You will have problems with this due to the limited precision available to real numbers. For example, 1/3 and 1/10 cannot be exactly represented in binary floating point arithmetic.
A first step would be converting the floating point number into a rational, but doing that is risky. How, for example, do you know for sure that 0.52941176470588235947190014485386200249195098876953125 is intended as 9/17 and not as 4768517252509937/9007199254740992 ?
My recommendation would be not to do what you are trying to do. You would be better off using a pair of rationals, such as [-3, 2] or [9, 17] instead of -1.5 and 0.52941176470588235947190014485386200249195098876953125
