pub fn to_float(arg: &Value) -> Result<f32, String>
Convert arg to a float and do some error checks (without casting)