v2

[shawnbot]

v2 replaces several of its old incomplete or non-compliant shims with @WebReflection's excellent ie8 and dom4 libraries (both referenced now as submodules, as is es5-shim), and introduces an actual test suite that runs in both IE8 and modern browsers. Some things to do before merging/releasing:

• flesh out unit tests (particularly, more DOM+CSS) More unit tests! #32
• update the README with links to new libs, etc.
• update the LICENSE to reference included libs
• consider Additional work on CSSStyleDeclaration #17: replace brain-dead CSS OM shim with a more complete implementation
• consider jQuery compatibility unit tests, e.g. for jQuery UI breaks aight if loaded after #24

And, because this all started as a way to get IE8 working with d3, let's also consider doing the following:

• add d3-specific unit tests (e.g. for the d3.selection#style("opacity") shim, see d3.syle("opacity") shim fails if there's no opacity set #13)
• d3.selection.sort() doesn't work in IE8 #1: fix d3.selection#sort() in aight.d3.js
• review d3.v2 loads but not d3.v3 #25

[usmonster]

I think the Node.prototype.textContent shim I added in #27 was more complete and necessary for reasons outlined in the issue. This one creates a regression with non-Element nodes and re-breaks, for example, certain common uses of jQuery.

[WebReflection]

I'd be happy to update ie8 with your Node.prototype.textContent proposal, can you point me at the shim you are mentioning? Thank you

[usmonster]

It's basically in the pull request I referenced, or you can just look at the element-properties.js file in the "js" directory of the current version. I'm also planning on making a pull request to add it to polyfill, which would also include tests. (Please also note that I haven't tested on IE 6 or 7, as those were not target browsers, and I'm not sure it will apply to them.)

[shawnbot]

Thanks guys. @WebReflection, let me know when this is done in ie8 and I'll update here.

[WebReflection]

it should be done in 0.2.1 … br returns \n and everything else should be patched as it was here (well, more or less, it's optimized/retested code)

Please let me know if there's any problem with such version. Regards

[WebReflection]

if standards have those as enumerable I will … setters returning values are misleading for the reason you might return something different and expect that value but the operation you thought was the result of a setter will break such expectations ;-)

var o = {};
Object.defineProperty(o, 'test', {set: function(){return 123}});
var value = (o.test = 456);

// you see what I mean ?
value // 456

If it's style, drop it, 'cause it's a misleading one. Cheers

[usmonster]

Quite right! I meant it's a style I use in certain other contexts where I might actually want to return something useful from a set operation. (Though, really, here I just blindly copied what the original author did.)

On the subject of copying what others do, some quick testing shows that textContent is both enumerable and configurable (or at least the property can be deleted) in modern browsers, though I haven't found anything about it in specs.

[WebReflection]

I made them configurable but if I try to make them enumerable too everything fails. I think configurable only might be OK t we don't have many options here … I also wrote a comment explaining this so enumerable: false it is

[usmonster]

Turns out neither configurable nor enumerable are used in IE8 anyway (according to Microsoft--see "Requirements" section near bottom), even though enumerable: false is known to cause an error in IE8 for accessor descriptors such as this one... which you discovered the hard way. :] In any case, I think what you've done is good to go.

(Just out of curiosity, were you testing in standards mode or compatibility mode (or both) when you got the error?)

[WebReflection]

all tests against this page, bear in mind it's interactive (and yes, some browser such safari fails 'cause not fully compliant with W3C specs)

tests are usually under standard mode as quirk one I believe is not a real-world scenario anymore (hopefully)

[shawnbot]

• update textContent shim per @usmonster's fix via ie8

[WebReflection]

@shawnbot current ie8 textContent is as standard as it was with previous element-prototype file

[shawnbot]

Thank you so much, guys! I'll update the submodule as soon as I'm able and check off that item. @usmonster, can you suggest some unit tests for this?

[WebReflection]

@shawnbot all new entries in the textContent property for all constructors prototypes have been added as tests in ie8, not sure you should re-test in here ;-)

[shawnbot]

Great, thanks!

[lusse]

is this going to get merged into master or not? aight.d3.js isn't working with the latest d3 version. I'm guessing this version does?