CLN: _wrap_applied_output (#36160) · pandas-dev/pandas@1234234 (original) (raw)

`@@ -1192,113 +1192,104 @@ def _wrap_applied_output(self, keys, values, not_indexed_same=False):

1192

`return self.obj._constructor()

1193

`elif isinstance(first_not_none, DataFrame):

1194

`return self._concat_objects(keys, values, not_indexed_same=not_indexed_same)

1195

else:

1196

key_index = self.grouper.result_index if self.as_index else None

1197

-

1198

if isinstance(first_not_none, Series):

1199

this is to silence a DeprecationWarning

1200

TODO: Remove when default dtype of empty Series is object

1201

kwargs = first_not_none._construct_axes_dict()

1202

backup = create_series_with_explicit_dtype(

1203

dtype_if_empty=object, **kwargs

1204

)

1205

-

1206

values = [x if (x is not None) else backup for x in values]

1207

1195

1208

v = values[0]

1209

-

1210

if not isinstance(v, (np.ndarray, Index, Series)) and self.as_index:

1211

values are not series or array-like but scalars

1212

self._selection_name not passed through to Series as the

1213

result should not take the name of original selection

1214

of columns

1215

return self.obj._constructor_sliced(values, index=key_index)

1196

key_index = self.grouper.result_index if self.as_index else None

1197

+

1198

if isinstance(first_not_none, Series):

1199

this is to silence a DeprecationWarning

1200

TODO: Remove when default dtype of empty Series is object

1201

kwargs = first_not_none._construct_axes_dict()

1202

backup = create_series_with_explicit_dtype(dtype_if_empty=object, **kwargs)

1203

+

1204

values = [x if (x is not None) else backup for x in values]

1205

+

1206

v = values[0]

1207

+

1208

if not isinstance(v, (np.ndarray, Index, Series)) and self.as_index:

1209

values are not series or array-like but scalars

1210

self._selection_name not passed through to Series as the

1211

result should not take the name of original selection

1212

of columns

1213

return self.obj._constructor_sliced(values, index=key_index)

1214

+

1215

if isinstance(v, Series):

1216

all_indexed_same = all_indexes_same((x.index for x in values))

1217

+

1218

GH3596

1219

provide a reduction (Frame -> Series) if groups are

1220

unique

1221

if self.squeeze:

1222

applied_index = self._selected_obj._get_axis(self.axis)

1223

singular_series = len(values) == 1 and applied_index.nlevels == 1

1224

+

1225

assign the name to this series

1226

if singular_series:

1227

values[0].name = keys[0]

1228

+

1229

GH2893

1230

we have series in the values array, we want to

1231

produce a series:

1232

if any of the sub-series are not indexed the same

1233

OR we don't have a multi-index and we have only a

1234

single values

1235

return self._concat_objects(

1236

keys, values, not_indexed_same=not_indexed_same

1237

)

1216

1238

1239

still a series

1240

path added as of GH 5545

1241

elif all_indexed_same:

1242

from pandas.core.reshape.concat import concat

1243

+

1244

return concat(values)

1245

+

1246

if not all_indexed_same:

1247

GH 8467

1248

return self._concat_objects(keys, values, not_indexed_same=True)

1249

+

1250

Combine values

1251

vstack+constructor is faster than concat and handles MI-columns

1252

stacked_values = np.vstack([np.asarray(v) for v in values])

1253

+

1254

if self.axis == 0:

1255

index = key_index

1256

columns = v.index.copy()

1257

if columns.name is None:

1258

GH6124 - propagate name of Series when it's consistent

1259

names = {v.name for v in values}

1260

if len(names) == 1:

1261

columns.name = list(names)[0]

1217

1262

`else:

1218

if isinstance(v, Series):

1219

all_indexed_same = all_indexes_same((x.index for x in values))

1220

-

1221

GH3596

1222

provide a reduction (Frame -> Series) if groups are

1223

unique

1224

if self.squeeze:

