tc39/proposal-object-getownpropertydescriptors

Name: proposal-object-getownpropertydescriptors

Owner: Ecma TC39

Description: ECMAScript proposal for Object.getOwnPropertyDescriptors

Created: 2016-01-20 17:37:57.0

Updated: 2018-04-05 07:03:37.0

Pushed: 2016-07-30 17:18:43.0

Homepage: https://tc39.github.io/proposal-object-getownpropertydescriptors

Size: 4183

Language: CSS

GitHub Committers

User	Most Recent Commit	# Commits

Other Committers

User	Email	Most Recent Commit	# Commits

README

Object.getOwnPropertyDescriptors Proposal (Polyfill)

Champion

At stage 0 Rick Waldron agreed to champion this proposal. However the current official Champion is Jordan Harband.

Status

This proposal is currently in stage 4 of the TC39 process.

This proposal could be identical to a Reflect.getOwnPropertyDescriptors one but for consistency with other plural versions it's described as an Object public static method.

Motivation

There is not a single method in ECMAScript capable of simplifying a proper copy between two objects. In these days more than ever, where functional programming and immutable objects are essential parts of complex applications, every framework or library is implementing its own boilerplate in order to properly copy properties between composed objects or prototypes.

There is a lot of confusion and most of the time undesired behavior when it comes to fallback to Object.assign because it copies in a way that swallows behavior: it directly accesses properties and symbols instead of their descriptors, discarding possible accessors which could result into an hazard when it come to composing more complex objects or classes? prototypes.

Retrieving all descriptors, enumerable or not, is also key to implementing composition over classes and their prototypes, since by default they have non-enumerable methods and accessors.

Also decorators could easily grab at once all descriptors from another class or mixin and assign them through Object.defineProperties. Filtering undesired descriptors would be simpler too, as well as less repetitive each time is needed.

Last, but not least, a shallow copy between two unknown objects would be free of surprises compared to what Object.assign would do.

FAQs

Should there be a Reflect.getOwnPropertyDescriptors ?

Since the main goal of this proposal is to simplify some common boilerplate and be consistent with the fact there is a singular version of the method but not a plural one, it might be further consistent to have the plural version of the current Reflect.getOwnPropertyDescriptor method too.

Update: The committee has previously decided that Reflect is solely to mirror Proxy traps, so this is not an option.

Proposed Solution

As plural version of Object.getOwnPropertyDescriptor, this proposal is about retrieving in one single operation all possible own descriptors of a generic object.

A polyfill of such proposal would look like the following:

!Object.hasOwnProperty('getOwnPropertyDescriptors')) {
ject.defineProperty(
Object,
'getOwnPropertyDescriptors',
{
  configurable: true,
  writable: true,
  value: function getOwnPropertyDescriptors(object) {
    return Reflect.ownKeys(object).reduce((descriptors, key) => {
      return Object.defineProperty(
        descriptors,
        key,
        {
          configurable: true,
          enumerable: true,
          writable: true,
          value: Object.getOwnPropertyDescriptor(object, key)
        }
      );
    }, {});
  }
}

Illustrative Examples

The polyfill shows an alternative, ES2015 friendly, way that improves the boilerplate needed for engines compatible with ES5 or partially with ES2015.

Now that Object.getOwnPropertyDescriptors is in place, all it's needed in order to make a real shallow copy or clone operation between two objects, is shown in the following example:

t shallowClone = (object) => Object.create(
ject.getPrototypeOf(object),
ject.getOwnPropertyDescriptors(object)


t shallowMerge = (target, source) => Object.defineProperties(
rget,
ject.getOwnPropertyDescriptors(source)

Possible objects based mixin solutions could also benefit from this proposal:

mix = (object) => ({
th: (...mixins) => mixins.reduce(
(c, mixin) => Object.create(
  c, Object.getOwnPropertyDescriptors(mixin)
), object)


ultiple mixins example
a = {a: 'a'};
b = {b: 'b'};
c = {c: 'c'};
d = mix(c).with(a, b);

Let's say you wanted a version of Object.assign that uses [[DefineOwnProperty]]/[[GetOwnProperty]] instead of [[Set]]/[[Get]], to avoid side effects and copy setters/getters, but still use enumerability as the distinguishing factor.

Before this proposal, such a method would look like:

tion completeAssign(target, ...sources) {
urces.forEach(source => {
// grab keys descriptors
let descriptors = Object.keys(source).reduce((descriptors, key) => {
  descriptors[key] = Object.getOwnPropertyDescriptor(source, key);
  return descriptors;
}, {});
// by default, Object.assign copies enumerable Symbols too
// so grab and filter Symbols as well
Object.getOwnPropertySymbols(source).forEach(sym => {
  let descriptor = Object.getOwnPropertyDescriptor(source, sym);
  if (descriptor.enumerable) {
    descriptors[sym] = descriptor;
  }
});
Object.defineProperties(target, descriptors);
;
turn target;

However, if Object.getOwnPropertyDescriptors was available, above boilerplate would look like:

completeAssign = (target, ...sources) =>
urces.reduce((target, source) => {
let descriptors = Object.getOwnPropertyDescriptors(source);
Reflect.ownKeys(descriptors).forEach(key => {
  if (!descriptors[key].enumerable) {
    delete descriptors[key];
  }
});
return Object.defineProperties(target, descriptors);
 target);