fix: allow arrays with Nones in single argument usage of ak.where

[ikrommyd]

Fixes #3138

This used to use ak.operations.to_numpy which allowed Nones to go through until #2757 changed it to use to_backend_array with allow_missing=False. I don't think there is any harm in setting allow_missing=True here.

[ikrommyd]

All these tests error on main with ValueError: Content.to_nplike cannot convert 'None' values to np.ma.MaskedArray unless the 'allow_missing' parameter is set to True

[pfackeldey]

This change makes sense to me, but I seem to not understand yet why it was changed in #2757? Maybe there was a reason for this?

[ikrommyd]

This change makes sense to me, but I seem to not understand yet why it was changed in #2757? Maybe there was a reason for this?

@agoose77 is there a reason to not allow missing values for the single argument case of ak.where? I’m tagging since that was your PR.

[ikrommyd]

Discussed briefly with Angus at SciPy about this and the reasoning was probably that this was the safest thing to do when .attrs was added. However, .nonzero() seems to always returns an array with integers and it doesn't matter if the numpy array is masked or not so awkward can always convert it into an awkward array again so it should be safe to enable this.

[ianna]

@ikrommyd - Thanks for adding the tests! They look visually identical to NumPy on my end. It’d be nice to also have a direct comparison with NumPy.

[ikrommyd]

@ikrommyd - Thanks for adding the tests! They look visually identical to NumPy on my end. It’d be nice to also have a direct comparison with NumPy.

added that too! 😃

[ianna]

@ikrommyd - Thanks for adding the tests! They look visually identical to NumPy on my end. It’d be nice to also have a direct comparison with NumPy.

added that too! 😃

Thanks @ikrommyd - have you read my message on Slack? Thanks.

[ikrommyd]

Oh sorry. I just did no

@ikrommyd - Thanks for adding the tests! They look visually identical to NumPy on my end. It’d be nice to also have a direct comparison with NumPy.

added that too! 😃

Thanks @ikrommyd - have you read my message on Slack? Thanks.

Oh sorry. I just did now. I thought your comment here was a request to make this change. My bad.

[ianna]

@ikrommyd - Thanks! Please, see my comment above.

[ianna]

@ikrommyd - Thanks!