I need some feed back, could this become a standard? is it usable? should I make a Mandelbrot renderer to test it???
Any feedback is helpful
I don’t think this is entirely useful, for similar reasons as outlined here: http://www.boost.org/doc/libs/1_64_0/libs/rational/rational.html
There is no consensus on the best way to represent a floating point as a fraction and preserve all the properties people want. It’s not too difficult to add the “correct” method yourself for each application specifically.
For example, some people would want float_to_rational(1.0/3.0) to equal 1/3, whereas some people would want the literal float representation 33333333333333331/100000000000000000. (example from above)
The makers of boost::rational decided not to implement that single function given that problem.
So to answer your questions with my opinion:
No, this can’t become a standard because there’s already no universal agreement, for good reasons from all sides;
Yes it’s usable as long as you understand how the conversion works and any limitations (not having the most accurate representation of pi is a little worrisome for many applications)
It would be interesting to see a Mandelbrot renderer comparison I guess? Not essential for testing but why not 
float_to_rational(1.0/3.0) to equal 1/3