Nuking misleading properties in `Object.getOwnPropertyDescriptor`

Nathan Wall nathan.wall at live.com
Sat Mar 9 19:54:37 PST 2013


Given that `defineProperty` uses properties on the prototype of the descriptor[1] and `getOwnPropertyDescriptor` returns an object which inherits from `Object.prototype`, the following use-case is volatile:

    function copy(from, to) {
        for (let name of Object.getOwnPropertyNames(from))
            Object.defineProperty(to, name,
                Object.getOwnPropertyDescriptor(from, name));
    }

If a third party script happens to add `get`, `set`, or `value` to `Object.prototype` the `copy` function breaks.

    Object.prototype.get = function() { };

    var x = { foo: 1 };
    var y = { };
    copy(x, y); // TypeError: property descriptors must not specify a value or
                // be writable when a getter or setter has been specified

It's misleading that you can't trust `getOwnPropertyDescriptor` to give you a descriptor which you can use to define a property with `defineProperty`.

A script which doesn't use `defineProperty` could very well define a `get` property on `Object.prototype`, unknowingly breaking other scripts which do use `defineProperty`. This is a hazard in the language.

I think one of the following would be an appropriate solution:

(1) "Nuke" the special properties (`get`, `set`, and `value` when any of them is not defined) on a descriptor returned by `getOwnPropertyDescriptor` which shouldn't be inherited through the descriptor's prototype. By "nuke" them, I mean specify that they be defined as `undefined`, much like `callee` and `caller` in strict mode. (I don't think strict mode is required in this case, though, since I highly doubt anyone could be relying on this behavior.)

(2) Use the introduction of `Reflect.defineProperty` as an opportunity to "fix" this function to only inspect own properties on the property descriptor so that the descriptor's prototype doesn't matter. This is similar to the addition of `Number.isNaN` to fix the broken `isNaN`.

Nathan


[1] https://mail.mozilla.org/pipermail/es-discuss/2012-November/026705.html 		 	   		  


More information about the es-discuss mailing list