Skip to content

Commit c2a0eac

Browse files
rhshadrachjbrockmendel
rhshadrach
authored and
jbrockmendel
committed
CLN: _wrap_applied_output (pandas-dev#36160)
1 parent 4480b4a commit c2a0eac

File tree

1 file changed

+91
-100
lines changed

1 file changed

+91
-100
lines changed

‎pandas/core/groupby/generic.py

Lines changed: 91 additions & 100 deletions
Original file line numberDiff line numberDiff line change
@@ -1229,113 +1229,104 @@ def _wrap_applied_output(self, keys, values, not_indexed_same=False):
12291229
return self.obj._constructor()
12301230
elif isinstance(first_not_none, DataFrame):
12311231
return self._concat_objects(keys, values, not_indexed_same=not_indexed_same)
1232-
else:
1233-
key_index = self.grouper.result_index if self.as_index else None
1234-
1235-
if isinstance(first_not_none, Series):
1236-
# this is to silence a DeprecationWarning
1237-
# TODO: Remove when default dtype of empty Series is object
1238-
kwargs = first_not_none._construct_axes_dict()
1239-
backup = create_series_with_explicit_dtype(
1240-
dtype_if_empty=object, **kwargs
1241-
)
1242-
1243-
values = [x if (x is not None) else backup for x in values]
12441232

1245-
v = values[0]
1246-
1247-
if not isinstance(v, (np.ndarray, Index, Series)) and self.as_index:
1248-
# values are not series or array-like but scalars
1249-
# self._selection_name not passed through to Series as the
1250-
# result should not take the name of original selection
1251-
# of columns
1252-
return self.obj._constructor_sliced(values, index=key_index)
1233+
key_index = self.grouper.result_index if self.as_index else None
1234+
1235+
if isinstance(first_not_none, Series):
1236+
# this is to silence a DeprecationWarning
1237+
# TODO: Remove when default dtype of empty Series is object
1238+
kwargs = first_not_none._construct_axes_dict()
1239+
backup = create_series_with_explicit_dtype(dtype_if_empty=object, **kwargs)
1240+
1241+
values = [x if (x is not None) else backup for x in values]
1242+
1243+
v = values[0]
1244+
1245+
if not isinstance(v, (np.ndarray, Index, Series)) and self.as_index:
1246+
# values are not series or array-like but scalars
1247+
# self._selection_name not passed through to Series as the
1248+
# result should not take the name of original selection
1249+
# of columns
1250+
return self.obj._constructor_sliced(values, index=key_index)
1251+
1252+
if isinstance(v, Series):
1253+
all_indexed_same = all_indexes_same((x.index for x in values))
1254+
1255+
# GH3596
1256+
# provide a reduction (Frame -> Series) if groups are
1257+
# unique
1258+
if self.squeeze:
1259+
applied_index = self._selected_obj._get_axis(self.axis)
1260+
singular_series = len(values) == 1 and applied_index.nlevels == 1
1261+
1262+
# assign the name to this series
1263+
if singular_series:
1264+
values[0].name = keys[0]
1265+
1266+
# GH2893
1267+
# we have series in the values array, we want to
1268+
# produce a series:
1269+
# if any of the sub-series are not indexed the same
1270+
# OR we don't have a multi-index and we have only a
1271+
# single values
1272+
return self._concat_objects(
1273+
keys, values, not_indexed_same=not_indexed_same
1274+
)
12531275

