stencillogic / num-bigfloat Goto Github PK

Could num-traits be implemented for BigFloat with the use of a feature flag? This would make it useable for libraries that depend on it.

assert!(BigFloat::from(-1).frac() == BigFloat::from(0))
is not equal to true

-0 is not equal to 0

Is this a bug?

some code for debugging

    let a = BigFloat::from(-1).frac();  
    let b = BigFloat::from(0);  
    assert!(a==b);

In the following situation:

        println!("input: {}", input);
        let val = match BigFloat::parse(input) {
            Some(s) => s,
            None => return None,
        };
        println!("val: {}", val.to_string());

The following are being treated as though they are identical:

input: 100000000000000000000000000000000000000000000.00000000000000000000000000000000000000001
val: 1.000000000000000000000000000000000000000e+39
input: 100000000000000000000000000000000000000000000.00000000000000000000000000000000000000002
val: 1.000000000000000000000000000000000000000e+39

When I compare them using .cmp() it returns 0.

The epsilon's mantissa is backwards making the epsilon value useless.

Minimal Reproducer

use num_bigfloat::BigFloat;

fn main() {
    let one = BigFloat::parse("01").unwrap();
    println!("{one}");
}

Expected Output

1.000000000000000000000000000000000000000

Actual Output

1.000000000000000000000000000000000000000e+39

Commentary

This behavior holds for any input that meets the following criteria:

1. The fractional part (i.e., digits following the decimal point) is equal to zero.
2. The number of leading zeros is equal to the number of remaining digits in the integral part.

Here are some more examples with shortened outputs:

While using the num-bigloat crate, we were hitting a panic mentioning that thread 'main' panicked at 'attempt to add with overflow'

After some investigation, we noticed that any operation that results in a value where the result is -1 > x < 1, panics.

For example, the code snippet below results in a panic.

use num_bigfloat::BigFloat;

fn main() {
    let x = BigFloat::from(1);
    let y = BigFloat::from(2);
    let z = x / y;
    println!("{z:?}");
}