Revert .oindex and .vindex additions in _ElementwiseFunctionArray, NativeEndiannessArray, and BoolTypeArray classes

[andersy005]

As noted in #8909, the use of .oindex and .vindex properties in coding/* appears to have broken some backends (e.g. scipy). This PR reverts those changes. We plan to bundle these changes into a separate backends feature branch (see this comment, which will be merged once we are confident about its impact on downstream dependencies.

• ClosesScipyArrayWrapper' object has no attribute 'oindex' #8909
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

[andersy005]

@ocraft, when you get a chance do you mind try this branch and confirming whether this fixes your issue in #8909? as @dcherian noted, the minimal verifiable example you provided in #8909 (comment) wasn't reproducible on his end.

[ocraft]

Ok, I started from scratch and created venv with this set of packages (newest version of xarray and scipy with deps):

Package Version
--------------- -----------
numpy 1.26.4
packaging 24.0
pandas 2.2.1
pip 22.0.4
python-dateutil 2.9.0.post0
pytz 2024.1
scipy 1.13.0
setuptools 58.1.0
six 1.16.0
tzdata 2024.1
xarray 2024.3.0

and my example results in error as in #8909 (comment)
After uninstalling xarray and installing it from this pull request with

pip install https://github.com/pydata/xarray/pull/8921/commits/721aaf1f4dc88d8801717246941d9e63379c9aef

I got:

NotImplementedError: _ElementwiseFunctionArray._oindex_get method should be overridden

[andersy005]

@dcherian /@ocraft, i believe i've managed to revert the breaking changes. however, the temporary fixes turned out to be messier than i expected :(

[dcherian]

yuck but ok :)

[andersy005]

@ocraft, when you get a chance can you confirm this fully addresses the issue on your end?

[ocraft]

@ocraft, when you get a chance can you confirm this fully addresses the issue on your end?

I have this error: NotImplementedError: StackedBytesArray._oindex_get method should be overridde

[dcherian]

Can we add the reproducer as a test case please

[dcherian]

Can we add the reproducer as a test case please

@andersy005 can we add a test here please?

[andersy005]

Can we add the reproducer as a test case please

@andersy005 can we add a test here please?

i wanted to mention that, as i revisit this issue to add a test, i'm learning the temporary fix may become more complex than initially thought. should we attempt to fix this going forward (as part of the adding support for oindex/vindex to BackendArray) rather than reverting to the previous behavior?

[andersy005]

i can confirm that this issue is fully resolved on backend-indexing branch

In [1]: importxarrayasxr
...: ...: ds=xr.Dataset()
...: ds['A'] =xr.DataArray([[1, 'a'], [2,'b']],dims=['x','y'])
...: ds.to_netcdf('/tmp/test.nc')
...: ds2=xr.open_dataset('/tmp/test.nc')
...: ds2.sel(y=[1]).to_netcdf('/tmp/test.nc')
In [2]: In [2]: xr.show_versions()
INSTALLEDVERSIONS------------------commit: e96e70e04dc21607173d7cdcd9dba2c121d6aab6python: 3.12.3|packagedbyconda-forge| (main, Apr152024, 18:35:20) [Clang16.0.6 ]
python-bits: 64OS: DarwinOS-release: 23.4.0machine: arm64processor: armbyteorder: littleLC_ALL: NoneLANG: en_US.UTF-8LOCALE: ('en_US', 'UTF-8')
libhdf5: Nonelibnetcdf: Nonexarray: 2024.3.1.dev39+g04cca97a0pandas: 2.2.2numpy: 1.26.4scipy: 1.13.0netCDF4: Nonepydap: Noneh5netcdf: Noneh5py: Nonezarr: Nonecftime: Nonenc_time_axis: Noneiris: Nonebottleneck: Nonedask: Nonedistributed: Nonematplotlib: Nonecartopy: Noneseaborn: Nonenumbagg: Nonefsspec: Nonecupy: Nonepint: Nonesparse: Noneflox: Nonenumpy_groupies: Nonesetuptools: 69.5.1pip: 24.0conda: Nonepytest: Nonemypy: NoneIPython: 8.22.2sphinx: None