pandas breaking changes

3.1.1

5 breaking changes, first published November 11, 2026.

• Signature changed pandas.tests.indexes.datetimes.methods.test_tz_localize.test_dti_tz_localize_nonexistent_shift_past_last_transition
  Was def test_dti_tz_localize_nonexistent_shift_past_last_transition(); now def test_dti_tz_localize_nonexistent_shift_past_last_transition(tz). pypi.org
• Signature changed pandas.tests.io.test_common.TestCommonIOCapabilities.test_read_fspath_all
  Was def test_read_fspath_all(self, reader, module, path, datapath); now def test_read_fspath_all(self, reader, module, path, path_cls, datapath). pypi.org
• Signature changed pandas.tests.io.test_common.TestCommonIOCapabilities.test_stringify_path_fspath
  Was def test_stringify_path_fspath(self); now def test_stringify_path_fspath(self, path_cls). pypi.org
• Signature changed pandas.tests.io.test_common.TestCommonIOCapabilities.test_write_fspath_all
  Was def test_write_fspath_all(self, writer_name, writer_kwargs, module, tmp_path); now def test_write_fspath_all(self, writer_name, writer_kwargs, module, path_cls, tmp_path). pypi.org
• Signature changed pandas.tests.io.test_common.TestCommonIOCapabilities.test_write_fspath_hdf5
  Was def test_write_fspath_hdf5(self, tmp_path); now def test_write_fspath_hdf5(self, path_cls, tmp_path). pypi.org

3.1.0

992 breaking changes, first published October 22, 2026.

• Signature changed pandas.api.extensions.ExtensionArray.isna
  Was def isna(self) -> np.ndarray | ExtensionArraySupportsAnyAll; now def isna(self) -> np.ndarray | ExtensionArrayNaResult. pypi.org
• Signature changed pandas.api.types.union_categoricals
  Was def union_categoricals(to_union, sort_categories: bool = False, ignore_order: bool = False) -> Categorical; now def union_categoricals(to_union: Sequence[CategoricalIndex | Series | Categorical], sort_categories: bool = False... pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.__init__
  Was def __init__(self, obj: NDFrameT, keys: _KeysArgType | None = None, level: IndexLabel | None = None, grouper: ops.BaseGrouper | None = None, exclusions: frozenset[Hashable] | None = None, selection: IndexLabel | None = None, as_index... pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.aggregate
  Was def aggregate(self, func=None, *args, engine=None, engine_kwargs=None, **kwargs); now def aggregate(self, func=None, *args, engine=None, engine_kwargs=None, **kwargs) -> DataFrame. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.cummax
  Was def cummax(self, numeric_only: bool = False, **kwargs) -> NDFrameT; now def cummax(self, numeric_only: bool = False, skipna: bool = True, **kwargs) -> NDFrameT. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.cummin
  Was def cummin(self, numeric_only: bool = False, **kwargs) -> NDFrameT; now def cummin(self, numeric_only: bool = False, skipna: bool = True, **kwargs) -> NDFrameT. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.cumprod
  Was def cumprod(self, numeric_only: bool = False, *args, **kwargs) -> NDFrameT; now def cumprod(self, numeric_only: bool = False, skipna: bool = True, *args, **kwargs) -> NDFrameT. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.cumsum
  Was def cumsum(self, numeric_only: bool = False, *args, **kwargs) -> NDFrameT; now def cumsum(self, numeric_only: bool = False, skipna: bool = True, *args, **kwargs) -> NDFrameT. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.describe
  Was def describe(self, percentiles=None, include=None, exclude=None) -> NDFrameT; now def describe(self, percentiles=None, include=None, exclude=None) -> DataFrame. pypi.org
• Signature changed pandas.api.typing.DataFrameGroupBy.ewm
  Was def ewm(self, com: float | None = None, span: float | None = None, halflife: float | str | Timedelta | None = None, alpha: float | None = None, min_periods: int | None = 0, adjust: bool = True, ignore_na: bool = False, times... pypi.org

And 982 more in this version.

