Ross Barnowski

Thanks @AlexDFischer , this gives me a stronger sense of how we might organize things: > kasteleyn_orientation is only relevant for planar graphs In that case, I wonder if it...

> However, this approach doesn't preserve backward compatibility with the current behavior of johnson, which returns paths by default. Whether or not things are backward compatible will depend on the...

A gentle nudge here - Fedora 43 is out (as of 10/25) with Python3.14 as the default interpreter!

Gentle ping @dinzim - have you had a chance to digest the above comments?

I agree that `tree/distance_measures` (now that it exists) is a natural place for this to live. Unfortunately we now have to weigh that against breaking backward compatibility. I think the...

Yes, floating point numbers may cause issues here. Unfortunately this is inherent to floating point representations. The [general recommendation](https://networkx.org/documentation/latest/tutorial.html#floating-point-considerations) is to convert values to integers to the degree of precision...

> Basically the difference is the importance of minimality of 0.5 edges, intuativly 1 weight edges are much more practical but sometimes you need to work with 0.5, so you...

Python 3.14.2 was just released so this should now be resolved. Let's leave open/pinned for now for those still affected. I think NX should also explicitly blocklist 3.14.1 in our...

Quick status update: NetworkX 3.6.1 has been released with Python 3.14.1 explicitly blocklisted, so attempting to install NX 3.6.1 with Python 3.14.1 will fail at install time. I propose to...

> Before I dive in, could you point me in the right direction? Are there any specific algorithmic approaches you had in mind when you suggested this? Nope! Nothing specific...