nunit-dev team mailing list archive

[Charlie Poole <charlie@xxxxxxxxx>]

Hi Simone,

>> On Fri, Oct 21, 2011 at 2:53 PM, Simone Busoli <simone.busoli@xxxxxxxxx>
>> wrote:
>> > Hi, unless I'm missing something - Charlie said on the "bug" page that
>> > it's
>> > not trivial - all I would do is insert an additional routine call before
>> > the
>> > default fallback to x.Equals(y) in NUnitEqualityComparer which would:
>>
>> Two issues about this:
>>
>> 1. Inserting the change right before the fallback x.Equals(y) sounds
>> good to me at least
>> for now. It will mess up folks with a class that implements
>> ICollection, but I think we
>> can live with that.
>
> Yes, I thought about it, I think there's no perfect solution here. Putting
> the new code before the object.Equals fallback we choose to give precedence
> to the implementation of IEnumerable, ICollection, .. over IEquatable. That
> said, I think it's unusual to implement both IEnumerable and IEquatable.

I'm OK with this solution.

>>
>> 2. The code itself is going to be reflection-based and reflection on
>> generic types is
>> a bit hairy, but I think it's doable.
>>
>> > check if both x and y implement IEquatable<>
>> > check that they both implement IEquatable of the same T
>> > if the two above apply, call the IEquatable<T>.Equals method somehow,
>> > using
>> > reflection perhaps
>>
>> I think you can check the types first and ensure that they are the
>> same type. I'm
>> not sure there's a case for a type T1 that implements IEquatable<T2>, but
>> I
>> could be wrong.
>
> I actually did it differently, there's no point in them being both
> IEquatable, it's sufficient for x to be IEquatable of y, and it's what I did
> in the code.

Should we allow for the opposite? That is, y is IEquatable<x>?

>> > In any case, I'm not sure we're still supporting .NET 1.x, in that case
>> > this
>> > would go under a pre-processing direct
>>
>> We are in 2.6, so put it all under #if CLR_2_0 || CLR_4_0.
>
> I committed the changes by branching the trunk, not sure it is correct. They
> are in my branch so you can take a look.

Yes, I've seen it and replied on the bug itself. Waiting to hear if you want to
have it integrated or will be working on it further.

Charlie
>>
>> Charlie
>> _____________
>> > Mailing list: https://launchpad.net/~nunit-dev
>> > Post to     : nunit-dev@xxxxxxxxxxxxxxxxxxx
>> > Unsubscribe : https://launchpad.net/~nunit-dev
>> > More help   : https://help.launchpad.net/ListHelp
>> >
>> >
>
>