Alejandro Candioti
Alejandro Candioti
> Doing so would change behavior for weighted graphs. When edges have a `"weight"` attribute, it is used in calculating the distances - I don't think that is the intention...
-removed / comment doesn't apply-
> > You could pick one method or the other dynamically based on the number of edges. The problem is that number_of_edges is not cached so it is O(E). >...
Would you share the logic you used for selection? Although adding auto should go in a separate PR. I'd imagine you just take random intervals and try to estimate density?...
Good job, I approve this PR. @dschult would you take a look at this one when you have time?
Docs look good to me. You don't need to explain it on super detail there. You can leave that for the docs page in a separate PR. We could add...
Nice! Be careful with those random graphs though. They tend to be disconnected and their shortest paths be of only a few hops (exactly taylored for the skip infinity optimization)....
> Thank you @amcandio and @dschult for guiding throughout the PR > > some tasks I am planning to work on next: > > 1. Currently the Floyd warshall (FW)...
A bit late to the party here but @Aditya-Shandilya1182 can you add unit tests that would fail without this fix?
Thanks for the PR! I don’t think this approach should be added to NetworkX itself. It adds complexity and doesn’t fully solve the problem—many algorithms beyond the main wrappers would...