Quaternions has no support for autodiff or symbolic expressions

[baggepinnen]

Definitions like Quaternion(s::Real, v1::Real, v2::Real, v3::Real, n::Bool = false) prevents the use of packages like DualNumbers to perform automatic differentiation.
http://www.juliadiff.org/

The type specification ::Real further makes it impossible to use any kind of symbolic math software.

This could be worth considering to make the package useful in more situations.

[SimonDanisch]

Pull requests to change this are very welcome and easy to write :)

[bzinberg]

Definitions like Quaternion(s::Real, v1::Real, v2::Real, v3::Real, n::Bool = false) prevents the use of packages like DualNumbers to perform automatic differentiation.

Is this actually true? I tried to use ForwardDiff with Quaternions and only found one obstacle (granted, I didn't do thorough testing, it was a means to an end). In particular, ForwardDiff.Dual is a subtype of Real.

The issue I ran into is this line in Quaternions.qrotation:

Quaternions.jl/src/Quaternion.jl

Line 197 in 08d7fa7

thetaT = convert(eltype(u), theta)

I was trying to differentiate with respect to theta, and u was a Vector{Float64}, so I got the (correct) error that theta could not be converted to a Float64.

I don't see any reason this conversion needs to happen. Indeed, autodiff worked for me after overriding the method to not have that line.

I'll try to send in a PR.

[baggepinnen]

You're right, the Real annotation could not have been the problem..

[hyrodium]

I'll close this issue because this seems stale. Please comment here if you have further thoughts.