It should be possible to write fully precise reals using mach.text.numeric.floats.writefloat

[pineapplemachine]

Right now the implementation of Burger's algorithm used to serialize floats only accepts doubles. To make it also accept reals should only mean changing some constants depending on the floating point type being serialized.

As part of this addition, it should also be possible to make the implementation more optimal for the float type.

[pineapplemachine]

This issue partially addressed by the addition of mach.text.numeric.hexfloats which, though it outputs data in a different format, does not suffer from any such limitation.