split method is ambiguous on its own

[mx00s]

Calling split on a single-element NonEmpty returns the same thing as a two-element NonEmpty when all the elements are the same.

    #[test]
    fn non_empty_split() {
        let one_element: NonEmpty<i32> = NonEmpty {
            head: 0,
            tail: vec![],
        };
        assert_eq!(one_element.len(), 1);

        let two_elements: NonEmpty<i32> = NonEmpty {
            head: 0,
            tail: vec![0],
        };
        assert_eq!(two_elements.len(), 2);

        // This assertion fails because they return the same thing, namely `(0, &[], 0)`!
        assert_ne!(one_element.split(), two_elements.split());
    }

Although this situation can be disambiguated, e.g. by checking the length, it's a potential source of confusion and bugs.

Changing the return type to (&T, &[T], Option<&T>) and returning (0, &[], None) for the one-element case and (0, &[], Some(0)) for the two-element case would resolve the ambiguity.

[FintanH]

That makes sense 👍 I'll add a patch to change it to return Option for the last element