BUG: fix df.where(cond) when cond is empty (#21947)

Author: pajachiet

doc/source/whatsnew/v0.24.0.txt

@@ -1047,7 +1047,7 @@ Removal of prior version deprecations/changes
 Performance Improvements
 ~~~~~~~~~~~~~~~~~~~~~~~~
 
-- Slicing Series and Dataframes with an monotonically increasing :class:`CategoricalIndex`
+- Slicing Series and DataFrames with an monotonically increasing :class:`CategoricalIndex`
   is now very fast and has speed comparable to slicing with an ``Int64Index``.
   The speed increase is both when indexing by label (using .loc) and position(.iloc) (:issue:`20395`)
   Slicing a monotonically increasing :class:`CategoricalIndex` itself (i.e. ``ci[1000:2000]``)
@@ -1150,7 +1150,7 @@ Timezones
 - Fixed bug where :meth:`DataFrame.describe` and :meth:`Series.describe` on tz-aware datetimes did not show `first` and `last` result (:issue:`21328`)
 - Bug in :class:`DatetimeIndex` comparisons failing to raise ``TypeError`` when comparing timezone-aware ``DatetimeIndex`` against ``np.datetime64`` (:issue:`22074`)
 - Bug in ``DataFrame`` assignment with a timezone-aware scalar (:issue:`19843`)
-- Bug in :func:`Dataframe.asof` that raised a ``TypeError`` when attempting to compare tz-naive and tz-aware timestamps (:issue:`21194`)
+- Bug in :func:`DataFrame.asof` that raised a ``TypeError`` when attempting to compare tz-naive and tz-aware timestamps (:issue:`21194`)
 - Bug when constructing a :class:`DatetimeIndex` with :class:`Timestamp`s constructed with the ``replace`` method across DST (:issue:`18785`)
 - Bug when setting a new value with :meth:`DataFrame.loc` with a :class:`DatetimeIndex` with a DST transition (:issue:`18308`, :issue:`20724`)
 - Bug in :meth:`DatetimeIndex.unique` that did not re-localize tz-aware dates correctly (:issue:`21737`)
@@ -1313,6 +1313,7 @@ Reshaping
 - Bug in :func:`pandas.concat` when joining resampled DataFrames with timezone aware index (:issue:`13783`)
 - Bug in :meth:`Series.combine_first` with ``datetime64[ns, tz]`` dtype which would return tz-naive result (:issue:`21469`)
 - Bug in :meth:`Series.where` and :meth:`DataFrame.where` with ``datetime64[ns, tz]`` dtype (:issue:`21546`)
+- Bug in :meth:`DataFrame.where` with an empty DataFrame and empty ``cond`` having non-bool dtype (:issue:`21947`)
 - Bug in :meth:`Series.mask` and :meth:`DataFrame.mask` with ``list`` conditionals (:issue:`21891`)
 - Bug in :meth:`DataFrame.replace` raises RecursionError when converting OutOfBounds ``datetime64[ns, tz]`` (:issue:`20380`)
 - :func:`pandas.core.groupby.GroupBy.rank` now raises a ``ValueError`` when an invalid value is passed for argument ``na_option`` (:issue:`22124`)

pandas/core/generic.py

@@ -8142,7 +8142,7 @@ def _where(self, cond, other=np.nan, inplace=False, axis=None, level=None,
             # This is a single-dimensional object.
             if not is_bool_dtype(cond):
                 raise ValueError(msg.format(dtype=cond.dtype))
-        else:
+        elif not cond.empty:
             for dt in cond.dtypes:
                 if not is_bool_dtype(dt):
                     raise ValueError(msg.format(dtype=dt))

pandas/tests/frame/test_indexing.py

@@ -2877,6 +2877,14 @@ def test_where_none(self):
                                     'on mixed-type'):
             df.where(~isna(df), None, inplace=True)
 
+    def test_where_empty_df_and_empty_cond_having_non_bool_dtypes(self):
+        # see gh-21947
+        df = pd.DataFrame(columns=["a"])
+        cond = df.applymap(lambda x: x > 0)
+
+        result = df.where(cond)
+        tm.assert_frame_equal(result, df)
+
     def test_where_align(self):
 
         def create():