REF: dont consolidate in is_mixed_type (#36873)

Author: jbrockmendel

pandas/core/frame.py

@@ -631,7 +631,6 @@ def _is_homogeneous_type(self) -> bool:
         if self._mgr.any_extension_types:
             return len({block.dtype for block in self._mgr.blocks}) == 1
         else:
-            # Note: consolidates inplace
             return not self._is_mixed_type
 
     @property

pandas/core/generic.py

@@ -5472,8 +5472,15 @@ def _consolidate(self, inplace: bool_t = False):
 
     @property
     def _is_mixed_type(self) -> bool_t:
-        f = lambda: self._mgr.is_mixed_type
-        return self._protect_consolidate(f)
+        if len(self._mgr.blocks) == 1:
+            return False
+
+        if self._mgr.any_extension_types:
+            # Even if they have the same dtype, we cant consolidate them,
+            #  so we pretend this is "mixed'"
+            return True
+
+        return self.dtypes.nunique() > 1
 
     def _check_inplace_setting(self, value) -> bool_t:
         """ check whether we allow in-place setting with this type of value """
@@ -6253,8 +6260,7 @@ def fillna(
         axis = self._get_axis_number(axis)
 
         if value is None:
-
-            if self._is_mixed_type and axis == 1:
+            if len(self._mgr.blocks) > 1 and axis == 1:
                 if inplace:
                     raise NotImplementedError()
                 result = self.T.fillna(method=method, limit=limit).T

pandas/core/indexing.py

@@ -1535,8 +1535,7 @@ def _setitem_with_indexer(self, indexer, value):
         info_axis = self.obj._info_axis_number
 
         # maybe partial set
-        # _is_mixed_type has the side effect of consolidating in-place
-        take_split_path = self.obj._is_mixed_type
+        take_split_path = len(self.obj._mgr.blocks) > 1
 
         # if there is only one block/type, still have to take split path
         # unless the block is one-dimensional or it can hold the value

pandas/core/internals/managers.py

@@ -655,12 +655,6 @@ def _consolidate_check(self) -> None:
         self._is_consolidated = len(dtypes) == len(set(dtypes))
         self._known_consolidated = True
 
-    @property
-    def is_mixed_type(self) -> bool:
-        # Warning, consolidation needs to get checked upstairs
-        self._consolidate_inplace()
-        return len(self.blocks) > 1
-
     @property
     def is_numeric_mixed_type(self) -> bool:
         return all(block.is_numeric for block in self.blocks)

pandas/tests/frame/test_dtypes.py

@@ -230,18 +230,6 @@ def test_constructor_list_str_na(self, string_dtype):
     def test_is_homogeneous_type(self, data, expected):
         assert data._is_homogeneous_type is expected
 
-    def test_is_homogeneous_type_clears_cache(self):
-        ser = pd.Series([1, 2, 3])
-        df = ser.to_frame("A")
-        df["B"] = ser
-
-        assert len(df._mgr.blocks) == 2
-
-        a = df["B"]  # caches lookup
-        df._is_homogeneous_type  # _should_ clear cache
-        assert len(df._mgr.blocks) == 1
-        assert df["B"] is not a
-
     def test_asarray_homogenous(self):
         df = pd.DataFrame({"A": pd.Categorical([1, 2]), "B": pd.Categorical([1, 2])})
         result = np.asarray(df)

pandas/tests/internals/test_internals.py

@@ -272,13 +272,6 @@ def test_attrs(self):
         assert mgr.nblocks == 2
         assert len(mgr) == 6
 
-    def test_is_mixed_dtype(self):
-        assert not create_mgr("a,b:f8").is_mixed_type
-        assert not create_mgr("a:f8-1; b:f8-2").is_mixed_type
-
-        assert create_mgr("a,b:f8; c,d: f4").is_mixed_type
-        assert create_mgr("a,b:f8; c,d: object").is_mixed_type
-
     def test_duplicate_ref_loc_failure(self):
         tmp_mgr = create_mgr("a:bool; a: f8")