>
> Yes that's certainly true, but it is useful to be able to cut and
> paste some floating-point number produced in some hairy 2D display and
> use it later.
Understood. But if the user selects to display fewer digits, plain old
cut&paste has no hope of working here. Right now, there isn't even a way of
*specifying* that you want as many digits as necessary to guarantee
re-readability, except specifying a too-high fpprec.
Yes. We should rethink this. And perhaps follow what Macsyma does.
What does Macsyma do?
What do you mean when you say bfloat doesn't have a decimal output
> form that guarantees re-readibility?
Calculate 1.0b0/3 - XXX, where XXX is the cut&pasted copy of the result of
1.0b0/3. The result is not zero.