The type of the eps for complex numbers is currently complex. Thus, if we compare two complex numbers with FloatCmp::eq(a, b, 1e-15), this will be false, even if the respective real/imag entries are closer than 1e-15 to each other. IMHO, this is counterintuitive.
One way to fix this is to accept an arithmetic eps and forward this a a complex({eps, eps}) or to alwas enforce the eps to be an arithmetic type and apply this in all components.
The type of the eps for complex numbers is currently complex. Thus, if we compare two complex numbers with FloatCmp::eq(a, b, 1e-15), this will be false, even if the respective real/imag entries are closer than 1e-15 to each other. IMHO, this is counterintuitive.
One way to fix this is to accept an arithmetic eps and forward this a a complex({eps, eps}) or to alwas enforce the eps to be an arithmetic type and apply this in all components.