Currently integers to floats “automatic” conversion is successful only if the type is not a Union but it fails otherwise:
a = Float64[1.0,2.0]
a = Float64[1,2] # ok
a = Union{Float64,String}[1.0,"aaa"]
a = Union{Float64,String}[1,"aaa"] # MethodError
From a user-centric point of view, is there a way that integer to float could be handled automatically even in cases of Unions without ambiguity or performance losses ?