[BUGFIX beta] Fix a recent regression with Ember.A

[workmanw]

Due to a recent regression, Ember.A(null) would return null instead of []. See issue #13284.

[mixonic]

In the native version, you rely on truthiness as a check for arr, but in this version you look explicitly for null instead of !arr. Is there a reason for the difference?

[workmanw]

@mixonic Yea that's a good point. I think I was just trying to be more explicit about what was happening. Someone reading the following might feel that it's redundant:

A = function (arr = []) {
  if (!arr) { arr = []; }
  return EmberArray.detect(arr) ? arr : NativeArray.apply(arr);
};

But it's definitely inconsistent between the two. So let's fix this, which do you think is prefered?

[mixonic]

Dunno- I would defer to the old behavior since this is a regression fix. Can you perhaps try it out on 2.4 or and older version and simply match that? Thanks man.

[workmanw]

@mixonic Ha. It turns out the old behavior was inconsistent too. 586d4c6#diff-7. I think I'll just use the truthiness check and add a code comment. Then see what others think.

[mixonic]

😭 truthiness was the mode with prototype extensions enabled (which is the most common), so I agree normalizing around that as a BUGFIX makes sense 👍

[krisselden]

why bother then with the default arg?

[krisselden]

the default arg is redundant to handling an explicitly undefined or null, please remove.

[krisselden]

also, the comment is unneeded with the regression test.

[krisselden]

This regression is unreleased?

[workmanw]

Chatted with @krisselden briefly on Slack. We agreed that keeping the default args was unnecessary in this case. As a result of that I removed the code comment as well.

This affects beta, it does not affect any release.

[mixonic]

Thanks @workmanw

[workmanw]

Anytime! 😄