fix bug #GH23020

Author: makbigc

doc/source/whatsnew/v0.24.0.rst

@@ -1564,6 +1564,12 @@ Sparse
 - Bug in :meth:`SparseArray.nonzero` and :meth:`SparseDataFrame.dropna` returning shifted/incorrect results (:issue:`21172`)
 - Bug in :meth:`DataFrame.apply` where dtypes would lose sparseness (:issue:`23744`)
 
+IntegerArray
+^^^^^^^^^^^^
+
+- Bug in :func:`pandas.merge` when merging on an Integer extension array (:issue:`23020`)
+
+
 Build Changes
 ^^^^^^^^^^^^^
 

pandas/core/reshape/merge.py

@@ -19,7 +19,7 @@
     is_bool_dtype, is_categorical_dtype, is_datetime64_dtype,
     is_datetime64tz_dtype, is_datetimelike, is_dtype_equal, is_float_dtype,
     is_int64_dtype, is_integer, is_integer_dtype, is_list_like, is_number,
-    is_numeric_dtype, needs_i8_conversion)
+    is_numeric_dtype, needs_i8_conversion, is_extension_array_dtype)
 from pandas.core.dtypes.missing import isnull, na_value_for_dtype
 
 from pandas import Categorical, DataFrame, Index, MultiIndex, Series, Timedelta
@@ -1605,6 +1605,10 @@ def _factorize_keys(lk, rk, sort=True):
 
         lk = ensure_int64(lk.codes)
         rk = ensure_int64(rk)
+    elif is_extension_array_dtype(lk) and is_extension_array_dtype(rk):
+        klass = libhashtable.Factorizer
+        lk = ensure_object(lk)
+        rk = ensure_object(rk)
     elif is_integer_dtype(lk) and is_integer_dtype(rk):
         # GH#23917 TODO: needs tests for case where lk is integer-dtype
         #  and rk is datetime-dtype

pandas/tests/reshape/merge/test_merge.py

@@ -1326,6 +1326,16 @@ def test_merging_with_bool_or_int_cateorical_column(self, category_column,
             CDT(categories, ordered=ordered))
         assert_frame_equal(expected, result)
 
+    def test_merge_on_int_array(self):
+        # GH  23020
+        df = pd.DataFrame({'A': pd.Series([1, 2, np.nan], dtype='Int64'),
+                           'B': 1})
+        result = pd.merge(df, df, on='A')
+        expected = pd.DataFrame({'A': pd.Series([1, 2, np.nan], dtype='Int64'),
+                                 'B_x': 1,
+                                 'B_y': 1})
+        assert_frame_equal(result, expected, check_dtype=True)
+
 
 @pytest.fixture
 def left_df():