BUG: unwrap setitem indexer for ArrowExtensionArray

[lukemanley]

• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/v2.0.0.rst file if fixing a bug or adding a new feature.

Setting values in a pyarrow-backed column with a non-scalar indexer was raising.

Fixes the following:

import pandas as pd

df = pd.DataFrame({"a": [1, 2, 3], "b": [4, 5, 6]}, dtype="int64[pyarrow]")

# IndexError: index 1 is out of bounds for axis 0 with size 1
df.loc[[0, 1], "b"] = 0

[phofl]

Not sure, but it might make more sense to handle this in setitem of the ArrowExtensionArray?

[lukemanley]

I was 50/50 on whether to put it here or in ArrowExtensionArray.setitem. I ended up leaning towards this as it seems consistent with this method and may be helpful beyond ArrowExtensionArray (e.g. third-party EAs). But happy to move it if setitem is a better place for it.

[phofl]

I am more concerned with our setiteim for Arrow diverging from the numpy setitem call, but not sure if this is correct either.

cc @jbrockmendel Could you weigh in here?

[jbrockmendel]

id want to handle it correctly in the ArrowArray method. could add a test to the extension base tests for handling this case correctly

[lukemanley]

ok, I will change this to be handled within ArrowExtensionArray.__setitem__.

I guess the typing on ExtensionArray.__setitem__ and ArrowExtensionArray.__setitem__ should change as well?

pandas/pandas/core/arrays/base.py

Line 351 in 1488157

def __setitem__(self, key: int | slice | np.ndarray, value: Any) -> None:

pandas/pandas/core/arrays/arrow/array.py

Line 878 in 1488157

def __setitem__(self, key: int | slice | np.ndarray, value: Any) -> None:

[phofl]

Yeah would make sense

[mroeschke]

Thanks @lukemanley