Changed Attributes

[jridgewell]

By returning the changed attributes, we're able to avoid a second #apply.

Re: #85 (comment)
Re: #120 (comment)

[jridgewell]

Did some digging into inlining, seems changedAttributes won't be inlined do to the #apply call, but updateAttributes and everything it does gets inlined into elementOpen:

Inlined getData called from updateAttribute.
# snip...
Inlined updateAttribute called from updateAttributes.
Inlined updateAttributes called from exports.elementOpen.

For reference, here's the full log.

[jridgewell]

Pinging this again. Not seeing any difference on the creation or selection performance tests.

[sparhami]

It is a bit confusing, but the perf tests use the -min.js from the dist folder (generated by js-min). I tried running those and I definitely do see a difference between your changes and the current state.

[jridgewell]

Yup, I run gulp js-min before the each test, one for master and one for the changes. I'll rerun on another computer to see if I get anything different.

[jridgewell]

Browser	Test	Branch	Result (ms)
Chrome	Creation	Master	3.2
Chrome	Creation	changedAttributes	3.2
Chrome	Selection	Master	0.32
Chrome	Selection	changedAttributes	0.32
FireFox	Creation	Master	8.2
FireFox	Creation	changedAttributes	7.4
FireFox	Selection	Master	0.64
FireFox	Selection	changedAttributes	0.63

[sparhami]

I tried this a while ago and wasn't able to reproduce your results. I don't plan on changing this for now.