Tune performance of angular <-> cartesian conversions

[chrisnatali]

No description provided.

[chrisnatali]

Happens much faster in bulk.

[vr2262]

I assume you mean the functions in here? /networker/geomath.py

[chrisnatali]

Yah. The spherical point project on segment function was much slower than the corresponding cartesian point projection.

A lot had to do with the angular to 3-space coordinate conversion...it ran faster when done in bulk, but it was difficult to bulk them in the context of finding the nearest point on a segment. I tried.

So I decided to revert to cartesian based projection onto segment by default (even for angular coordinates) and add a configuration option for doing this spherically. This is how it was done in legacy versions and it hasn't introduced a lot of error thus far.

[vr2262]

So... is it worth continued effort to improve the spherical version?

[chrisnatali]

It's a low priority at this point