Skip to content

Commit 55f59e9

Browse files
jorisvandenbosschejorisvandenbossche
committed
API: ignore empty range/object dtype in Index setop operations (string dtype compat) (pandas-dev#60797)
(cherry picked from commit ee06e71)
1 parent 3143f44 commit 55f59e9

24 files changed

+990
-95
lines changed

‎doc/source/whatsnew/v2.3.0.rst

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -59,6 +59,16 @@ Increased minimum version for Python
5959

6060
pandas 2.3.0 supports Python 3.10 and higher.
6161

62+
.. _whatsnew_230.api_changes:
63+
64+
API changes
65+
~~~~~~~~~~~
66+
67+
- When enabling the ``future.infer_string`` option: Index set operations (like
68+
union or intersection) will now ignore the dtype of an empty ``RangeIndex`` or
69+
empty ``Index`` with object dtype when determining the dtype of the resulting
70+
Index (:issue:`60797`)
71+
6272
.. ---------------------------------------------------------------------------
6373
.. _whatsnew_230.deprecations:
6474

‎doc/source/whatsnew/v3.0.0.rst

Lines changed: 841 additions & 0 deletions
Large diffs are not rendered by default.

‎pandas/core/indexes/base.py

Lines changed: 26 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -6387,6 +6387,24 @@ def _find_common_type_compat(self, target) -> DtypeObj:
63876387
"""
63886388
target_dtype, _ = infer_dtype_from(target)
63896389

6390+
if using_string_dtype():
6391+
# special case: if left or right is a zero-length RangeIndex or
6392+
# Index[object], those can be created by the default empty constructors
6393+
# -> for that case ignore this dtype and always return the other
6394+
# (https://github.com/pandas-dev/pandas/pull/60797)
6395+
from pandas.core.indexes.range import RangeIndex
6396+
6397+
if len(self) == 0 and (
6398+
isinstance(self, RangeIndex) or self.dtype == np.object_
6399+
):
6400+
return target_dtype
6401+
if (
6402+
isinstance(target, Index)
6403+
and len(target) == 0
6404+
and (isinstance(target, RangeIndex) or target_dtype == np.object_)
6405+
):
6406+
return self.dtype
6407+
63906408
# special case: if one dtype is uint64 and the other a signed int, return object
63916409
# See https://github.com/pandas-dev/pandas/issues/26778 for discussion
63926410
# Now it's:
@@ -7005,6 +7023,14 @@ def insert(self, loc: int, item) -> Index:
70057023

70067024
arr = self._values
70077025

7026+
if using_string_dtype() and len(self) == 0 and self.dtype == np.object_:
7027+
# special case: if we are an empty object-dtype Index, also
7028+
# take into account the inserted item for the resulting dtype
7029+
# (https://github.com/pandas-dev/pandas/pull/60797)
7030+
dtype = self._find_common_type_compat(item)
7031+
if dtype != self.dtype:
7032+
return self.astype(dtype).insert(loc, item)
7033+
70087034
try:
70097035
if isinstance(arr, ExtensionArray):
70107036
res_values = arr.insert(loc, item)

‎pandas/tests/frame/constructors/test_from_dict.py

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -3,8 +3,6 @@
33
import numpy as np
44
import pytest
55

6-
from pandas._config import using_string_dtype
7-
86
from pandas import (
97
DataFrame,
108
Index,
@@ -44,7 +42,6 @@ def test_constructor_single_row(self):
4442
)
4543
tm.assert_frame_equal(result, expected)
4644

47-
@pytest.mark.xfail(using_string_dtype(), reason="columns inferring logic broken")
4845
def test_constructor_list_of_series(self):
4946
data = [
5047
OrderedDict([["a", 1.5], ["b", 3.0], ["c", 4.0]]),

‎pandas/tests/frame/indexing/test_coercion.py

