I am experiencing an issue with the latest version of DynamicPolynomials (0.6) when I evaluate a polynomial with BigFloat coefficients at another BigFloat. I isolated a simple example that shows that I get different results if I set the variable c to Float64(0) or BigFloat(0).
julia> using DynamicPolynomials
julia> @polyvar a b c
(a, b, c)
julia> F = BigFloat(5)*a*b*c + BigFloat(1)*a*b^2
5.0abc + ab²
julia> DynamicPolynomials.subs(F, c=>BigFloat(0))
0.0
julia> DynamicPolynomials.subs(F, c=>0)
ab²
On the contrary, for the version 0.4.6 of the package I get in both cases ab² as a result.
*I tried as well with the same polynomial and coefficient types Float64, and I get the same results
Does anyone know what is the reason please ? Are BigFloats not compatible with DynamicPolynomials ?