Teach LambdaEqualityHelper to deal with null equality contribution

[ethomson]

No description provided.

[dahlbyk]

How about https://github.com/dahlbyk/libgit2sharp/compare/equalityhelper?

[nulltoken]

@ethomson I'd indeed prefer not to make LambdaEqualityHelper public. I kind of like @dahlbyk's counter proposal.

FWIW, I found a clear case in LibGit2Sharp which demonstrates this issue.

[Fact]
public void CanRetrieveHashCodeFromAnUnbornBranch()
{
    string path = InitNewRepository();
    using (var repo = new Repository(path))
    {
        Assert.DoesNotThrow(() => repo.Head.GetHashCode());
    }
}

[nulltoken]

Rather than using 0 (which would level down the previous computation steps to 1), how about using a bit larger number (eg. The Java 31 or whatever...)

[ethomson]

I don't think I'm following entirely; should XORing with zero versus one change the distribution significantly?

Note that Joshua Bloch's hashCode is the reference implantation that I think you're referring to, and uses 0 in this case for null elements.

[nulltoken]

Duh. Ignore me.

I don't know why, but I forgot that ^ doesn't refer to a "power-of" function in C#. Sorry for the noise. 😊

[ethomson]

I brought in @dahlbyk 's tests. This is one of the few times I wish for java's package protection...!

[dahlbyk]

This is one of the few times I wish for java's package protection...!

We could achieve a similar effect with InternalsVisibleTo, though that feels like cheating.

[ethomson]

Squashed, rebased, etc.

[nulltoken]

Very nice catch!

❤️ to both of you.