Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

BUG: correctly instantiate subclassed DataFrame/Series in groupby apply #45363

Merged
merged 2 commits into from
Jan 16, 2022
Merged

BUG: correctly instantiate subclassed DataFrame/Series in groupby apply #45363

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
365ff86
BUG: correctly instantiate subclassed DataFrame/Series in groupby apply
jorisvandenbossche Jan 14, 2022
7f9dfde
update finalize tests
jorisvandenbossche Jan 14, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 4 additions & 12 deletions pandas/core/groupby/ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -753,6 +753,7 @@ def apply(
zipped = zip(group_keys, splitter)

for key, group in zipped:
group = group.__finalize__(data, method="groupby")
object.__setattr__(group, "name", key)

# group might be modified
Expand Down Expand Up @@ -1000,6 +1001,7 @@ def _aggregate_series_pure_python(
splitter = get_splitter(obj, ids, ngroups, axis=0)

for i, group in enumerate(splitter):
group = group.__finalize__(obj, method="groupby")
res = func(group)
res = libreduction.extract_result(res)

Expand Down Expand Up @@ -1243,13 +1245,7 @@ def _chop(self, sdata: Series, slice_obj: slice) -> Series:
# fastpath equivalent to `sdata.iloc[slice_obj]`
mgr = sdata._mgr.get_slice(slice_obj)
# __finalize__ not called here, must be applied by caller if applicable
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are we now doing __finalize__ in all of the places this is called? if so, better to do it here? (and on L1260)?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, there is a comment there about this:

pandas/pandas/core/groupby/ops.py

Line 1263 in 83ea173

# __finalize__ not called here, must be applied by caller if applicable

(from #37461)
But I don't know the reason that it was originally decided that finalize must be called by the caller. In any case I don't see a reason to not move it to _chop now.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

But I don't know the reason that it was originally decided that finalize must be called by the caller. In any case I don't see a reason to not move it to _chop now.

#37461 OP says it is for perf to not call it on each iteration. This PR removes that motivation, so go for it!


# fastpath equivalent to:
# `return sdata._constructor(mgr, name=sdata.name, fastpath=True)`
obj = type(sdata)._from_mgr(mgr)
object.__setattr__(obj, "_flags", sdata._flags)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are these still getting pinned?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

__finalize__ should take care of that (there is a test_finalize.py test that was changed in the diff, which was testing that this was not yet implemented. Although it was testing it with attrs and not with flags, but both should be handled by finalize).

object.__setattr__(obj, "_name", sdata._name)
return obj
return sdata._constructor(mgr, name=sdata.name, fastpath=True)


class FrameSplitter(DataSplitter):
Expand All @@ -1261,11 +1257,7 @@ def _chop(self, sdata: DataFrame, slice_obj: slice) -> DataFrame:
# return sdata.iloc[:, slice_obj]
mgr = sdata._mgr.get_slice(slice_obj, axis=1 - self.axis)
# __finalize__ not called here, must be applied by caller if applicable

# fastpath equivalent to `return sdata._constructor(mgr)`
obj = type(sdata)._from_mgr(mgr)
object.__setattr__(obj, "_flags", sdata._flags)
return obj
return sdata._constructor(mgr)


def get_splitter(
Expand Down
2 changes: 1 addition & 1 deletion pandas/tests/generic/test_finalize.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -746,6 +746,7 @@ def test_categorical_accessor(method):
"method",
[
operator.methodcaller("sum"),
lambda x: x.apply(lambda y: y),
lambda x: x.agg("sum"),
lambda x: x.agg("mean"),
lambda x: x.agg("median"),
Expand All @@ -764,7 +765,6 @@ def test_groupby_finalize(obj, method):
"method",
[
lambda x: x.agg(["sum", "count"]),
lambda x: x.apply(lambda y: y),
lambda x: x.agg("std"),
lambda x: x.agg("var"),
lambda x: x.agg("sem"),
Expand Down
23 changes: 23 additions & 0 deletions pandas/tests/groupby/test_groupby_subclass.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@

from pandas import (
DataFrame,
Index,
Series,
)
import pandas._testing as tm
Expand Down Expand Up @@ -64,6 +65,28 @@ def test_groupby_preserves_metadata():
for _, group_df in custom_df.groupby("c"):
assert group_df.testattr == "hello"

# GH-45314
def func(group):
assert isinstance(group, tm.SubclassedDataFrame)
assert hasattr(group, "testattr")
return group.testattr

result = custom_df.groupby("c").apply(func)
expected = tm.SubclassedSeries(["hello"] * 3, index=Index([7, 8, 9], name="c"))
tm.assert_series_equal(result, expected)

def func2(group):
assert isinstance(group, tm.SubclassedSeries)
assert hasattr(group, "testattr")
return group.testattr

custom_series = tm.SubclassedSeries([1, 2, 3])
custom_series.testattr = "hello"
result = custom_series.groupby(custom_df["c"]).apply(func2)
tm.assert_series_equal(result, expected)
result = custom_series.groupby(custom_df["c"]).agg(func2)
tm.assert_series_equal(result, expected)


@pytest.mark.parametrize("obj", [DataFrame, tm.SubclassedDataFrame])
def test_groupby_resample_preserves_subclass(obj):
Expand Down