GH1641 Pandas 3.0 support #1643

loicdiridollou · 2026-01-22T20:16:09Z

Closes #xxxx (Replace xxxx with the Github issue number)
Tests added (Please use assert_type() to assert the type of any return value)
If I used AI to develop this pull request, I prompted it to follow AGENTS.md.

loicdiridollou · 2026-01-22T20:35:54Z

I did not remove all the pytest warn bounded, there is already a ton of stuff to look at so thought I would do it in a follow up. @cmp0xff has already made the changes for dropping the dependency on 3.10 so did not do all of those.

cmp0xff · 2026-01-23T12:35:28Z

pandas-stubs/core/frame.pyi

        col_fill: Hashable = ...,
        drop: _bool = ...,
-        inplace: Literal[False] = False,
+        inplace: bool = False,


inplace=True still returns None. Please fix and also add a test.

Reverted and added

cmp0xff · 2026-01-23T12:37:05Z

pandas-stubs/core/frame.pyi

-    @overload
-    def stack(
-        self,
-        level: IndexLabel = ...,
-        *,
-        future_stack: Literal[True],
-    ) -> Self | Series: ...
-    @overload
    def stack(
        self,
        level: IndexLabel = ...,
-        dropna: _bool = ...,
-        sort: _bool = ...,
-        future_stack: Literal[False] = False,
+        future_stack: Literal[True] = True,
    ) -> Self | Series: ...


Both implementations still coexist. I guess we should keep the old implementation.

According to the old warning message, the future_stack=False is a deprecated implementation so I thought we should not let the user use them, unless they willfully ignore the warning. What do you think?

cmp0xff · 2026-01-23T12:47:08Z

tests/arrays/test_arrow_string_array.py

-        arr_arrow = cast("ArrowStringArray", pd.array([pd.NA], str))
-        check(assert_type(arr_arrow, ArrowStringArray), ArrowStringArray, float)
-        assert pd.isna(assert_type(arr_arrow.dtype.na_value, NAType | float))
+    # pandas-dev/pandas#63567


pandas-dev/pandas#54466 will need to be propagated, maybe in a separate PR.

Suggested change

# pandas-dev/pandas#63567

# TODO: pandas-dev/pandas#54466 should give BaseStringArray after 3.0

cmp0xff · 2026-01-23T12:48:53Z

tests/arrays/test_base_string_array.py

+    check(pd.array([*data, *data, np.nan], dtype), BaseStringArray, dtype_na)

    if TYPE_CHECKING:
        # TODO: pandas-dev/pandas#54466 should give BaseStringArray after 3.0


pandas-dev/pandas#54466 will need to be propagated, maybe in a separate PR.

Yes I plan to do those in a separate one, otherwise it is just a massive bunch of changes and so easy to miss something.

cmp0xff · 2026-01-23T12:49:50Z

tests/arrays/test_datetime_array.py

+    check(assert_type(arr.month_name(), np_1darray_object), BaseStringArray)
+    check(assert_type(arr.day_name(), np_1darray_object), BaseStringArray)


The following needs to work

Suggested change

check(assert_type(arr.month_name(), np_1darray_object), BaseStringArray)

check(assert_type(arr.day_name(), np_1darray_object), BaseStringArray)

check(assert_type(arr.month_name(), BaseStringArray), BaseStringArray)

check(assert_type(arr.day_name(), BaseStringArray), BaseStringArray)

cmp0xff · 2026-01-23T13:07:05Z

tests/test_groupby.py


-        # interpolate
-        if PD_LTE_23:
-            check(assert_type(GB_DF.resample("ME").interpolate(), DataFrame), DataFrame)


Do we need negative tests here?

cmp0xff · 2026-01-23T13:07:23Z

tests/test_groupby.py

-    if PD_LTE_23:
-        with pytest_warns_bounded(
-            FutureWarning,
-            r"The provided callable <function (sum|mean) .*> is currently using ",


Probably need to disable the previous implementation and add negative tests in if TYPE_CHECKING_INVALID_USAGE

cmp0xff · 2026-01-23T13:07:35Z

tests/test_groupby.py

-    # aggregate
-    with pytest_warns_bounded(
-        FutureWarning,
-        r"The provided callable <function (sum|mean) .*> is currently using ",


Probably need to disable the previous implementation and add negative tests in if TYPE_CHECKING_INVALID_USAGE

cmp0xff · 2026-01-23T13:07:44Z

tests/test_windowing.py

-    if PD_LTE_23:
-        with pytest_warns_bounded(
-            FutureWarning,
-            r"The provided callable <function (sum|mean) .*> is currently using ",