3.0.6

1 breaking change, first published September 17, 2026.

• Signature changed pandas.RangeIndex.searchsorted
  Was def searchsorted(self, value, side: Literal['left', 'right'] = 'left', sorter: NumpySorter | None = None) -> npt.NDArray[np.intp] | np.intp; now def searchsorted(self, value: NumpyValueArrayLike | ExtensionArray, side: Literal['left'... pypi.org

3.0.3

1 breaking change, first published May 11, 2026.

• Signature changed pandas.io.common.IOArgs.__init__
  Was def __init__(self, filepath_or_buffer: str | BaseBuffer, encoding: str, mode: str, compression: CompressionDict, should_close: bool = False) -> None; now def __init__(self, filepath_or_buffer: str | BaseBuffer, encoding: str, mode... pypi.org

3.0.2

3 breaking changes, first published March 31, 2026.

• Signature changed pandas.RangeIndex.sort_values
  Was def sort_values(self, *, return_indexer: bool = False, ascending: bool = True, na_position: NaPosition = 'last', key: Callable | None = None) -> Self | tuple[Self, np.ndarray | RangeIndex]; now def sort_values(self, *, return_indexer... pypi.org
• Signature changed pandas.Timedelta.to_numpy
  Was def to_numpy(self, dtype: npt.DTypeLike = ..., copy: bool = False) -> np.timedelta64; now def to_numpy(self, dtype: npt.DTypeLike | None = ..., copy: bool = False) -> np.timedelta64. pypi.org
• Signature changed pandas.Timestamp.fromisoformat
  Was @classmethod def fromisoformat(cls, date_string: str) -> Self; now @classmethod def fromisoformat(cls, date_string: str, /) -> Self. pypi.org

3.0.0

674 breaking changes, first published January 21, 2026.

• Removed
  pandas 3.0.0 removed features that were deprecated in previous releases. Upgrade to pandas 2.3 and fix warnings before moving to 3.0. github.com
• Behavior changed
  pandas 3.0.0 supports Python 3.11 and higher, so older Python versions can no longer install it. github.com
• Behavior changed
  Copy-on-Write gives consistent copy and view semantics and removes SettingWithCopyWarning, so code relying on chained assignment views may change results. github.com
• Behavior changed
  Strings now use a dedicated string data type by default, which can change dtypes and behavior of existing string columns. github.com
• Default changed
  Datetime-like data now infers a new default resolution, which can change results for existing code. github.com
• Signature changed pandas.api.extensions.ExtensionArray.fillna
  Was def fillna(self, value: object | ArrayLike | None = None, method: FillnaOptions | None = None, limit: int | None = None, copy: bool = True) -> Self; now def fillna(self, value: object | ArrayLike, limit: int | None = None, copy: bool =... pypi.org
• Signature changed pandas.api.extensions.ExtensionArray.map
  Was def map(self, mapper, na_action=None); now def map(self, mapper, na_action: Literal['ignore'] | None = None). pypi.org
• Signature changed pandas.api.extensions.ExtensionArray.ravel
  Was def ravel(self, order: Literal['C', 'F', 'A', 'K'] | None = 'C') -> ExtensionArray; now def ravel(self, order: Literal['C', 'F', 'A', 'K'] | None = 'C') -> Self. pypi.org
• Signature changed pandas.api.extensions.ExtensionArray.T
  Was T: ExtensionArray; now T: Self. pypi.org
• Signature changed pandas.api.extensions.ExtensionArray.transpose
  Was def transpose(self, *axes: int) -> ExtensionArray; now def transpose(self, *axes: int) -> Self. pypi.org

And 664 more in this version.

Check your code

These are the changes. To see which of them reach your code, scan a public repository, or have your coding agent call check_code with your manifest or lockfile over MCP. The API reads every change in every version, not only the newest. A fix pull request opens for the changes that reach a connected repository. Nothing merges or deploys.