Error when impostors list is empty

[varunagrawal]

Numpy throws an error in np.hstack(impostors) when impostors is empty.
What would be the recommended way to fix this?

[perimosocordiae]

If the impostors list is empty, the rest of the algorithm doesn't make much sense to run. That being said, the code should handle that case more gracefully. A simple check before calling hstack could avoid the error, but the main fit() function should handle the lack of impostors in some way as well.

[varunagrawal]

Exactly! If a particular dataset happens to provide no impostors, shouldn't the code just display the appropriate nearest neighbors?

[varunagrawal]

Hi,
Still waiting for a suggestion on the fix. I can work on the PR once I get the suggestion.

[perimosocordiae]

Sorry for the late response. In the fit method, we could check the size of impostors and raise a warning if it's empty. At that point, it probably makes sense to return early.

[anirudt]

@perimosocordiae : Return early with the linear transformation as an identity matrix, i.e. do nothing, right?

[perimosocordiae]

Yep, that makes sense.