Finding a reasonably accurate fractional approximation to a number, even allowing a certain degree of suboptimality in the estimation, is not computationally ef

Finding a reasonably accurate fractional approximation to a number, even allowing a certain degree of suboptimality in the estimation, is not computationally efficient. So, computing with fractions almost immediately becomes computing with bigints, unless you get lucky and stuff cancels out.