PartialEq implementations for Option and Result are painfully restrictive

[spacejam]

current implementation:

impl<T> PartialEq<Option<T>> for Option<T> where
    T: PartialEq<T>

impl<T, E> PartialEq<Result<T, E>> for Result<T, E> where
    E: PartialEq<E>,
    T: PartialEq<T>, 

desired implementation:

impl<A, B> PartialEq<Option<B>> for Option<A> where 
    A: PartialEq<B>

impl<TA, TB, EA, EB> PartialEq<Result<TB, EB>> for Result<TA, EA> where
    EA: PartialEq<EB>,
    TA: PartialEq<TB>, 

This would allow nested types that can be compared to be compared without mapping over the structure. I have a particularly painful usecase in sled, where I return a special zero-copy read pointer for some operations that allows me to avoid clones of requested buffers without using an Arc. This pointer type implements PartialEq<T> where T: AsRef<[u8]> so it can be compared against vectors, slices, etc... But the database returns responses to requests as something like Result<Option<ZeroCopyPointer>> which really makes it cumbersome to do things like assert!(Ok(Some(zcp)) == Ok(Some(vec![1,2,3,4]))).

I see a little discussion of something similar at #20927, but it's not totally clear to me based on that if this change would be out of the question due to possible backwards-compatibility concerns.

Is there a major reason why this is impossible with today's rust? I'm happy to implement it if not.

[varkor]

This is covered by #20063. See #20063 (comment) in particular.

[spacejam]

@varkor thanks for the info! Hoping this will be feasible one day.