BUG: `ValueError: ndarray is not C-contiguous` for `cummax` on nullable dtypes

[MarcoGorelli]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

In [5]: df = pd.DataFrame({'a': [1,1,2], 'b': [4,5,6], 'i': [0,1,2]}, dtype='Int64')[::-1]

In [6]: df.groupby('a')['b'].cummax()
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[6], line 1
----> 1 df.groupby('a')['b'].cummax()

File ~/pandas-dev/pandas/core/groupby/groupby.py:4945, in GroupBy.cummax(self, numeric_only, **kwargs)
   4886 """
   4887 Cumulative max for each group.
   4888
   (...)
   4942 bull    6   9
   4943 """
   4944 skipna = kwargs.get("skipna", True)
-> 4945 return self._cython_transform(
   4946     "cummax", numeric_only=numeric_only, skipna=skipna
   4947 )

File ~/pandas-dev/pandas/core/groupby/generic.py:684, in SeriesGroupBy._cython_transform(self, how, numeric_only, **kwargs)
    681 obj = self._obj_with_exclusions
    683 try:
--> 684     result = self._grouper._cython_operation(
    685         "transform", obj._values, how, 0, **kwargs
    686     )
    687 except NotImplementedError as err:
    688     # e.g. test_groupby_raises_string
    689     raise TypeError(f"{how} is not supported for {obj.dtype} dtype") from err

File ~/pandas-dev/pandas/core/groupby/ops.py:932, in BaseGrouper._cython_operation(self, kind, values, how, axis, min_count, **kwargs)
    928 assert kind in ["transform", "aggregate"]
    930 cy_op = WrappedCythonOp(kind=kind, how=how, has_dropped_na=self.has_dropped_na)
--> 932 return cy_op.cython_operation(
    933     values=values,
    934     axis=axis,
    935     min_count=min_count,
    936     comp_ids=self.ids,
    937     ngroups=self.ngroups,
    938     **kwargs,
    939 )

File ~/pandas-dev/pandas/core/groupby/ops.py:546, in WrappedCythonOp.cython_operation(self, values, axis, min_count, comp_ids, ngroups, **kwargs)
    542 self._validate_axis(axis, values)
    544 if not isinstance(values, np.ndarray):
    545     # i.e. ExtensionArray
--> 546     return values._groupby_op(
    547         how=self.how,
    548         has_dropped_na=self.has_dropped_na,
    549         min_count=min_count,
    550         ngroups=ngroups,
    551         ids=comp_ids,
    552         **kwargs,
    553     )
    555 return self._cython_op_ndim_compat(
    556     values,
    557     min_count=min_count,
   (...)
    561     **kwargs,
    562 )

File ~/pandas-dev/pandas/core/arrays/masked.py:1602, in BaseMaskedArray._groupby_op(self, how, has_dropped_na, min_count, ngroups, ids, **kwargs)
   1599 if how == "rank" and kwargs.get("na_option") in ["top", "bottom"]:
   1600     result_mask[:] = False
-> 1602 res_values = op._cython_op_ndim_compat(
   1603     self._data,
   1604     min_count=min_count,
   1605     ngroups=ngroups,
   1606     comp_ids=ids,
   1607     mask=mask,
   1608     result_mask=result_mask,
   1609     **kwargs,
   1610 )
   1612 if op.how == "ohlc":
   1613     arity = op._cython_arity.get(op.how, 1)

File ~/pandas-dev/pandas/core/groupby/ops.py:331, in WrappedCythonOp._cython_op_ndim_compat(self, values, min_count, ngroups, comp_ids, mask, result_mask, **kwargs)
    329 if result_mask is not None:
    330     result_mask = result_mask[None, :]
--> 331 res = self._call_cython_op(
    332     values2d,
    333     min_count=min_count,
    334     ngroups=ngroups,
    335     comp_ids=comp_ids,
    336     mask=mask,
    337     result_mask=result_mask,
    338     **kwargs,
    339 )
    340 if res.shape[0] == 1:
    341     return res[0]

