smcallis
smcallis
I have the same problem on Ubuntu 16.04 (gcc 7.2.0): ``` > make cd tests && ../flisp unittest.lsp fatal error: (type-error apply functionSegmentation fault (core dumped) Makefile:23: recipe for target...
Happy to bang out a quick change however you like, could you summarize the final thinking in that thread on how people would like this to work?
Any thoughts on merging this?
I think this likely amounts to honoring the :initform keyword when a given slot is unset after parsing the json, I'll try to write a pull request.
Submitted PR #2
We might consider a new method called ARPIST that was recently published too: https://arxiv.org/pdf/2201.00261.pdf I think I've found a set of points that has huge relative error with the current...
The issue was actually the opposite here, we weren't cancelling enough digits. Using `atan2(|AxB|, A*B)` to get the edge lengths is ill behaved when the vectors are nearly parallel. Kahan...
Sorry yes I provided the wrong points initially, it's the _x_2 points that were giving me trouble. It's definitely way at the boundaries of precision. We're actually not necessarily interested...
There's Euler's (other) formula as well which does no subtractions or tangents: > cos E = (cos α + cos β + cos c)/(4*cos(α/2)*cos(β/2)*cos(γ/2)) [ref (page 23)](http://fer3.com/arc/img/121573.cesaro-sph-gometry.pdf) Since α, β,...
Yeah I suppose if we think of it as $\cos (x) = 1 - \dfrac{x^2}{2}$ for small x, then it's not very friendly at all precision wise.