The "then" makes it seem like we should mention the order list e.g.,
"Clusters are then extracted from the order list using...", though not sure what the right term is here

going into that level of details will be hard, and it'll make this docstring very long, I'm not sure if we should do that here.

What about just removing the "then" then?

"Clusters are extracted using a DBSCAN-like method ... "

Though:

"Clusters are then extracted order list using a DBSCAN-like method"

isn't that much longer. Or are you saying that if we use the term "order list" we need to explain it further?

for me "then" here implies "as the next step", but sure, you're the native speaker here 😁

No I agree that "then" means "as the next step", I was trying to add the info of "next step after what"?

I know above we mention "Unlike DBSCAN, it keeps cluster hierarchy for a variable neighborhood radius", but then we talk about some other stuff, and it is not obvious what step happens before cluster extraction? i.e., what is the "then" following?

In the paragraph below (can't add this comment that far down), does the sentence:

" This implementation deviates from the original OPTICS by first performing
k-nearest-neighborhood searches on all points to identify core sizes, then
computing only the distances to unprocessed points when constructing the
cluster order."

mean that the original OPTICS algorithm does not only compute unprocessed points, or is it saying that the KNN part differs and this is just the step that follows (but does not differ from the original OPTICS)?

the original OPTICS algorithm proposed in the paper calculates core distance and reachability distances for all objects as a first step, which we don't. It's not easy to explain really, I had a hard time converting that to code.

I guess this comment is not for the average user, and rather for somebody picky who reads the paper and then comes looks at the implementation here. You could argue it could be a comment in the code maybe.

Ahh I think I understand the meaning now.

Does "unprocessed point" mean the same in both implementations? Does the original OPTICS algorithm re-compute reachability distances, updating if they are smaller? i.e., is the second part of the algorithm the same here as in the original paper?

Oh I'd have to read the paper and compare with implementation to be able to answer that, can't really confidently respond to this question easily

Yeah fair, I just looked at it and am still confused 🙃