[TIMOB-25650] Update to babel 7 ecosystem

[sgtcoolguy]

https://jira.appcelerator.org/browse/TIMOB-25650

• Update to babel 7 ecosystem.
• Add quick unit test that polyfilling works.
• Replace wrench usage with fs-extra.
• Bump version to 0.4.14

[sgtcoolguy]

@ewanharris Note that I have not yet tried dropping this into the SDK to see if it breaks it. I simply ran the existing tests and adde one new one for polyfilling. (the tests are not particularly comprehensive here, I'd be happy if anyone had some more tests to write - maybe just a number of files to pass through jsanalyze with different options to verify the output?).

[garymathews]

Should we move this outside of the if (opts.transpile) { and add some logic to include core-js to always polyfill?

[sgtcoolguy]

Well I think the issue was raised suggesting we should just always polyfill. I happen to agree we probably should always polyfill and should enable transpile by default, but didn't try to alter that behavior here. We should open a ticket for each and consider it as a team. It may make more sense in a 8.0.0 timeline, though it shouldn't be a breaking change.

[sgtcoolguy]

Sort of related: tidev/titanium-sdk#10112

[ewanharris]

Couple comments. Kitchensink currently errors out due to the core-js path in the app not being as expected.

[ewanharris]

It might be safer to pin the babel packages versions, from their previous betas breaking changes are still possible up until the RC.

[ewanharris]

core-js residing under @babel/polyfill seems to be causing issues for me in kitchensink as the require path is core-js/modules/es7.symbol.async-iterator which isn't the location as it's under @babel/polyfill, nuking the package-lock makes puts core-js back in the top level (but causes a whole host of changes).

I think something like the below for resolving the modules might work, but would require us to bump our min Node to 8.9 for the paths option in require.resolve (so we know we're getting the exact dependency @babel/polyfill wants), this makes kitchensink happy.

const polyfillPkgJSONKLocation = require.resolve(path.join('@babel/polyfill', 'package.json'));
const polyfillPath = path.dirname(polyfillPkgJSONKLocation);
const polyfillPkgJSON = fs.readJSONSync(polyfillPkgJSONKLocation);
const polyfillDest =  path.join(modulesDir, '@babel/polyfill');
if (!fs.existsSync(polyfillDest)) {
	fs.copySync(polyfillPath, polyfillDest);
}
for (const dependency of Object.keys(polyfillPkgJSON.dependencies)) {
	const dependencyPath = path.dirname(require.resolve(dependency, { paths: [ polyfillPath ] }));
	const moduleDstDir = path.join(modulesDir, dependency);
	if (!fs.existsSync(moduleDstDir)) {
		fs.copySync(dependencyPath, moduleDstDir);
	}
}

[sgtcoolguy]

@ewanharris I did not pin the babel versions but I did take your polyfill dependency copying suggestion and implemented it with recursion. I had to tweak what you had somewhat for it to work, though - and had to bump the min node to 8.9.0.

[ewanharris]

@sgtcoolguy, I'm just catching up on things but will get round to this as soon as I can. What timeframe do you think we're looking to land this in, babel shipped what they hope was the last beta last week, and are hoping to go RC->GA fairly quick (https://github.com/babel/babel/releases/tag/v7.0.0-beta.56). Do we want to land it pre-GA or hold off til GA (maybe landing around the 8.0.0 timeframe due to the node version requirements?)

[sgtcoolguy]

@ewanharris I've updated this PR. It now uses the babel 7.0.0 GA version. It also bumps the node minimum to 8.9.0 and therefore bumps the version to 1.0.0 of this package.