self.unique - different on GitHub and pipy file

Nobody has claimed this yet.

Hi!

I'm no expert on neither UMAP nor Python, but I have been using the UMAP extension on-and-off for the last 3 months. Previously I've had no error using it, but now a value error is appearing where I do not believe it should.

Using the "precomputed" metric, the following value error appears:
ValueError: unique is poorly defined on a precomputed metric

Looking at the code in umap_.py file on GitHub (https://github.com/lmcinnes/umap/blob/master/umap/umap_.py) row 1500 is states:

elif self.metric == "precomputed":
if self.unique:
raise ValueError("unique is poorly defined on a precomputed metric")

but in the pipy file (downloaded using pip install umap-learn, found https://pypi.org/project/umap-learn/) states:

elif self.metric == "precomputed":
if self.unique is False:
raise ValueError("unique is poorly defined on a precomputed metric")

The second statement seems contradictory to me. Since it says that unique is poorly defined for a precomputed metric it should logically be if self.unique like in the first case.

Is this an error?

Start by comparing the precomputed-metric branch in umap_.py on GitHub with the installed umap-learn package described in the issue. Check which released source and version contain each condition, then verify the intended behavior for self.unique; done means the discrepancy and correct condition are documented or resolved in the published package.

Written by the indexing model from the issue text.