Sheehan Olver

actually that's an "Illegal Instruction" so might be a different bug

@andreasnoack what was the motivation for having the diagonal repeated twice?

But Taylor series are not polynomials. And the Issue makes clear we are discussing the setting where they are used for autodiff.

Your example the series goes on forever so is not a polynomial. In this package it's `a_0 + a_1 x + … + a_n x^n + O(x^{n+1})` which is certainly...

Might point is if you called it ε instead of ` x ` you would have no issue with ` isless ` … that is when it’s infinitesimally small. but...

> DualNumbers isn't much used the plan at one point was to have ` ForwardDiff.Dual ` to move to DualNumbers.jl. Just would need someone to take the initiative

(btw, the use of “Taylor series” in this package seems like a misnomer: that term implies convergence in a neighbourhood of 0. The actual data structure seems to actually represent...

1 + ε^2 is greater than 1

A truncated Taylor series does not have the O(x^n) part. Neither does a truncated Fourier

What's a weak dependency?