`Display` trait associated with the imaginary part of underflow negative numbers

[seanxnie]

Hi, I have encountered an issue when printing an imaginary part of underflow negative numbers using the Display trait when a floating point number is underflow. Here is an example code:

use num::complex::Complex64;
fn main() {
    let cn = Complex64::new(2.0, -1.0e-330);
    println!("{:}", cn);
}

The output is
2+-0i

The output of the imaginary part is unreasonable due to the +- sign. The expected output is 2-0i or 2+0i. This is caused by the conditional statement $im < $T::zero() of fmt_re_im function in write_complex!() macro in lib.rs file. When $im is underflow, the statement is equal, but the Display trait of floating point numbers still outputs the negative sign. Maybe $im <= $T::zero() is a better choice. I'm hoping for a better idea for this issue. Thanks.

Sean

[redweasel]

Can confirm, although simply -0.0 is enough to confuse it, as that is actually the floating point -0 which is equal in value to 0, but has a different representation. It can be reliably checked for using the signum or copysign functions.