '{op}'"
+                else:
+                    msg = "category type does not support"
                 with pytest.raises(TypeError, match=msg):
                     get_result()
 
@@ -1954,6 +1994,34 @@ def get_result():
                 tm.assert_equal(result, expected)
                 return
 
+        if (
+            op in ["mad", "min", "max", "skew"]
+            and isinstance(values, Categorical)
+            and len(keys) == 1
+        ):
+            # Categorical doesn't implement, so with numeric_only=True
+            #  these are dropped and we get an empty DataFrame back
+            result = get_result()
+            expected = df.set_index(keys)[[]]
+
+            # with numeric_only=True, these are dropped, and we get
+            # an empty DataFrame back
+            if len(keys) != 1:
+                # Categorical is special without 'observed=True'
+                lev = Categorical([0], dtype=values.dtype)
+                mi = MultiIndex.from_product([lev, lev], names=keys)
+                expected = DataFrame([], columns=[], index=mi)
+            else:
+                # all columns are dropped, but we end up with one row
+                # Categorical is special without 'observed=True'
+                lev = Categorical([0], dtype=values.dtype)
+                ci = Index(lev, name=keys[0])
+                expected = DataFrame([], columns=[], index=ci)
+            # expected = df.set_index(keys)[columns]
+
+            tm.assert_equal(result, expected)
+            return
+
     result = get_result()
     expected = df.set_index(keys)[columns]
     if override_dtype is not None: