Test ES6 code in PhantomJS

[bajtos]

Right now, we use PhantomJS headless browser to run (some of) loopback tests in a browser environment. Unfortunately, PhantomJS does not support ES6 yet - see ariya/phantomjs#14506

To allow us to write ES6 code, we need to add some sort of ES6-to-ES5 transpiler to our Grunt pipeline.

[ritch]

@bajtos anyone else able to take this on?

@kjdelisle @pthieu ??

[bajtos]

anyone else able to take this on?

Ah, sure, of course.

So far, we are using browserify to build the browser bundle (see test/karma.conf.js#L19) and adding es5-shim to add missing ES5 APIs (see test/karma.conf.js#L23).

Unfortunately, ES6 is not only about extra APIs, we need to add ES6-to-ES5 transpiler to the mix. I think it should be enough to configure karma-browserify to transform the sources (transpile, perhaps using Babel?), the plugin already supports custom transformations - see https://github.com/nikku/karma-browserify. We may need to add karma-es6-shim to the mix too.

That's a quick intro. I didn't have time to look into this issue, so take my comment with a grain of salt.

[pthieu]

I may be able to take this on. Have transpiled and used grunt in production; but not together with tests. Might need some hand-holding. What's the priority on this?
@ritch @bajtos ^

[superkhau]

IMO, this is very important to having the loopback repo pass (I think there are windows issues related to phantom too).

[bajtos]

What's the priority on this?

I would prefer to get this done ASAP. We have already spent a lot of time and effort to allow us to use ES6 in our code, I think this is the last missing piece.

On the other hand, this is not related to PhantomJS passing/failing in the current code, therefore it's not a critical problem.

I have explicitly disabled ES6 syntax in eslint config of loopback. For upstream dependencies like strong-remoting and loopback-datasource-juggler, we are relying on downstream CI builds to detect the situation where an ES6 code would break PhantomJS tests.