1276+
# still a series
1277+
# path added as of GH 5545
1278+
elif all_indexed_same:
1279+
from pandas.core.reshape.concat import concat
1280+
1281+
return concat(values)
1282+
1283+
if not all_indexed_same:
1284+
# GH 8467
1285+
return self._concat_objects(keys, values, not_indexed_same=True)
1286+
1287+
# Combine values
1288+
# vstack+constructor is faster than concat and handles MI-columns
1289+
stacked_values = np.vstack([np.asarray(v) for v in values])
1290+
1291+
if self.axis == 0:
1292+
index = key_index
1293+
columns = v.index.copy()
1294+
if columns.name is None:
1295+
# GH6124 - propagate name of Series when it's consistent
1296+
names = {v.name for v in values}
1297+
if len(names) == 1:
1298+
columns.name = list(names)[0]
12541299
else:
1255-
if isinstance(v, Series):
1256-
all_indexed_same = all_indexes_same((x.index for x in values))
1257-
1258-
# GH3596
1259-
# provide a reduction (Frame -> Series) if groups are
1260-
# unique
1261-
if self.squeeze:
1262-
applied_index = self._selected_obj._get_axis(self.axis)
1263-
singular_series = (
1264-
len(values) == 1 and applied_index.nlevels == 1
1265-
)
1266-
1267-
# assign the name to this series
1268-
if singular_series:
1269-
values[0].name = keys[0]
1270-
1271-
# GH2893
1272-
# we have series in the values array, we want to
1273-
# produce a series:
1274-
# if any of the sub-series are not indexed the same
1275-
# OR we don't have a multi-index and we have only a
1276-
# single values
1277-
return self._concat_objects(
1278-
keys, values, not_indexed_same=not_indexed_same
1279-
)
1280-
1281-
# still a series
1282-
# path added as of GH 5545
1283-
elif all_indexed_same:
1284-
from pandas.core.reshape.concat import concat
1285-
1286-
return concat(values)
1287-
1288-
if not all_indexed_same:
1289-
# GH 8467
1290-
return self._concat_objects(keys, values, not_indexed_same=True)
1291-
1292-
# Combine values
1293-
# vstack+constructor is faster than concat and handles MI-columns
1294-
stacked_values = np.vstack([np.asarray(v) for v in values])
1295-
1296-
if self.axis == 0:
1297-
index = key_index
1298-
columns = v.index.copy()
1299-
if columns.name is None:
1300-
# GH6124 - propagate name of Series when it's consistent
1301-
names = {v.name for v in values}
1302-
if len(names) == 1:
1303-
columns.name = list(names)[0]
1304-
else:
1305-
index = v.index
1306-
columns = key_index
1307-
stacked_values = stacked_values.T
1308-
1309-
result = self.obj._constructor(
1310-
stacked_values, index=index, columns=columns
1311-
)
1300+
index = v.index
1301+
columns = key_index
1302+
stacked_values = stacked_values.T
13121303

1313-
elif not self.as_index:
1314-
# We add grouping column below, so create a frame here
1315-
result = DataFrame(
1316-
values, index=key_index, columns=[self._selection]
1317-
)
1318-
else:
1319-
# GH#1738: values is list of arrays of unequal lengths
1320-
# fall through to the outer else clause
1321-
# TODO: sure this is right? we used to do this
1322-
# after raising AttributeError above
1323-
return self.obj._constructor_sliced(
1324-
values, index=key_index, name=self._selection_name
1325-
)
1304+
result = self.obj._constructor(stacked_values, index=index, columns=columns)
13261305

1327-
# if we have date/time like in the original, then coerce dates
1328-
# as we are stacking can easily have object dtypes here
1329-
so = self._selected_obj
1330-
if so.ndim == 2 and so.dtypes.apply(needs_i8_conversion).any():
1331-
result = _recast_datetimelike_result(result)
1332-
else:
1333-
result = result._convert(datetime=True)
1306+
elif not self.as_index:
1307+
# We add grouping column below, so create a frame here
1308+
result = DataFrame(values, index=key_index, columns=[self._selection])
1309+
else:
1310+
# GH#1738: values is list of arrays of unequal lengths
1311+
# fall through to the outer else clause
1312+
# TODO: sure this is right? we used to do this
1313+
# after raising AttributeError above
1314+
return self.obj._constructor_sliced(
1315+
values, index=key_index, name=self._selection_name
1316+
)
1317+
1318+
# if we have date/time like in the original, then coerce dates
1319+
# as we are stacking can easily have object dtypes here
1320+
so = self._selected_obj
1321+
if so.ndim == 2 and so.dtypes.apply(needs_i8_conversion).any():
1322+
result = _recast_datetimelike_result(result)
1323+
else:
1324+
result = result._convert(datetime=True)
13341325

1335-
if not self.as_index:
1336-
self._insert_inaxis_grouper_inplace(result)
1326+
if not self.as_index:
1327+
self._insert_inaxis_grouper_inplace(result)
13371328

1338-
return self._reindex_output(result)
1329+
return self._reindex_output(result)
13391330

13401331
def _transform_general(
13411332
self, func, *args, engine="cython", engine_kwargs=None, **kwargs

0 commit comments

Comments
 (0)