Get ready for pandas 3 copy-on-write

[dcherian]

What is your issue?

This line fails with pd.set_options("mode.copy_on_write", True)

xarray/xarray/tests/__init__.py

Line 329 in c9d3084

assert all(obj.data.flags.writeable for obj in obj.variables.values())

We'll need to fix this before Pandas 3 is released in April:

xarray/xarray/tests/__init__.py

Line 329 in c9d3084

assert all(obj.data.flags.writeable for obj in obj.variables.values())

Here's a test

def example():
    obj = Dataset()
    obj["dim2"] = ("dim2", 0.5 * np.arange(9))
    obj["time"] = ("time", pd.date_range("2000-01-01", periods=20)
    print({k: v.data.flags for k, v in obj.variables.items()})
    return obj

example()
pd.set_options("mode.copy_on_write", True)
example()

[phofl]

I would recommend to either copy the data they come from a pandas object or set the flags to writeable if you are ok with the implications or if you don't modify the data inplace anyway. That said, modifying the underlying data of an Index is a really bad idea if you or your users want to re-use the Index. That can lead to segfaults among other things

Be aware if you set the flag to writeable, this is not possible in all scenarios, e.g.

idx = pd.Index([1, 2, 3], dtype="int64[pyarrow]")
vals = idx.to_numpy()
vals.flags.writeable = True

This will raise, arrow protects against this and does not allow you to set the flag to writeable

[dcherian]

Thanks @phofl I suspect this assertion is overly broad and we didn't want to do it for Index-backed Xarray variables.