[[Enumerate]] in the presence of prototype changes and property deletions.

Gareth Smith gds at doc.ic.ac.uk
Tue Apr 16 09:29:05 PDT 2013


This is my first post here - please let me know if I do something wrong!

I've filed an issue here: https://bugs.ecmascript.org/show_bug.cgi?id=1444

Section 8.3.12 of the current draft[1] defines the behaviour of
[[Enumerate]], and describes what should happen if properties are added
or removed during enumeration:

"If a property that has not yet been visited during enumeration is
deleted, then it will not be visited. If new properties are added to the
object being enumerated during enumeration, the newly added properties
are not guaranteed to be visited in the active enumeration."

It is not so clear on what happens if we enter the loop with an
enumerable property shadowing a non-enumerable one, and then delete the
enumerable one before it is visited.

Perhaps the "nicest" reading of the existing text would be to treat the
revealed properties the same way we would treat "added" properties. So,
in this case, the "revealed" property would be checked for enumerability
and skipped. This doesn't seem to be the only possible reading of the
spec however.  For example, while spidermonkey seems to behave as I had
expected, V8 disagrees, and visits the non-enumerable property.

Essentially the same trick can be played by mutating __proto__ during
enumeration.

Here is a code snippit which demonstrates this behaviour, assuming that
x happens to come before z in the enumeration order:

  var a = {};
  Object.defineProperty(a, 'z', {value: 'secret', enumerable: false});
  var b = {x:0, z:"tricky", __proto__:a};
  var ret = "We have not visited a.z";
  var deleted_z_yet = false;
  for (var i in b) {
      if(i=='x') {
          delete b.z ;
          deleted_z_yet = true;
      }
      if (i=='z' && deleted_z_yet) {
          ret = "we have visited a.z" ;
      }
  }

A related issue is that the "informative algorithm" in this section
seems to return a static list, which can't react to deleted
properties. I've filed that issue here:
https://bugs.ecmascript.org/show_bug.cgi?id=1443

Finally, FWIW, when I filed these issues, I got a long wait (5 mins or
so), followed by "internal server error" each time. However, when I
tried to re-submit, I got a faster response informing me that my
previous submission had succeeded. I don't know who to tell about this,
and the "Help" link at the top of the bugzilla page 404s, so I thought
I'd mention it here.

Thanks,

Gareth.

[1] http://people.mozilla.org/~jorendorff/es6-draft.html#sec-8.3.12 if
you're reading the HTML version.


More information about the es-discuss mailing list