if I gather it right, then what you propose is basically a Computer Algebra System (i.e., a system for manipulating reals by symbolic expressions). What specifically would the goal of the library be?
It is not the arithmetic that is different to other approaches, but the way the reals are defined. Reals in this context are functions generating digits. Arithmetic operations are composition of these functions to obtain digits. And comparison operations iterate the digits of the operands for deciding. This has (in theory) infinite precision, in practice we can limit it to a few thousands or millions. A lot more than what double can provide. Notice I’m not including in my proposal any transcendental function. My scope now is only the four basic arithmetic operations, and comparison operators. The algorithms are not something new, they were researched and published decades ago. I just want to reimplement them because I never saw them in the form of open source modern c++. I have a good use case for this in the area of simulation, but I’m sure there is a lot more places you can use them, e.g. physics computations, geometrical computations, etc…