The Flattr button and author attribution text in the banner are partially hidden in Opera because it is mispositioned.

To improve readability of the list, I propose to:

• Drop/hide obsolete browsers, e. g. Firefox <17 (Fx 17 is ESR, so it will be around for at least a few months), Chrome <21 (usage numbers are near-zero or actually zero).
• Make columns hideable (so one can only display the columns he is interested in). This can be achieved with a small JS (which I'm happy to provide, but it may take some time before I can get to it).
• Highlight current row (and, if possible, column) on mouseover.
• A way to unfocus browser, by clicking on the focused browser again
• Separate browser+versions into 2 headers
• Fixed header
• Replace bottom notes with inline tooltips (for convenience)
• Expand subtests by clicking on entire cell (except link) not just "down" arrow

I'm not sure, which node version is been targeted in the table, but the below mentioned works fine for v0.11.7

$ node -v
v0.11.7

# Note : --harmony enables all the flags except type_of in V8, but for node we need to do the below.
$ node --use-strict $(node --v8-options | grep harm | awk '{print $1}'| xargs) # enabling all harmony flags

# let keyword works.
> eval('(function () { let foobarbaz2 = 123; return foobarbaz2 == 123; }())')
true

# Iterators and for-of loop for them works fine.
function *Counter(){
  var n = 0;
  while(1<2) {
    yield n;
    ++n;
  }
}

var CountIter = new Counter();

for(let value of CountIter) {
  console.log(value);
} 

# Generator yield works fine.
> eval('(function* () { yield 5; }())');
{}

# Sets works fine.
> var set = new Set();
> set.add(123); 
> set.has(123) && set.size === 1
> true

# Proxies works fine, Proxy.create() is present, not the latest spec though.
> Proxy
{}

# Block level function definition works 
>> eval('{function f(){}} typeof f == "undefined"');
true

# Some of the String functions 
> String.prototype.repeat
[Function: repeat]
> String.prototype.startsWith
[Function: startsWith]
> String.prototype.endsWith
[Function: endsWith]

# Array functions 
> Array.prototype.find
[Function: find]
> Array.prototype.findIndex
[Function: findIndex]

# Finally Math
 > Math.imul
[Function: imul]

data-es6.js#L2567 need a var as it's in strict mode?

L653 needs a fix for node?

Happy to send a PR :)

The outline around the "Current Browser" column is being displayed over the info window that is displayed when you hover over (c)

Because of the two separate script tags in the row marked let, the css selector td:nth-child(3) { outline: #AAF solid 3px; } is not being applied to the "this browser" column for that row.

What is the easiest & fastest way to add a column (eg new browser version) for your tables please? I was about to add FF14 for you, & I looked at it, & said, "Table layouts = cool, not cool I have to hand- edit a new column." What tools do you use/recommend, or can this be automated please? New browser versions are added every few months now.

Hi! A test for destructuring (see http://wiki.ecmascript.org/doku.php?id=harmony:destructuring) is missing. Maybe something like this could work:

test((function () {
  try {
    return eval('(function() { var { foo } = { foo: 5 }; ; return typeof foo === "number" })()');
  } catch(error) {
    return false;
  }
})())

I think only V8 with --harmony supports WeakSet but it would be good to have it in the table for completeness sake.

Check this out: http://mxr.mozilla.org/mozilla-central/source/js/src/tests/ecma_5/extensions/extension-methods-reject-null-undefined-this.js#18

E.g. String#quote, toJSON on most objects, toSource on most objects, etc.

Right now it uses Firefox's nonstandard syntax. function* () { } is the correct syntax.

E.g. a\u200c\u200d is a valid identifier as per ES5, but not all browsers/engines support it (yet).

This can easily be tested as follows:

var supportsZeroWidth = (function() {
    try {
        Function('var a\u200c\u200d');
        return true;
    } catch(e) { }
}());

supportsZeroWidth will be true if ZWJ and ZWNJ characters are supported in IdentifierPart, else undefined (which is falsy).

I'm not sure I understand the ES6 "pragmas" test, as there is nothing in ES6 that matches:

use strict;

(not to be confused with the string version of the same words)

ES5 defines getter and setter initializers: http://es5.github.com/#x11.1.5
Information on support can be found here: http://robertnyman.com/javascript/javascript-getters-setters.html#regular-getters-and-setters
Likewise on MDN:

• https://developer.mozilla.org/en/JavaScript/Reference/Operators/Special/set
• https://developer.mozilla.org/en/JavaScript/Reference/Operators/Special/get

… this is for iterators: http://people.mozilla.org/~jorendorff/es6-draft.html#sec-properties-of-the-array-prototype-object

{keys, entries} is available in Firefox 28, values since Fx 25.

Seems, Chrome supports it:

var object = {};
var symbol = Symbol();
object[symbol] = 42;
Object.getOwnPropertyNames(object).length === 0 && object[symbol] === 42;

P.S.
although,
http://jsperf.com/symbol-vs-weakmap/3
may be i write a bad test or Chrome's support is not very good.

master.js looks really crazy to me.
Why not using jQuery or Zepto for manipulating the DOM?

