Visionmedia JSpec → 3.0

Consider this (simplified) code:

  function toArray(iterable) {
    var length = iterable.length, results = new Array(length);
    while (length--) results[length] = iterable[length];
    return results;
  }

I.e., note how the returned array is exactly the same, but it is build by copying in reverse order.

and this test:

  it "should copy the array"
    var args = function() { return toArray(arguments) }(9,8,7);
    args.should.eql [9,8,7]
  end

This work in Firefox, but fails in IE7 where it responds with "expected [7,8,9] to eql [9,8,7]"

It appears that the hash value that JSpec calculates for arrays is dependent on how the array is built. This is because the hash value algorithm iterates over a hash, even if the value is an array.

I expect the same issue can occur with testing for object equality. For example, {x:1, y:2}.should.eql {y:2, x:1} would probably fail, where I think it should succeed.

[I just tested this, it does indeed fail, also in firefox]

I think the hash algorithm should not use a "for (key in object)" loop (in the +each+ function) if the object is an array: it should really iterate over the indexes instead to preserve the correct order. If it is an object I would suggest it should sort on the keys first before calling the callback functions.