Skip to content

BUG: Set dtypes of new columns when stacking (#36991) #40127

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 17 commits into from
Mar 10, 2021
Merged

BUG: Set dtypes of new columns when stacking (#36991) #40127

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
1b90519
BUG: Set dtypes of new columns when stacking (#36991)
maroth96 Feb 28, 2021
f709330
Fix PEP8 issues
maroth96 Feb 28, 2021
23ff7ba
Fix pre-commit issues and add GH comment
maroth96 Feb 28, 2021
9cb2055
Use MultiIndex.from_arrays
maroth96 Feb 28, 2021
75cb56c
Reformat
maroth96 Feb 28, 2021
61164b2
Remove unnecessary list()
maroth96 Mar 2, 2021
611841a
Fix mypy error
maroth96 Mar 2, 2021
d8e7517
Add whatsnew entry
maroth96 Mar 2, 2021
8016c7f
Rewrite test such that it compares the series
maroth96 Mar 3, 2021
ce161c7
Refactor such that unique_groups is no longer needed
maroth96 Mar 3, 2021
bc69788
Extract method for stacking the column index
maroth96 Mar 3, 2021
379bde5
Reformat
maroth96 Mar 3, 2021
17e841d
Separate complex expression and add a comment
maroth96 Mar 3, 2021
9a7b29e
Add function annotation and explanatory comments
maroth96 Mar 6, 2021
76b5465
Merge branch 'master' into b-36991
maroth96 Mar 6, 2021
c09697b
Rewrite loop as list comprehension
maroth96 Mar 9, 2021
c6ab291
Add typing to _stack_multi_column_index
maroth96 Mar 10, 2021
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
11 changes: 9 additions & 2 deletions pandas/core/reshape/reshape.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -643,8 +643,15 @@ def _convert_level_number(level_num, columns):
levs.append(np.take(lev, level_codes))
tuples = list(zip(*levs))
unique_groups = [key for key, _ in itertools.groupby(tuples)]
new_names = this.columns.names[:-1]
new_columns = MultiIndex.from_tuples(unique_groups, names=new_names)
new_columns = MultiIndex.from_arrays(
[
Index(new_level, dtype=level.dtype)
if None not in new_level
Copy link
Contributor

Choose a reason for hiding this comment

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

do we have any tests that hit the None case?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, specifically test_stack_nan_in_multiindex_columns.

None is a tricky case---it is allowed for some Index types (like CategoricalIndex) but not others (like Int64Index). I am just avoiding levels with None entirely.

I've also looked at the way None is handled elsewhere, and it's not totally consistent:

>>> MultiIndex.from_arrays([[1, 2, None]]).levels[0]
Int64Index([1, 2], dtype='int64')

But:

>>> Index([1, 2, None])
Index([1, 2, None], dtype='object')

Perhaps in the future it would be better if Index accepted None values, even when a dtype is specified. Then Index([1, 2, None], dtype='int64') would return Int64Index([1, 2], dtype='int64'). Then we wouldn't need such a condition here. Thoughts?

Copy link
Contributor

Choose a reason for hiding this comment

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

hmm interesting, ok can you open an issue specifically showing the MI vs Index cases. I agree we should do something about this. ok for here on this PR.

else new_level
for new_level, level in zip(zip(*unique_groups), this.columns.levels)
],
names=this.columns.names[:-1],
)
else:
new_columns = this.columns.levels[0]._rename(name=this.columns.names[0])
unique_groups = new_columns
Expand Down
24 changes: 24 additions & 0 deletions pandas/tests/frame/test_stack_unstack.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1052,6 +1052,30 @@ def test_stack_preserve_categorical_dtype(self, ordered, labels):

tm.assert_series_equal(result, expected)

@pytest.mark.parametrize("ordered", [False, True])
@pytest.mark.parametrize("labels", [list("yxz"), list("yzx")])
@pytest.mark.parametrize("labels2", [list("uv"), list("vu")])
def test_multi_stack_preserve_categorical_dtype(self, ordered, labels, labels2):
# GH-36991
cidx = pd.CategoricalIndex(labels, categories=list("xyz"), ordered=ordered)
cidx2 = pd.CategoricalIndex(labels2, categories=list("uv"), ordered=ordered)
sorted_cidx = pd.CategoricalIndex(
list("xyz"), categories=list("xyz"), ordered=ordered
)
sorted_cidx2 = pd.CategoricalIndex(
list("uv"), categories=list("uv"), ordered=ordered
)

midx = MultiIndex.from_product([cidx, cidx2, [1, 2, 3]], names=list("abc"))
df = DataFrame(np.random.randn(5, midx.size), columns=midx)
result = df.stack(["a", "b"])

expected = MultiIndex.from_product(
[df.index, sorted_cidx, sorted_cidx2], names=[None, "a", "b"]
)

tm.assert_equal(result.index, expected)

def test_stack_preserve_categorical_dtype_values(self):
# GH-23077
cat = pd.Categorical(["a", "a", "b", "c"])
Expand Down