Function relative_error_with_floor 

pub fn relative_error_with_floor(
    numerical: f64,
    reference: f64,
    floor: f64,
) -> f64

Normalized relative error with an explicit absolute floor.

floor must be non-negative. Use this only when a problem has a meaningful non-default scale; otherwise prefer relative_error.

Examples

use solver::validation::relative_error_with_floor;

let e = relative_error_with_floor(0.0002, 0.0, 1e-3);
assert!((e - 0.2).abs() < 1e-15);