>
>
>I don't know how to go
>about that. Be that as it may, cf should handle special cases
>for which it knows particular algorithms, and fall back on something
>like cf(bfloat(expr)) in the general case.
>
Square roots are a special case as their cf expansion is periodic. So
there is
a "finite result". Bfloat computations would not detect this.
Elements of a more general numberfield can be treated symbolically as
well, i.e.
one can do it with integer arithmetic (except for the floor function).
However I think one still needs to compute an infinite number of terms.
So in that case there seems to be no advantage of using integer arithmetic.
Michel