I want to implement a scoring function which returns float like values (but never NaN etc.) and find the minimum values of a list with min_by. As min_by requires its values to implement core::cmp::Ord, I can't use f32 and f64 (which don't have a total ordering).
Is there any floating point alternative in the library implementing core::cmp:Ord and if not, how do I build an alternative that excludes non-real values?