File ~/pandas-dev/pandas/core/groupby/ops.py:477, in WrappedCythonOp._call_cython_op(self, values, min_count, ngroups, comp_ids, mask, result_mask, **kwargs)
    474     if self.how != "rank":
    475         # TODO: should rank take result_mask?
    476         kwargs["result_mask"] = result_mask
--> 477     func(
    478         out=result,
    479         values=values,
    480         labels=comp_ids,
    481         ngroups=ngroups,
    482         is_datetimelike=is_datetimelike,
    483         mask=mask,
    484         **kwargs,
    485     )
    487 if self.kind == "aggregate" and self.how not in ["idxmin", "idxmax"]:
    488     # i.e. counts is defined.  Locations where count<min_count
    489     # need to have the result set to np.nan, which may require casting,
    490     # see GH#40767. For idxmin/idxmax is handled specially via post-processing
    491     if result.dtype.kind in "iu" and not is_datetimelike:
    492         # if the op keeps the int dtypes, we have to use 0

File groupby.pyx:2287, in pandas._libs.groupby.group_cummax()

File <stringsource>:663, in View.MemoryView.memoryview_cwrapper()

File <stringsource>:353, in View.MemoryView.memoryview.__cinit__()

ValueError: ndarray is not C-contiguous

Issue Description

groupby-cummax raises for nullable integers in this case

Note that for pyarrow-backed integers it works fine

Expected Behavior

2    6
1    5
0    5
Name: b, dtype: Int64

Installed Versions

INSTALLED VERSIONS

commit : b552dc9
python : 3.10.12
python-bits : 64
OS : Linux
OS-release : 5.15.167.4-microsoft-standard-WSL2
Version : #1 SMP Tue Nov 5 00:21:55 UTC 2024
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : C.UTF-8
LOCALE : en_US.UTF-8

pandas : 3.0.0.dev0+618.gb552dc95c9
numpy : 1.26.4
dateutil : 2.9.0.post0
pip : 25.0
Cython : 3.0.11
sphinx : 8.1.3
IPython : 8.32.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.13.3
blosc : None
bottleneck : 1.4.2
fastparquet : 2024.11.0
fsspec : 2025.2.0
html5lib : 1.1
hypothesis : 6.125.2
gcsfs : 2025.2.0
jinja2 : 3.1.5
lxml.etree : 5.3.0
matplotlib : 3.10.0
numba : 0.61.0
numexpr : 2.10.2
odfpy : None
openpyxl : 3.1.5
psycopg2 : 2.9.10
pymysql : 1.4.6
pyarrow : 19.0.0
pyreadstat : 1.2.8
pytest : 8.3.4
python-calamine : None
pytz : 2025.1
pyxlsb : 1.0.10
s3fs : 2025.2.0
scipy : 1.15.1
sqlalchemy : 2.0.37
tables : 3.10.1
tabulate : 0.9.0
xarray : 2024.9.0
xlrd : 2.0.1
xlsxwriter : 3.2.2
zstandard : 0.23.0
tzdata : 2025.1
qtpy : None
pyqt5 : None

Spotted in Narwhals

[ruoyeruolan]

The error is caused by the slice and it actually is not an error. When you use slicing operations like [::-1] in pandas/NumPy, it creates a view of the original data rather than a new array. This view has the following characteristics:

• It references the same memory as the original array
• It uses a negative stride to traverse the elements in reverse order
• It becomes non-contiguous in memory

1. C-contiguous: Data elements are stored sequentially in memory (row by row)
2. Non-contiguous: Data elements are not stored sequentially, requiring jumps in memory when accessing them in order

To solve it, just call .copy() on the dataframe.
It will allocates new memory and creates a fresh array with the data arranged in proper C-contiguous order

PS:

• PyArrow-backed arrays typically materialize a new copy when slicing operations occur
• This happens automatically behind the scenes, creating a properly formatted array

[MarcoGorelli]

shouldn't the copy be done (if necessary) on pandas' side?

note that cumsum and cumprod are both supported here

[rhshadrach]

The error is caused by the slice and it actually is not an error.

The OP is two valid operations, one followed by the other. That is a bug. That users can work around this bug is useful, but it still needs to be fixed.