john verzani

It seems like a new method is needed to resolve this special case: ``` julia> ForwardDiff.derivative(x -> ℯ^x, 1/2) ERROR: MethodError: ^(::Irrational{:ℯ}, ::ForwardDiff.Dual{ForwardDiff.Tag{var"#1#2",Float64},Float64,1}) is ambiguous. Candidates: ^(::Irrational{:ℯ}, x::Number) in Base.MathConstants...

add test_composable to test suite

2

The tests in `test_composable.jl` are not run, to avoid installing dependencies not otherwise used. However, #162 showed that by running them bugs can be exposed.

This polynomial shows that multiplication for different polynomial types can vary significantly: ``` julia> rts = [1.0, sqrt(2), sqrt(3)]; julia> ls = [2, 3, 4]; julia> x, xx = variable(Polynomial{Float64}),...

CUDA

3

In https://discourse.julialang.org/t/using-polynomials-jl-with-cuda-jl/80580 we find the following hangs: ``` using CUDA, Polynomials a = CuVector([1.0f0, 1.0f0, 1.0f0]) Polynomial(a) ```

In #297 support for roots and fitting using BigFloat polynomials was added just by adding a dependency on `GenericLinearAlgebra`. Unfortunately, this little change causes downstream breakage (https://github.com/JuliaLinearAlgebra/GenericLinearAlgebra.jl/pull/71#issuecomment-743928205). Until that is...

This came up in #225 . Would be nice to duck type this value to operations associated with rings.

rework round and trunc

1

This introduces these semantics for round and trunc * both pass on non real values * trunc and round return Sym objects (this is a change) * round only uses...

This came up in https://discourse.julialang.org/t/how-to-round-whole-matrix-like-numpy-round/84897 What should be the proper semantics of `round` for `Sym` objects: * sympy.round() fails on symbolic objects (seems excessive, but what should the result of...

There are testing issues as the sympy that gets picked up by CI is now 1.10, not 1.9 and (hopefully) small differences have accumulated.