BUG: preserve timezone info when writing empty tz-aware series to HDF5

doc/source/whatsnew/v1.2.0.rst

@@ -476,6 +476,7 @@ I/O
 - Bug in :meth:`DataFrame.to_html`, :meth:`DataFrame.to_string`, and :meth:`DataFrame.to_latex` ignoring the ``na_rep`` argument when ``float_format`` was also specified (:issue:`9046`, :issue:`13828`)
 - Bug in output rendering of complex numbers showing too many trailing zeros (:issue:`36799`)
 - Bug in :class:`HDFStore` threw a ``TypeError`` when exporting an empty :class:`DataFrame` with ``datetime64[ns, tz]`` dtypes with a fixed HDF5 store (:issue:`20594`)
+- Bug in :class:`HDFStore` was dropping timezone information when exporting :class:`Series` with ``datetime64[ns, tz]`` dtypes with a fixed HDF5 store (:issue:`20594`)
 
 Plotting
 ^^^^^^^^

pandas/io/pytables.py

@@ -45,7 +45,6 @@
     is_string_dtype,
     is_timedelta64_dtype,
 )
-from pandas.core.dtypes.generic import ABCExtensionArray
 from pandas.core.dtypes.missing import array_equivalent
 
 from pandas import (
@@ -63,6 +62,7 @@
 from pandas.core.arrays import Categorical, DatetimeArray, PeriodArray
 import pandas.core.common as com
 from pandas.core.computation.pytables import PyTablesExpr, maybe_expression
+from pandas.core.construction import extract_array
 from pandas.core.indexes.api import ensure_index
 
 from pandas.io.common import stringify_path
@@ -2968,11 +2968,12 @@ def write_array_empty(self, key: str, value: ArrayLike):
         node._v_attrs.value_type = str(value.dtype)
         node._v_attrs.shape = value.shape
 
-    def write_array(self, key: str, value: ArrayLike, items: Optional[Index] = None):
-        # TODO: we only have one test that gets here, the only EA
+    def write_array(self, key: str, obj: FrameOrSeries, items: Optional[Index] = None):
+        # TODO: we only have a few tests that get here, the only EA
         #  that gets passed is DatetimeArray, and we never have
         #  both self._filters and EA
-        assert isinstance(value, (np.ndarray, ABCExtensionArray)), type(value)
+
+        value = extract_array(obj, extract_numpy=True)
 
         if key in self.group:
             self._handle.remove_node(self.group, key)
@@ -3077,7 +3078,7 @@ def read(
     def write(self, obj, **kwargs):
         super().write(obj, **kwargs)
         self.write_index("index", obj.index)
-        self.write_array("values", obj.values)
+        self.write_array("values", obj)
         self.attrs.name = obj.name
 
 

pandas/tests/io/pytables/test_timezones.py

@@ -296,8 +296,11 @@ def test_timezones_fixed_format_frame_non_empty(setup_path):
         tm.assert_frame_equal(result, df)
 
 
-@pytest.mark.parametrize("dtype", ["datetime64[ns, UTC]", "datetime64[ns, US/Eastern]"])
-def test_timezones_fixed_format_frame_empty(setup_path, dtype):
+def test_timezones_fixed_format_frame_empty(setup_path, tz_aware_fixture):
+    # GH 20594
+
+    dtype = pd.DatetimeTZDtype(tz=tz_aware_fixture)
+
     with ensure_clean_store(setup_path) as store:
         s = Series(dtype=dtype)
         df = DataFrame({"A": s})
@@ -306,6 +309,30 @@ def test_timezones_fixed_format_frame_empty(setup_path, dtype):
         tm.assert_frame_equal(result, df)
 
 
+def test_timezones_fixed_format_series_nonempty(setup_path, tz_aware_fixture):
+    # GH 20594
+
+    dtype = pd.DatetimeTZDtype(tz=tz_aware_fixture)
+
+    with ensure_clean_store(setup_path) as store:
+        s = Series([0], dtype=dtype)
+        store["s"] = s
+        result = store["s"]
+        tm.assert_series_equal(result, s)
+
+
+def test_timezones_fixed_format_series_empty(setup_path, tz_aware_fixture):
+    # GH 20594
+
+    dtype = pd.DatetimeTZDtype(tz=tz_aware_fixture)
+
+    with ensure_clean_store(setup_path) as store:
+        s = Series(dtype=dtype)
+        store["s"] = s
+        result = store["s"]
+        tm.assert_series_equal(result, s)
+
+
 def test_fixed_offset_tz(setup_path):
     rng = date_range("1/1/2000 00:00:00-07:00", "1/30/2000 00:00:00-07:00")
     frame = DataFrame(np.random.randn(len(rng), 4), index=rng)