I guess I was reasoning just in a mathematical sense. That does sound weird but I understand nothing stops me to do something like that.
With respect to type stability, I understand float() would solve the problem, but I always try to avoid conversion when I can. In particular, I like to write functions that are as general as possible, and in this example this would create problems if the function f were defined only for some specific type.
For instance, if I define f(x::Rational) = x^2 - 1, with the line x1, x2 = float(x1), float(x2) I get
julia> root_1d(f,1//30,10//3)
ERROR: MethodError: no method matching f(::Float64)
while commenting it out generates the right result (or close to it).
What I’d like to do here is get a type stable version of the above but without having to restrict it to use floats.
P.S. how do I add quotes of previous posts?