[AryanRG]

Description

Selecting features from a stacked array can leave unused variable labels in its index. to_unstacked_dataset() then tries to select those missing variables and raises KeyError.

This removes unused labels when finding the variables to unstack. Tests cover NumPy and Dask arrays, different level selections, and preservation of the input and coordinate metadata.

Regression and nearby stacking tests pass, as do the configured pre-commit hooks. The local type check reports the same errors on the unchanged base.

Checklist

• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.
    Tools: Codex, used for implementation, tests, review and the description. Task: fix unstacking after feature selection while preserving the selected data and metadata.

[headtr1ck]

Thanks, this looks good! PandasMultiIndex.isel keeps pandas' unused levels, and unstack / from_dataframe already drop them with remove_unused_levels_categories, so doing the same here is consistent.

Could you add the PR number and your name to the whats-new entry, like the other entries?

- :py:meth:`DataArray.to_unstacked_dataset` no longer raises ``KeyError`` for
  unused variable labels left in the stacked index after selecting features
  (:pull:`11808`).
  By `Aryan R. Ganjoo <https://github.com/AryanRG>`_.

[AryanRG]

Added the PR number and author credit, thanks!

[Codex on behalf of AryanRG.]

[headtr1ck]

Thanks!

[welcome]

Congratulations on completing your first pull request! Welcome to Xarray! We are proud of you, and hope to see you again!