BUG: Concat multiple different ExtensionArray types (pandas-dev#22997)

Author: TomAugspurger

doc/source/whatsnew/v0.24.0.txt

@@ -627,6 +627,7 @@ update the ``ExtensionDtype._metadata`` tuple to match the signature of your
 - :meth:`Series.astype` and :meth:`DataFrame.astype` now dispatch to :meth:`ExtensionArray.astype` (:issue:`21185:`).
 - Slicing a single row of a ``DataFrame`` with multiple ExtensionArrays of the same type now preserves the dtype, rather than coercing to object (:issue:`22784`)
 - Added :meth:`pandas.api.types.register_extension_dtype` to register an extension type with pandas (:issue:`22664`)
+- Bug when concatenating multiple ``Series`` with different extension dtypes not casting to object dtype (:issue:`22994`)
 - Series backed by an ``ExtensionArray`` now work with :func:`util.hash_pandas_object` (:issue:`23066`)
 - Updated the ``.type`` attribute for ``PeriodDtype``, ``DatetimeTZDtype``, and ``IntervalDtype`` to be instances of the dtype (``Period``, ``Timestamp``, and ``Interval`` respectively) (:issue:`22938`)
 - :func:`ExtensionArray.isna` is allowed to return an ``ExtensionArray`` (:issue:`22325`).

pandas/core/dtypes/concat.py

@@ -560,11 +560,6 @@ def _concat_sparse(to_concat, axis=0, typs=None):
 
     fill_values = [x.fill_value for x in to_concat
                    if isinstance(x, SparseArray)]
-
-    if len(set(fill_values)) > 1:
-        raise ValueError("Cannot concatenate SparseArrays with different "
-                         "fill values")
-
     fill_value = fill_values[0]
 
     # TODO: Fix join unit generation so we aren't passed this.

pandas/core/internals/managers.py

@@ -1636,8 +1636,7 @@ def concat(self, to_concat, new_axis):
         # check if all series are of the same block type:
         if len(non_empties) > 0:
             blocks = [obj.blocks[0] for obj in non_empties]
-
-            if all(type(b) is type(blocks[0]) for b in blocks[1:]):  # noqa
+            if len({b.dtype for b in blocks}) == 1:
                 new_block = blocks[0].concat_same_type(blocks)
             else:
                 values = [x.values for x in blocks]

pandas/tests/extension/decimal/test_decimal.py

@@ -100,7 +100,9 @@ def test_hashable(self, dtype):
 
 
 class TestInterface(BaseDecimal, base.BaseInterfaceTests):
-    pass
+
+    pytestmark = pytest.mark.skipif(compat.PY2,
+                                    reason="Unhashble dtype in Py2.")
 
 
 class TestConstructors(BaseDecimal, base.BaseConstructorsTests):
@@ -112,7 +114,8 @@ def test_from_dtype(self, data):
 
 
 class TestReshaping(BaseDecimal, base.BaseReshapingTests):
-    pass
+    pytestmark = pytest.mark.skipif(compat.PY2,
+                                    reason="Unhashble dtype in Py2.")
 
 
 class TestGetitem(BaseDecimal, base.BaseGetitemTests):
@@ -174,7 +177,8 @@ class TestCasting(BaseDecimal, base.BaseCastingTests):
 
 
 class TestGroupby(BaseDecimal, base.BaseGroupbyTests):
-    pass
+    pytestmark = pytest.mark.skipif(compat.PY2,
+                                    reason="Unhashble dtype in Py2.")
 
 
 class TestSetitem(BaseDecimal, base.BaseSetitemTests):

pandas/tests/reshape/test_concat.py

@@ -1,24 +1,25 @@
 from warnings import catch_warnings, simplefilter
 from itertools import combinations
 from collections import deque
+from decimal import Decimal
 
 import datetime as dt
 import dateutil
 import numpy as np
 from numpy.random import randn
 
 from datetime import datetime
-from pandas.compat import StringIO, iteritems, PY2
+from pandas.compat import Iterable, StringIO, iteritems, PY2
 import pandas as pd
 from pandas import (DataFrame, concat,
                     read_csv, isna, Series, date_range,
                     Index, Panel, MultiIndex, Timestamp,
                     DatetimeIndex, Categorical)
-from pandas.compat import Iterable
 from pandas.core.dtypes.dtypes import CategoricalDtype
 from pandas.util import testing as tm
 from pandas.util.testing import (assert_frame_equal,
                                  makeCustomDataframe as mkdf)
+from pandas.tests.extension.decimal import to_decimal
 
 import pytest
 
@@ -2361,6 +2362,18 @@ def test_concat_datetime_timezone(self):
                                 index=idx1.append(idx1))
         tm.assert_frame_equal(result, expected)
 
+    @pytest.mark.skipif(PY2, reason="Unhashable Decimal dtype")
+    def test_concat_different_extension_dtypes_upcasts(self):
+        a = pd.Series(pd.core.arrays.integer_array([1, 2]))
+        b = pd.Series(to_decimal([1, 2]))
+
+        result = pd.concat([a, b], ignore_index=True)
+        expected = pd.Series([
+            1, 2,
+            Decimal(1), Decimal(2)
+        ], dtype=object)
+        tm.assert_series_equal(result, expected)
+
 
 @pytest.mark.parametrize('pdt', [pd.Series, pd.DataFrame, pd.Panel])
 @pytest.mark.parametrize('dt', np.sctypes['float'])