[mfonda]

I think there was a more general question here that was missed: can a == x && a == y ever be true for any arbitrary values of a, x, and y, where x != y.

From a logical point of view, no, this can never be true. I would suspect this can never be true in javascript, and could only be made true in a language where you can override == to always return true.

I think when most developers use the word "never", what they really mean is "never (within the current context)". This makes conversations a lot simpler. Imagine how difficult conversations would be if you always had to qualify never. "This can never be true (assuming a weird valueOf method hasn't been defined and assuming I didn't modify the javascript interpreter to always return true for == and assuming ...)".

[lmkg]

I think there are a few cases of non-transitive equalities in JavaScript, specifically around falsey values. Part of it also has to do with whether x is the right-side input or the left-side input, because that changes type coercion rules.

Aha, found one:

  ['0'] == 0
  > true
  [0] == 0
 > true
  [0] == ['0']
 > false

Transitivity of equality can also never be relied upon when dealing with Floating Point numbers. This is correct and unavoidable behavior, but still surprising to many developers.

[gsg]

That's not due to transitivity: ['0'] == ['0'] (and for that matter, ['0'] === ['0']) is also false. These operators compare objects for identity, not structural equality.

[russellsprouts]

Something like this could work, to some degree:

        a = {}
        a.valueOf = function() {
           var caller = arguments.callee.caller.toString();
           //parse the caller function, and return 
           //different values based on what comparison is being made. 
        }

But I think this is even worse than the getter on the global object.

[tonyarkles]

Here's a JavaScript example, based on what was in the article.

    var b = {c:0};
    b.valueOf = function() { this.c++; return this.c; }
    b == 1 && b == 2
    > true

[mfonda]

That only works for consecutive integers, not any arbitrary values.