Skip to content

BUG: get_indexer rountripping through string dtype #56013

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 9 commits into from
Nov 16, 2024
Merged

BUG: get_indexer rountripping through string dtype #56013

Show file tree
Hide file tree
Changes from 8 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
d5c7e9a
BUG: get_indexer rountripping through string dtype
phofl Nov 16, 2023
bc87192
Update test_indexing.py
phofl Nov 18, 2023
868850f
Merge branch 'main' into 55837
phofl Nov 21, 2023
e7e1d0b
Merge remote-tracking branch 'upstream/main' into 55837
jorisvandenbossche Oct 28, 2024
bdfd2d3
Merge remote-tracking branch 'upstream/main' into 55837
jorisvandenbossche Nov 4, 2024
cfb2f09
update teset
jorisvandenbossche Nov 4, 2024
f424a62
Merge remote-tracking branch 'upstream/main' into 55837
jorisvandenbossche Nov 14, 2024
d202057
remove pyarrow skip
jorisvandenbossche Nov 14, 2024
1609e89
more specific check for object/string combo
jorisvandenbossche Nov 14, 2024
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
2 changes: 1 addition & 1 deletion doc/source/whatsnew/v2.3.0.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,7 +118,7 @@ Interval

Indexing
^^^^^^^^
-
- Fixed bug in :meth:`Index.get_indexer` round-tripping through string dtype when ``infer_string`` is enabled (:issue:`55834`)
-

Missing
Expand Down
3 changes: 3 additions & 0 deletions pandas/core/indexes/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6556,6 +6556,9 @@ def _maybe_cast_listlike_indexer(self, target) -> Index:
"""
Analogue to maybe_cast_indexer for get_indexer instead of get_loc.
"""
if not hasattr(target, "dtype") and self.dtype == object:
# Avoid inference for object since we are casting back later anyway
return Index(target, dtype=self.dtype)
Copy link
Member

Choose a reason for hiding this comment

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

hmm this might prevent us from doing intentional inference in _maybe_downcast_for_indexing

Copy link
Member

Choose a reason for hiding this comment

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

@jbrockmendel could you think of a concrete situation where this might happen?

(the tests are all passing, so don't seem to cover that?)

Copy link
Member

Choose a reason for hiding this comment

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

Looking at _maybe_downcast_for_indexing more closely, we are downcasting object dtype index to the other dtype, because this can give a performance improvement, right?

This might actually also be true for object/string combo, but I think correctness if more important, and if missing values are handled differently by downcasting, we should avoid that for now for this combo, until we can solve that better.

I will just make the check a bit more specific for object dtype then (even more ugly, but at least fixing this issue)

return ensure_index(target)

@final
Expand Down
9 changes: 9 additions & 0 deletions pandas/tests/indexes/object/test_indexing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,15 @@ def test_get_indexer_with_NA_values(
expected = np.array([0, 1, -1], dtype=np.intp)
tm.assert_numpy_array_equal(result, expected)

def test_get_indexer_infer_string_missing_values(self):
# ensure the passed list is not cast to string but to object so that
# the None value is matched in the index
# https://github.com/pandas-dev/pandas/issues/55834
idx = Index(["a", "b", None], dtype="object")
result = idx.get_indexer([None, "x"])
expected = np.array([2, -1], dtype=np.intp)
tm.assert_numpy_array_equal(result, expected)


class TestGetIndexerNonUnique:
def test_get_indexer_non_unique_nas(self, nulls_fixture):
Expand Down
Loading