test for for-of optimization on arrays and add it for array type anno…

[hzoo]

was going to do this earlier but I closed the pr

[codecov-io]

Codecov Report

Merging #4747 into 7.0 will increase coverage by 0.11%.
The diff coverage is 96.66%.

@@            Coverage Diff             @@
##              7.0    #4747      +/-   ##
==========================================
+ Coverage   85.21%   85.33%   +0.11%     
==========================================
  Files         284      284              
  Lines        9958     9984      +26     
  Branches     2780     2788       +8     
==========================================
+ Hits         8486     8520      +34     
+ Misses        971      965       -6     
+ Partials      501      499       -2

Impacted Files	Coverage Δ
.../babel-plugin-transform-es2015-for-of/src/index.js	95.49% <96.66%> (+1.37%)	⬆️
packages/babel-traverse/src/path/context.js	85.34% <0%> (-1.73%)	⬇️
packages/babel-traverse/src/visitors.js	85.71% <0%> (-0.96%)	⬇️
...bel-plugin-transform-es2015-classes/src/vanilla.js	90.63% <0%> (ø)	⬆️
packages/babel-traverse/src/scope/index.js	79.04% <0%> (+0.23%)	⬆️
...ages/babel-traverse/src/path/inference/inferers.js	90.12% <0%> (+1.23%)	⬆️
...ackages/babel-traverse/src/path/inference/index.js	54.09% <0%> (+6.55%)	⬆️
packages/babel-types/src/flow.js	40.74% <0%> (+9.25%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cce83a0...0ff5946. Read the comment docs.

[DrewML]

Nice!

[danez]

This assumes that people have the correct type annotation. But what if they do what we do? Like having typeannotations but not verifying them, because then even if the type annotation is Array it could be an object.

[hzoo]

Hm good point - would we be ok with that assumption? @jridgewell thoughts?

[hzoo]

Linking @Kovensky's suggestion too http://astexplorer.net/#/dh8bycNK1A/2

[jridgewell]

I think the biggest improvement here is when we can prove it's an array, so @Kovensky's would work just fine for me.

[hzoo]

Updated http://astexplorer.net/#/zoFUOACwxU/3

edit: http://astexplorer.net/#/gist/687d23df3377f2d0669be386ef8d9f35/4f4436dae16f9f7d29a6a14239626ce2b5084a34

[hzoo]

We could remove the type annotation check I guess and it would just check for binding/array.from

[jridgewell]

How about with arguments?

[hzoo]

@danez I think if you are going to opt-in to types, it's on you to make sure you run flow/ts. So it's a good assumption to make

[hzoo]

via @ljharb: Array.of, new Array(), Array, String.prototype.match, RegExp.prototype.exec, and friends

[hzoo]

Didn't add the type check yet but we can

[hzoo]

I think we can land this as is since I didn't add the type part so it's safe

[jridgewell]

I'd rather see this in

babel/packages/babel-traverse/src/path/inference/inferers.js

Lines 148 to 150 in 622672d

	export function CallExpression() {
	return resolveCall(this.get("callee"));
	}

.

[hzoo]

I'd be happy to let you finish this up if you want to refactor

[jridgewell]

Let's merge this, I'll add this to #5835, and you'll promise to review it. 😉