Skip to content

Bug: GroupBy raising error with None in first level of MultiIndex #47351

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

Merged
merged 5 commits into from
Jun 27, 2022
Merged

Bug: GroupBy raising error with None in first level of MultiIndex #47351

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
66ea20e
Bug: GroupBy raising error with None in first level of MultiIndex
phofl Jun 14, 2022
f644ecb
Merge remote-tracking branch 'upstream/main' into 47348
phofl Jun 24, 2022
c48478d
Add test
phofl Jun 24, 2022
0fae4f6
Change whatsnew
phofl Jun 25, 2022
4bdb382
Merge remote-tracking branch 'upstream/main' into 47348
phofl Jun 25, 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
1 change: 1 addition & 0 deletions doc/source/whatsnew/v1.5.0.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -904,6 +904,7 @@ Groupby/resample/rolling
- Bug in :meth:`DataFrameGroupby.cumsum` with ``skipna=False`` giving incorrect results (:issue:`46216`)
- Bug in :meth:`.GroupBy.cumsum` with ``timedelta64[ns]`` dtype failing to recognize ``NaT`` as a null value (:issue:`46216`)
- Bug in :meth:`GroupBy.cummin` and :meth:`GroupBy.cummax` with nullable dtypes incorrectly altering the original data in place (:issue:`46220`)
- Bug in :meth:`DataFrame.GroupBy` raising error when ``None`` is in first level of :class:`MultiIndex` (:issue:`47348`)
Copy link
Member

Choose a reason for hiding this comment

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

.groupby

Copy link
Member Author

Choose a reason for hiding this comment

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

Thx for your help here. Changed whatsnew

- Bug in :meth:`GroupBy.cummax` with ``int64`` dtype with leading value being the smallest possible int64 (:issue:`46382`)
- Bug in :meth:`GroupBy.max` with empty groups and ``uint64`` dtype incorrectly raising ``RuntimeError`` (:issue:`46408`)
- Bug in :meth:`.GroupBy.apply` would fail when ``func`` was a string and args or kwargs were supplied (:issue:`46479`)
Expand Down
3 changes: 3 additions & 0 deletions pandas/core/groupby/grouper.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -835,6 +835,9 @@ def get_grouper(

# if the actual grouper should be obj[key]
def is_in_axis(key) -> bool:
if key is None:
return False
Copy link
Member

@rhshadrach rhshadrach Jun 14, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

With this change the following (which I think is technically valid) now raises:

df = DataFrame({None: [1, 1, 2, 2], 'b': [1, 1, 2, 3], 'c': [4, 5, 6, 7]})
print(df.groupby(by=[None]).sum())

# Without this change:
#    b   c
# 1  2   9
# 2  5  13

A few other thoughts...

It looks like this method is the only use of _is_label_like which explicitly excludes None from being "label-like".

Also, in our tests the items.get_loc(key) is never successful in the case of SeriesGroupBy (obj.ndim == 1). In fact, I'm not sure why you'd be looking in the index for the key except maybe axis=1. But axis=1 seems useless when working with a Series.

Copy link
Member Author

Choose a reason for hiding this comment

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

Conerted to draft for now. Can not think of a reason either why we would check the index for a Series

Copy link
Member

Choose a reason for hiding this comment

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

I think the right thing to do here is to have if obj.ndim == 1: return False inside the not-label-like block. The current behavior of finding something in the Series index leads to a scalar grouper that will always raise because it's not one dimensional:

class A:
    def __str__(self):
        return 'cA'
a = A()
ser = pd.DataFrame({'a': [1, 1, a], 'b': [3, 4, 5]}).set_index('a')['b']
gb = ser.groupby([a])

raises ValueError: Grouper for 'cA' not 1-dimensional, the grouper in this case being the numpy int 5.

Copy link
Member Author

Choose a reason for hiding this comment

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

Added this and the testcase you posted above


if not _is_label_like(key):
# items -> .columns for DataFrame, .index for Series
items = obj.axes[-1]
Expand Down
11 changes: 11 additions & 0 deletions pandas/tests/groupby/test_groupby.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2776,3 +2776,14 @@ def test_by_column_values_with_same_starting_value():
).set_index("Name")

tm.assert_frame_equal(result, expected_result)


def test_groupby_none_in_first_mi_level():
# GH#47348
arr = [[None, 1, 0, 1], [2, 3, 2, 3]]
ser = Series(1, index=MultiIndex.from_arrays(arr, names=["a", "b"]))
result = ser.groupby(level=[0, 1]).sum()
expected = Series(
[1, 2], MultiIndex.from_tuples([(0.0, 2), (1.0, 3)], names=["a", "b"])
)
tm.assert_series_equal(result, expected)