TST Extend tests for scipy.sparse/*array in sklearn/neighbors/tests/test_neighbors

[Charlie-XIAO]

Towards #27090.

Made some (maybe undesired) changes to pass the tests. Maybe I shouldn't have done so? Please let me know if this is not the proper way to go.

[github-actions]

✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: ac8253e. Link to the linter CI: here}

[glemaitre]

Otherwise, LGTM.

[Charlie-XIAO]

@glemaitre There change seems to (possibly) affect these public functionality:

• metrics.pairwise_distances when metric is a callable: X, Y can be sparse arrays
• metrics.pairwise.pairwise_kernels when metric is a callable: X, Y can be sparse arrays
• cluster.compute_optics_graph when metric is a callable: X can be sparse array
• cluster.HDBSCAN.fit when metric is a callable and when one of the following holds: (1) algorithm is "brute" (2) algorithm is "auto" (and X is sparse) (3) store_centers is not None: X can be sparse array
• manifold.trustworthiness when metric is a callable: X can be sparse array
• manifold.TSNE.fit when metric is a callable and method is "exact": X can be sparse array
• manifold.TSNE.fit_transform when metric is a callable and method is "exact": X can be sparse array
• metrics.pairwise_distances_chuncked when metric is a callable: X, Y can be sparse arrays
• manifold.Isomap.fit when metric is a callable: X can be sparse array
• manifold.Isomap.fit_transform when metric is a callable: X can be sparse array
• cluster.OPTICS.fit when metric is a callable: X can be sparse array

However I still need to test a bit so as to confirm that these functionality indeed changed.

[glemaitre]

I see a list growing. I assume that we should use a common entry in the changelog stating the estimators supporting the sparse arrays and add all PRs number and contributors in this entry.
You can ignore it for the moment. I will try to that at some point in main.

[Charlie-XIAO]

@glemaitre Seems that you have created that common entry in on main. I have now updated the changelog there, and validated that the functions and classed mentioned are indeed affected. Would you please take a look?

On the main branch:

===============================================================================================================================
                                                       Testing functions
===============================================================================================================================

`cluster.compute_optics_graph` when `metric` is callable and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`manifold.trustworthiness` when `metric` is callable and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`metrics.pairwise_distances` when `metric` is callable and `X` and `Y` are sparrays    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`metrics.pairwise_distances_chunked` when `metric` is callable and `X` and `Y` are sparrays    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`metrics.pairwise.pairwise_kernels` when `metric` is callable and `X` and `Y` are sparrays    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`

===============================================================================================================================
                                                        Testing classes
===============================================================================================================================

`cluster.HDBSCAN` when `metric` is callable, `algorithm` is 'brute' or 'auto', and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`cluster.OPTICS` when `metric` is callable and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`manifold.Isomap` when `metric` is callable and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`
`manifold.TSNE` when `metric` is callable and `X` is sparray    NotImplementedError: We have not yet implemented 1D sparse slices; please index using explicit indices, e.g. `x[:, [0]]`

In this PR:

===============================================================================================================================
                                                       Testing functions
===============================================================================================================================

`cluster.compute_optics_graph` when `metric` is callable and `X` is sparray    Passed
`manifold.trustworthiness` when `metric` is callable and `X` is sparray    Passed
`metrics.pairwise_distances` when `metric` is callable and `X` and `Y` are sparrays    Passed
`metrics.pairwise_distances_chunked` when `metric` is callable and `X` and `Y` are sparrays    Passed
`metrics.pairwise.pairwise_kernels` when `metric` is callable and `X` and `Y` are sparrays    Passed

===============================================================================================================================
                                                        Testing classes
===============================================================================================================================

`cluster.HDBSCAN` when `metric` is callable, `algorithm` is 'brute' or 'auto', and `X` is sparray    Passed
`cluster.OPTICS` when `metric` is callable and `X` is sparray    Passed
`manifold.Isomap` when `metric` is callable and `X` is sparray    Passed
`manifold.TSNE` when `metric` is callable and `X` is sparray    Passed

[glemaitre]

Thanks for updating the changelog.
It looks good to me.

[glemaitre]

Maybe @OmarManzoor you want to have a look.

[OmarManzoor]

Otherwise LGTM. Thanks @Charlie-XIAO