IE Non-Enumerable properties

[jridgewell]

I’m not sure what we were doing before with the FuncProto…

@jdalton: mind giving this a review?

[paulfalgout]

This seems to work in IE8. Beat me to it. If you're waiting for @jdalton to review.. it might be awhile.. he's been banned.

[jridgewell]

I've got it working in IE6 and 8, so I think we're set.

If you're waiting for @jdalton to review.. it might be awhile.. he's been banned.

Well, it's been one of those days.

[jashkenas]

What's the difference from the previous implementation?

[jridgewell]

The only important change is determining the prototype to compare to.

@@ -894,17 +896,21 @@

   // Keys in IE < 9 that won't be iterated by `for key in ...` and thus missed.
   var hasEnumBug = !{toString: null}.propertyIsEnumerable('toString');
-  var nonEnumerableProps = ['constructor', 'valueOf', 'isPrototypeOf', 'toString',
+  var nonEnumerableProps = ['valueOf', 'isPrototypeOf', 'toString',
                       'propertyIsEnumerable', 'hasOwnProperty', 'toLocaleString'];

   function collectNonEnumProps(obj, keys) {
     var nonEnumIdx = nonEnumerableProps.length;
-    var proto = typeof obj.constructor === 'function' ? FuncProto : ObjProto;
+    var constructor = obj.constructor;
+    var proto = (_.isFunction(constructor) && constructor.prototype) || ObjProto;
+
+    // Constructor is a special case.
+    var prop = 'constructor';
+    if (_.has(obj, prop) && !_.contains(keys, prop)) keys.push(prop);

     while (nonEnumIdx--) {
-      var prop = nonEnumerableProps[nonEnumIdx];
-      if (prop === 'constructor' ? _.has(obj, prop) : prop in obj &&
-        obj[prop] !== proto[prop] && !_.contains(keys, prop)) {
+      prop = nonEnumerableProps[nonEnumIdx];
+      if (prop in obj && obj[prop] !== proto[prop] && !_.contains(keys, prop)) {
         keys.push(prop);
       }
     }

[paulfalgout]

The only functional change I see really is using obj.constructor.prototype vs FuncProto

[jashkenas]

Kill the "fasley"s?

[megawac]

LGTM 👍

[jridgewell]

The only functional change I see really is using obj.constructor.prototype vs FuncProto

That's what I meant.

Kill the "fasley"s?

But they pass? And it's good to test edge cases.

---

Travis is erroring because of issues with 173a7fa. #2065

[paulfalgout]

That's what I meant.

You're too quick for me

The tests should say "falsey" I think right?

[JakeChampion]

"fasley" should be "falsey"

[jridgewell]

Already fixed.