This allows the floating-point formatter to use the same functions as the integral formatter. This was tested in D114001.