Lines changed: 1 addition & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -103,12 +103,7 @@ def test_26395(indexer_al):
103103
df["D"] = 0
104104

105105
indexer_al(df)["C", "D"] = 2
106-
expected = DataFrame(
107-
{"D": [0, 0, 2]},
108-
index=["A", "B", "C"],
109-
columns=pd.Index(["D"], dtype=object),
110-
dtype=np.int64,
111-
)
106+
expected = DataFrame({"D": [0, 0, 2]}, index=["A", "B", "C"], dtype=np.int64)
112107
tm.assert_frame_equal(df, expected)
113108

114109
with tm.assert_produces_warning(

‎pandas/tests/frame/indexing/test_indexing.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1206,7 +1206,7 @@ def test_loc_setitem_datetimelike_with_inference(self):
12061206
result = df.dtypes
12071207
expected = Series(
12081208
[np.dtype("timedelta64[ns]")] * 6 + [np.dtype("datetime64[ns]")] * 2,
1209-
index=Index(list("ABCDEFGH"), dtype=object),
1209+
index=list("ABCDEFGH"),
12101210
)
12111211
tm.assert_series_equal(result, expected)
12121212

‎pandas/tests/frame/indexing/test_insert.py

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -67,8 +67,7 @@ def test_insert_with_columns_dups(self):
6767
df.insert(0, "A", ["d", "e", "f"], allow_duplicates=True)
6868
df.insert(0, "A", ["a", "b", "c"], allow_duplicates=True)
6969
exp = DataFrame(
70-
[["a", "d", "g"], ["b", "e", "h"], ["c", "f", "i"]],
71-
columns=Index(["A", "A", "A"], dtype=object),
70+
[["a", "d", "g"], ["b", "e", "h"], ["c", "f", "i"]], columns=["A", "A", "A"]
7271
)
7372
tm.assert_frame_equal(df, exp)
7473

‎pandas/tests/frame/indexing/test_setitem.py

Lines changed: 22 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -146,18 +146,32 @@ def test_setitem_different_dtype(self):
146146
)
147147
tm.assert_series_equal(result, expected)
148148

149-
def test_setitem_empty_columns(self):
150-
# GH 13522
149+
def test_setitem_overwrite_index(self):
150+
# GH 13522 - assign the index as a column and then overwrite the values
151+
# -> should not affect the index
151152
df = DataFrame(index=["A", "B", "C"])
152153
df["X"] = df.index
153154
df["X"] = ["x", "y", "z"]
154155
exp = DataFrame(
155-
data={"X": ["x", "y", "z"]},
156-
index=["A", "B", "C"],
157-
columns=Index(["X"], dtype=object),
156+
data={"X": ["x", "y", "z"]}, index=["A", "B", "C"], columns=["X"]
158157
)
159158
tm.assert_frame_equal(df, exp)
160159

160+
def test_setitem_empty_columns(self):
161+
# Starting from an empty DataFrame and setting a column should result
162+
# in a default string dtype for the columns' Index
163+
# https://github.com/pandas-dev/pandas/issues/60338
164+
165+
df = DataFrame()
166+
df["foo"] = [1, 2, 3]
167+
expected = DataFrame({"foo": [1, 2, 3]})
168+
tm.assert_frame_equal(df, expected)
169+
170+
df = DataFrame(columns=Index([]))
171+
df["foo"] = [1, 2, 3]
172+
expected = DataFrame({"foo": [1, 2, 3]})
173+
tm.assert_frame_equal(df, expected)
174+
161175
def test_setitem_dt64_index_empty_columns(self):
162176
rng = date_range("1/1/2000 00:00:00", "1/1/2000 1:59:50", freq="10s")
163177
df = DataFrame(index=np.arange(len(rng)))
@@ -171,9 +185,7 @@ def test_setitem_timestamp_empty_columns(self):
171185
df["now"] = Timestamp("20130101", tz="UTC").as_unit("ns")
172186

173187
expected = DataFrame(
174-
[[Timestamp("20130101", tz="UTC")]] * 3,
175-
index=range(3),
176-
columns=Index(["now"], dtype=object),
188+
[[Timestamp("20130101", tz="UTC")]] * 3, index=range(3), columns=["now"]
177189
)
178190
tm.assert_frame_equal(df, expected)
179191

@@ -212,7 +224,7 @@ def test_setitem_period_preserves_dtype(self):
212224