OSX Mavericks features Safari 7, which should be included in the ES5/6 tables

I was just browsing your awesome site here:
http://kangax.github.io/es5-compat-table/#Array.prototype.indexOf

and its not possible to easily tell which column goes with which browser once you scroll past the headers. I wrote a plugin that solves that problem, here is an example that looks similar to your site:

http://mkoryak.github.io/floatThead/examples/hidden-columns/

I can open a pull request if you are open to using that plugin.

The ES6 table says that Iterators aren't supported in any version of Firefox, however my own testing in Firefox 23 indicates that they are supported.

https://bugzilla.mozilla.org/show_bug.cgi?id=885798

The “CH 7-10” column says callable regex is supported, but I’m getting a “NO” in the “this browser” column when using Chrome Canary v22.0.1187.0. I’m not sure if this is a typo, a false positive in the test, or a regression in Chrome. I haven’t tested other Chrome versions.

Typed objects and StructType aren't in the latest ES6 spec, and none of the tracked platforms implement it.

I know this is probably an odd request, but I am red/green color blind and I am not able to see the difference in the table, beyond the yes/no. Is there a possibility for colorblind (red/green) friendly colors?

Here's the spread(...) test:

test((function () {
          try {
            return eval('(function(args) { return typeof args !== "undefined"; })()');
          } catch(error) {
            return false;
          }
        })())

Maybe you meant eval('(function(args) { return typeof args !== "undefined"; })(...[1])'); ?

Same as Object.defineProperty

https://twitter.com/awbjs/status/422859931075629056

• u flag
• @@hasInstance and @@toPrimitive
• Array is subclassable
• RegExp is subclassable

First off, super useful data source!

I would like to use this data in an automated fashion in an node.js module, but concerned about licensing.

I wasn't sure if you intended to put a license on it, but it would be awesome if you released it as open source under something like zlib or MIT.

Cheers!

Uncaught ReferenceError: domready is not defined

title, cheers :)

Although it's not finalized how users can create iterable objects, it's definitely not going to be through a simple iterator property. Right now you're just testing against Firefox's nonstandard implementation.

I can't really see a good test for this, however, given that TC39 hasn't decided exactly how it will work. Probably via a symbol that you import from a standard module, but not clear.

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Math/imul

Suggested test:

try {
  eval('var a = () => 5;');
} catch (e) {
  return false;
}
return true;

As gh-pages is the default branch, and master has not been updated since … ever, I think you can remove master. It's a bit confusing to have them both around, with one being totally out-of-date and unneeded.

Perhaps add a count next to each feature of how many of the listed browsers have implemented the feature, and then have a checkbox to sort by most-implemented features?

I would find it very helpful to see Android browsers at a glance in this table as well. If anyone knows of a existing resource, I’ll happily contribute a PR!

The module test looks wrong, it's the same as the rest parameters test.

Traceur supports a lot of the features the table states it doesn't. Especially if you enable the experimental features. Things like destruction for instance is implemented already, and experimental support for let and const is in place (though it might kill your performance).

A recently-explored hack to pass multiline strings in javascript is to use comments (e.g. http://tomasz.janczuk.org/2013/05/multi-line-strings-in-javascript-and.html) This technique only works if Function.prototype.toString preserves the original content (and not some stripped version). This technique does not work in firefox 16

The checkbox option is inverted. When checked, the columns are hidden, and when unchecked (after having been checked initially), they are show.

https://github.com/kangax/es5-compat-table/blob/50dfabff13b963e7f681/index.html#L1132

Line 1132 of index.html should be changed to

document.body.className = this.checked ? 'hide-old-browsers' : '';

Here are some additional semantics you can test for let and const. All of them should only be run if the basic tests succeed (i.e. const x = 5; and let x = 5; do not produce errors when evaled).

const must have an initializer

try {
  eval('const x;');
} catch (e) {
  return e instanceof SyntaxError;
}
return false;

Cannot occur twice with the same identifier within a block

try {
  eval('let x; let x;');
} catch (e) {
  return e instanceof SyntaxError;
}
return false;

try {
  eval('const x = 5; const x = 5;');
} catch (e) {
  return e instanceof SyntaxError;
}
return false;

Temporal dead zone (use before declaration)

try {
  eval('x; const x = 5;');
} catch (e) {
  return e instanceof ReferenceError;
}
return false;

try {
  eval('x; let x;');
} catch (e) {
  return e instanceof ReferenceError;
}
return false;

Temporal dead zone still works with shadowing

try {
  eval('const x = 2; { x; const x = 5; }');
} catch (e) {
  return e instanceof ReferenceError;
}
return false;

try {
  eval('let x; { x; let x; }');
} catch (e) {
  return e instanceof ReferenceError;
}
return false;

Fresh per-loop iteration

var a = [];
for (let i = 0; i < 3; ++i) {
  a[i] = function () { return i; };
}

return typeof i === "undefined" && a[1]() === 1;

In particular, testing for Proxy.create and Proxy.createFunction will product false negatives in environments which only implement direct proxies, and not the nonstandard spec.

Suggested test:

typeof Proxy === "function" && new Proxy({}, { get: function () { return 5; }).foo === 5;

apparently Firefox fails the first part of this.

https://github.com/kangax/es5-compat-table/blob/50dfabff13b963e7f681/index.html#L34