Up to this point the c8::natural class has been a little different to the integer and rational classes. Unlike the other two it has supported c8::natural and c8::natural_digit operands. The thinking was that the c8::natural_digit versions would be significantly quicker than the c8::natural variants.

While this approach might have been correct early on in the development cycle, it hasn’t actually been correct for some time. The generic versions now handle special cases efficiently, and the overheads associated with natural number construction have been dramatically reduced. Removing the c8::natural_digit operations gives us a design that’s much cleaner because it avoids leaking something that was really intended to be in an internal implementation detail. The result is much easier to understand and use.

Having made this initial simplification, the naming of some of the digit array processing functions really needed to change. The new names reflect the numbers of digits of each source operand, and this sets the scene for even more specialized versions to be added over time.