result = DataFrame([])
213225
result["a"] = data
214226

215-
expected = DataFrame({"a": data}, columns=Index(["a"], dtype=object))
227+
expected = DataFrame({"a": data}, columns=["a"])
216228

217229
tm.assert_frame_equal(result, expected)
218230

@@ -939,7 +951,7 @@ def test_setitem_scalars_no_index(self):
939951
# GH#16823 / GH#17894
940952
df = DataFrame()
941953
df["foo"] = 1
942-
expected = DataFrame(columns=Index(["foo"], dtype=object)).astype(np.int64)
954+
expected = DataFrame(columns=["foo"]).astype(np.int64)
943955
tm.assert_frame_equal(df, expected)
944956

945957
def test_setitem_newcol_tuple_key(self, float_frame):

‎pandas/tests/frame/methods/test_dropna.py

Lines changed: 1 addition & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -182,12 +182,9 @@ def test_dropna_multiple_axes(self):
182182
with pytest.raises(TypeError, match="supplying multiple axes"):
183183
inp.dropna(how="all", axis=(0, 1), inplace=True)
184184

185-
def test_dropna_tz_aware_datetime(self, using_infer_string):
185+
def test_dropna_tz_aware_datetime(self):
186186
# GH13407
187-
188187
df = DataFrame()
189-
if using_infer_string:
190-
df.columns = df.columns.astype("str")
191188
dt1 = datetime.datetime(2015, 1, 1, tzinfo=dateutil.tz.tzutc())
192189
dt2 = datetime.datetime(2015, 2, 2, tzinfo=dateutil.tz.tzutc())
193190
df["Time"] = [dt1]

‎pandas/tests/frame/methods/test_reset_index.py

Lines changed: 31 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -4,8 +4,6 @@
44
import numpy as np
55
import pytest
66

7-
from pandas._config import using_string_dtype
8-
97
from pandas.core.dtypes.common import (
108
is_float_dtype,
119
is_integer_dtype,
@@ -646,7 +644,6 @@ def test_rest_index_multiindex_categorical_with_missing_values(self, codes):
646644
tm.assert_frame_equal(res, expected)
647645

648646

649-
@pytest.mark.xfail(using_string_dtype(), reason="TODO(infer_string) - GH#60338")
650647
@pytest.mark.parametrize(
651648
"array, dtype",
652649
[
@@ -783,3 +780,34 @@ def test_reset_index_false_index_name():
783780
result_frame.reset_index()
784781
expected_frame = DataFrame(range(5, 10), RangeIndex(range(5), name=False))
785782
tm.assert_frame_equal(result_frame, expected_frame)
783+
784+
785+
@pytest.mark.parametrize("columns", [None, Index([])])
786+
def test_reset_index_with_empty_frame(columns):
787+
# Currently empty DataFrame has RangeIndex or object dtype Index, but when
788+
# resetting the index we still want to end up with the default string dtype
789+
# https://github.com/pandas-dev/pandas/issues/60338
790+
791+
index = Index([], name="foo")
792+
df = DataFrame(index=index, columns=columns)
793+
result = df.reset_index()
794+
expected = DataFrame(columns=["foo"])
795+
tm.assert_frame_equal(result, expected)
796+
797+
index = Index([1, 2, 3], name="foo")
798+
df = DataFrame(index=index, columns=columns)
799+
result = df.reset_index()
800+
expected = DataFrame({"foo": [1, 2, 3]})
801+
tm.assert_frame_equal(result, expected)
802+
803+
index = MultiIndex.from_tuples([], names=["foo", "bar"])
804+
df = DataFrame(index=index, columns=columns)
805+
result = df.reset_index()
806+
expected = DataFrame(columns=["foo", "bar"])
807+
tm.assert_frame_equal(result, expected)
808+
809+
index = MultiIndex.from_tuples([(1, 2), (2, 3)], names=["foo", "bar"])
810+
df = DataFrame(index=index, columns=columns)
811+
result = df.reset_index()
812+
expected = DataFrame({"foo": [1, 2], "bar": [2, 3]})
813+
tm.assert_frame_equal(result, expected)

0 commit comments

Comments
 (0)