Probably need to disable the previous implementation and add negative tests in if TYPE_CHECKING_INVALID_USAGE

I think something is broken there:

import numpy as np from pandas import DataFrame df = DataFrame({"A": ["a"] * 4, "B": range(4)}) df.groupby("A").ewm(1).aggregate(sum) # fails df.groupby("A").ewm(1).aggregate("sum") # works df.groupby("A").ewm(1).aggregate(np.sum) # fails

Raised issue in pandas-dev/pandas#63855

Seems like they fixed it on main but it will remain broken on 3.0.0, we can decide to ignore the tests until 3.0.1 is released.

cmp0xff · 2026-01-23T13:07:52Z

tests/test_windowing.py

-    if PD_LTE_23:
-        with pytest_warns_bounded(
-            FutureWarning,
-            r"The provided callable <function (sum|mean) .*> is currently using ",


Probably need to disable the previous implementation and add negative tests in if TYPE_CHECKING_INVALID_USAGE

Same issue as pandas-dev/pandas#63855

cmp0xff · 2026-01-23T14:49:12Z

tests/series/test_series.py

    if TYPE_CHECKING_INVALID_USAGE:
-        s.clip(lower=lower, axis=1)  # type: ignore[call-overload] # pyright: ignore[reportArgumentType]
-        s.clip(lower=lower, axis="column")  # type: ignore[call-overload] # pyright: ignore[reportArgumentType]
+        s.clip(lower=lower, axis=1)  # type: ignore[call-overload] # pyright: ignore[reportCallIssue, reportArgumentType]


Suggested change

s.clip(lower=lower, axis=1) # type: ignore[call-overload] # pyright: ignore[reportCallIssue, reportArgumentType]

s.clip(lower=lower, axis=1) # type: ignore[call-overload] # pyright: ignore[reportCallIssue,reportArgumentType]

cmp0xff · 2026-01-23T14:50:40Z

tests/series/test_series.py

-            FutureWarning,
-            r"Logical ops \(and, or, xor\) between Pandas objects and dtype-less sequences "
-            r"\(e.g. list, tuple\) are deprecated",
-            lower="2.0.99",


Maybe negative tests

cmp0xff · 2026-01-23T14:50:57Z

tests/series/test_series.py

-            FutureWarning,
-            r"Logical ops \(and, or, xor\) between Pandas objects and dtype-less sequences "
-            r"\(e.g. list, tuple\) are deprecated",
-            lower="2.0.99",


Maybe negative tests

Dr-Irv · 2026-01-24T18:41:21Z

I had started a review, then @cmp0xff did a review, then @loicdiridollou pushed new commits, so I will discard that review, and start from the latest commit. Please let me handle this going forward.

loicdiridollou · 2026-01-24T18:56:41Z

I will hold on pushing anything until I get your review, there is a lot of changes so it will be easier like this!

loicdiridollou added 3 commits January 22, 2026 15:14

GH1641 Pandas 3.0 support

41911ed

GH1641 Drop python 3.10

fd2fa7a

GH1641 Fix missing 3.12 version split

9de72ab

loicdiridollou requested review from Dr-Irv and cmp0xff and removed request for Dr-Irv January 22, 2026 20:34

loicdiridollou added 2 commits January 22, 2026 16:48

Merge branch 'main' into gh1641_pandas_30

87f529a

GH1641 Fix missing 3.12 version split

cfb75eb

cmp0xff requested changes Jan 23, 2026

View reviewed changes

GH1641 PR Feedback

6ad17ac

	# pandas-dev/pandas#63567
	# TODO: pandas-dev/pandas#54466 should give BaseStringArray after 3.0

		check(assert_type(arr.month_name(), np_1darray_object), BaseStringArray)
		check(assert_type(arr.day_name(), np_1darray_object), BaseStringArray)

	s.clip(lower=lower, axis=1) # type: ignore[call-overload] # pyright: ignore[reportCallIssue, reportArgumentType]
	s.clip(lower=lower, axis=1) # type: ignore[call-overload] # pyright: ignore[reportCallIssue,reportArgumentType]

Uh oh!

GH1641 Pandas 3.0 support #1643

Are you sure you want to change the base?

GH1641 Pandas 3.0 support #1643

Uh oh!

Conversation

loicdiridollou commented Jan 22, 2026

Uh oh!

loicdiridollou commented Jan 22, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

loicdiridollou Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Dr-Irv commented Jan 24, 2026

Uh oh!

loicdiridollou commented Jan 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

loicdiridollou Jan 24, 2026 •

edited

Loading