1225

applied_index = self._selected_obj._get_axis(self.axis)

1226

singular_series = (

1227

len(values) == 1 and applied_index.nlevels == 1

1228

)

1229

-

1230

assign the name to this series

1231

if singular_series:

1232

values[0].name = keys[0]

1233

-

1234

GH2893

1235

we have series in the values array, we want to

1236

produce a series:

1237

if any of the sub-series are not indexed the same

1238

OR we don't have a multi-index and we have only a

1239

single values

1240

return self._concat_objects(

1241

keys, values, not_indexed_same=not_indexed_same

1242

)

1243

-

1244

still a series

1245

path added as of GH 5545

1246

elif all_indexed_same:

1247

from pandas.core.reshape.concat import concat

1248

-

1249

return concat(values)

1250

-

1251

if not all_indexed_same:

1252

GH 8467

1253

return self._concat_objects(keys, values, not_indexed_same=True)

1254

-

1255

Combine values

1256

vstack+constructor is faster than concat and handles MI-columns

1257

stacked_values = np.vstack([np.asarray(v) for v in values])

1258

-

1259

if self.axis == 0:

1260

index = key_index

1261

columns = v.index.copy()

1262

if columns.name is None:

1263

GH6124 - propagate name of Series when it's consistent

1264

names = {v.name for v in values}

1265

if len(names) == 1:

1266

columns.name = list(names)[0]

1267

else:

1268

index = v.index

1269

columns = key_index

1270

stacked_values = stacked_values.T

1271

-

1272

result = self.obj._constructor(

1273

stacked_values, index=index, columns=columns

1274

)

1263

index = v.index

1264

columns = key_index

1265

stacked_values = stacked_values.T

1275

1266

1276

elif not self.as_index:

1277

We add grouping column below, so create a frame here

1278

result = DataFrame(

1279

values, index=key_index, columns=[self._selection]

1280

)

1281

else:

1282

GH#1738: values is list of arrays of unequal lengths

1283

fall through to the outer else clause

1284

TODO: sure this is right? we used to do this

1285

after raising AttributeError above

1286

return self.obj._constructor_sliced(

1287

values, index=key_index, name=self._selection_name

1288

)

1267

result = self.obj._constructor(stacked_values, index=index, columns=columns)

1289

1268

1290

if we have date/time like in the original, then coerce dates

1291

as we are stacking can easily have object dtypes here

1292

so = self._selected_obj

1293

if so.ndim == 2 and so.dtypes.apply(needs_i8_conversion).any():

1294

result = _recast_datetimelike_result(result)

1295

else:

1296

result = result._convert(datetime=True)

1269

elif not self.as_index:

1270

We add grouping column below, so create a frame here

1271

result = DataFrame(values, index=key_index, columns=[self._selection])

1272

else:

1273

GH#1738: values is list of arrays of unequal lengths

1274

fall through to the outer else clause

1275

TODO: sure this is right? we used to do this

1276

after raising AttributeError above

1277

return self.obj._constructor_sliced(

1278

values, index=key_index, name=self._selection_name

1279

)

1280

+

1281

if we have date/time like in the original, then coerce dates

1282

as we are stacking can easily have object dtypes here

1283

so = self._selected_obj

1284

if so.ndim == 2 and so.dtypes.apply(needs_i8_conversion).any():

1285

result = _recast_datetimelike_result(result)

1286

else:

1287

result = result._convert(datetime=True)

1297

1288

1298

if not self.as_index:

1299

self._insert_inaxis_grouper_inplace(result)

1289

if not self.as_index:

1290

self._insert_inaxis_grouper_inplace(result)

1300

1291

1301

return self._reindex_output(result)

1292

return self._reindex_output(result)

1302

1293

1303

1294

`def _transform_general(

1304

1295

`self, func, *args, engine="cython", engine_kwargs=None, **kwargs