Fix complex.__repr__ handling of nan - #3184

Inside float_ops we have to_string that takes an f64 and returns it nicely formatted (it also includes handling of the NaN). The main issue is the sign we add, if its a nan for imag we must always prefix with a "+" (granted that a real part exists).

So I've made the change you suggested as:

if re == 0.0 {
    format!("{}j", float_ops::to_string(im))
} else if im.is_nan() {
    format!("({}+nanj)", float_ops::to_string(re))
} else {
    format!("({}{:+}j)", float_ops::to_string(re), float_ops::to_string(im))
}

It covered examples given in the issue:

>>>>> complex(0, float('nan'))
nanj
>>>>> complex(float('nan'), float('nan'))
(nan+nanj)
>>>>> complex(float('nan'), -float('nan'))
(nan+nanj)
>>>>> complex(-float('nan'), float('nan'))
(nan+nanj)

However, when I tested it with some random complex numbers, it does not go as planned:

>>>>> 10j
10.0j
>>>>> 10j
10.0j
>>>>> 1-10j
(1.0-10.0j)
>>>>> complex(10, float('10'))
(10.010.0j)
>>>>> complex(-10, float('10'))
(-10.010.0j)

I know that {:+} does not work because given value is not numerical, it can be fixed with checking if the im part is greater than 0 and add plus sign but it does not feel right.

Second to that, float_ops::to_string serialize it as float always. For example; in current -messy- implementation 10j will be 10j but new one will print it as 10.0j.

Right now, I am little bit confused how to proceed (maybe I misunderstood something?) . Open to any suggestion 😿

You can forward the f64 to floats_ops::to_string only if it has a fractional part (f64::fract) that is != 0.0.

@mdegis are you working on this issue? otherwise we will merge #3228

I've done some work on this issue but using to_lowercase() method is much simple approach than forwarding the f64 to floats_ops::to_string only if it has a fractional part (f64::fract) that is != 0.0.

I do not want to steal any ones work so, in the name of simplicity #3228 can be merged.

Thank you for considerate response. Do you want another issue to work on?

Thank you for considerate response. Do you want another issue to work on?

No, thank you.