885531 - For Map and Set, use SameValueZero as comparator

Status: RESOLVED DUPLICATE of bug 952870

(Core :: JavaScript Engine, defect)

Description

[Brandon Benvie [:benvie]]

A recent change to the ES6 spec changes the signatures of the Map and Set constructors to look like `Map(iterable = undefined, comparator = undefined)`. The comparator, if not undefined, must be the string "is" or a TypeError is thrown. This acts as a toggle for the algorithm used to determine equality, with the default comparison using `SameValueZero`, and "is" switching it to `SameValue`. `SameValueZero` only differs from `SameValue` in that it considers -0 and +0 as equal (it's like `===` but `SameValueZero(NaN, NaN)` is true).

This observably changes the default behavior of Sets and Maps, because prior to this change `SameValue` was the default (and only) comparator.

As currently implemented in SpiderMonkey:
> set.add(-0)
> set.has(0) // false

After implementing this change:
> set.add(-0)
> set.has(0) // true

To achieve the current behavior after this change:
> var set = new Set(undefined, "is")
> set.add(-0)
> set.has(0) // false


See the following ES6 spec sections (May 2013 draft):
* SameValueZero: 9.2.4
* The Map constructor: 15.14.1.1
* The Set constructor: 15.16.1.1
* Map.prototype.[get|set|has|delete] check `[[MapComparator]]`
* Set.prototype.[add|has|delete] check `[[SetComparator]]`

See also: https://github.com/rwldrn/tc39-notes/blob/master/es6/2013-01/jan-31.md#mapset-comparator


Note: this doesn't affect WeakMaps since they can't have 0/-0 as a key.

Comment 1

[Sankha Narayan Guria [:sankha]]

I am trying to write a patch for this one.

Comment 2

[4esn0k]

seems, in IE11 and in Chrome
map.get(+0) is not the same as map.get(-0),
may be it is better to change ES6 spec?

Comment 3

[Brandon Benvie [:benvie]]

The most recent update to the spec removed the constuctor comparator argument, but it still uses SameValueZero as the comparator. I'll inquire on es-discuss about this.

Comment 4

[Rick Waldron [:rwaldron]]

Relevant thread https://mail.mozilla.org/pipermail/es-discuss/2014-January/035847.html

Comment 6

[Florian Scholz (Open Web Docs)]

For docs see bug 952870 comment11.