Simple Rational Numbers Library for Modern C++ in a Single File

Floating point value types such as float, double and long double are usually represented according to the IEEE-754 specification, which requires them to store the exponent and mantissa of the number as binary numbers (instead of say, storing them as integers, like the decimal type in .NET). This leads to loss of precision and is sometimes error prone. (For example, 0.1 * 3 == 0.3 is false). That is why, this library represents numbers as ratios (with a numerator and a denominator). This leads to a correct representation of the values. The numerator and denominator are both of type ptrdiff_t by default (platform-sized integers), but this can be changed by declaring Ratio to use a different type (for example long).

The short answer is no. There really is no reason use this library. I simply made it as a side-thing. If precision is a concern, you will be better off with something like the GNU Multiple Precision Library (GMP). It implements numbers as a dynamic sequence of digits with a dynamically set exponent and is a battle-tested and well-optimized piece of software (unlike what you are currently looking at). There are also proposals for including a decimal128 type in the next version of C++, which might be a better choice still (it is simillar to decimal in .NET). There are free software implementations of in on the Internet. Fixed-point arithmetic is can also be a good option in some cases.

using namespace rational;

Rational x = 1; // equals 1 / 1
Rational y (1, 2); // equals 0.5
Rational z (10, 3); // equals 3.(3) or 3.3333..

Rational xPlusY = x + y; // equals 3 / 2 or 1.5

Rational result = xPlusY * z; // equals 30 / 6 or 5 / 1

std::cout << result; // prints "5/1"