BUG: Fixed merging on tz-aware

Closes pandas-dev#25014
TomAugspurger · Jan 30, 2019 · 58f2892 · 58f2892
1 parent b9f2e2b
commit 58f2892
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 0 deletions.
diff --git a/doc/source/whatsnew/v0.24.1.rst b/doc/source/whatsnew/v0.24.1.rst
@@ -23,6 +23,7 @@ Fixed Regressions
 - Bug in :meth:`DataFrame.itertuples` with ``records`` orient raising an ``AttributeError`` when the ``DataFrame`` contained more than 255 columns (:issue:`24939`)
 - Bug in :meth:`DataFrame.itertuples` orient converting integer column names to strings prepended with an underscore (:issue:`24940`)
 - Fixed regression in :class:`Index.intersection` incorrectly sorting the values by default (:issue:`24959`).
+- Fixed regression in :func:`merge` when merging an empty ``DataFrame`` with multiple timezone-aware columns on one of the timezone-aware columns (:issue:`25014`).
 
 .. _whatsnew_0241.enhancements:
 

diff --git a/pandas/core/internals/concat.py b/pandas/core/internals/concat.py
@@ -219,6 +219,11 @@ def get_reindexed_values(self, empty_dtype, upcasted_na):
 
         else:
             for ax, indexer in self.indexers.items():
+                # GH-25014: get_upcasted_na returns iNaT, but
+                # DatetimeArray.take expects NaT.
+                # TODO: update get_empty_dtype_and_na to use EAs earlier?
+                if is_datetime64tz_dtype(values) and fill_value == tslibs.iNaT:
+                    fill_value = tslibs.NaT
                 values = algos.take_nd(values, indexer, axis=ax,
                                        fill_value=fill_value)
 

diff --git a/pandas/tests/reshape/merge/test_merge.py b/pandas/tests/reshape/merge/test_merge.py
@@ -616,6 +616,24 @@ def test_merge_on_datetime64tz(self):
         assert result['value_x'].dtype == 'datetime64[ns, US/Eastern]'
         assert result['value_y'].dtype == 'datetime64[ns, US/Eastern]'
 
+    def test_merge_on_datetime64tz_empty(self):
+        # https://github.com/pandas-dev/pandas/issues/25014
+        dtz = pd.DatetimeTZDtype(tz='UTC')
+        right = pd.DataFrame({'date': [pd.Timestamp('2018', tz=dtz.tz)],
+                              'value': [4.0],
+                              'date2': [pd.Timestamp('2019', tz=dtz.tz)]},
+                             columns=['date', 'value', 'date2'])
+        left = right[:0]
+        result = left.merge(right, on='date')
+        expected = pd.DataFrame({
+            'value_x': pd.Series(dtype=float),
+            'date2_x': pd.Series(dtype=dtz),
+            'date': pd.Series(dtype=dtz),
+            'value_y': pd.Series(dtype=float),
+            'date2_y': pd.Series(dtype=dtz),
+        }, columns=['value_x', 'date2_x', 'date', 'value_y', 'date2_y'])
+        tm.assert_frame_equal(result, expected)
+
     def test_merge_datetime64tz_with_dst_transition(self):
         # GH 18885
         df1 = pd.DataFrame(pd.date_range(