Move to ES6

[lgeiger]

I would like to (re)evaluate if we should port our code base from CoffeeScript to ES6. Atom is slowly moving away from CoffeeScript as well.

My main motivation is to move Hydrogen closer to nteract and be able to use all ES6 features.

I just made a quick example plugin for Hydrogen in ES6 and was really surprised how easy it was.
We don't need a build step, we only need to add 'use babel'; on top of the file and Atom takes care of the rest.

I don't feel strongly against CoffeeScript but if think we should have this discussion.
@rgbkrk @n-riesco @nikitakit What do you think?

[rgbkrk]

I'm certainly more likely to contribute if it's ES6 instead of CoffeeScript.

[n-riesco]

Also note that since the latest Atom uses Node v4.4.5, in some cases 'use babel'; may be avoided too.

[lgeiger]

I'm certainly more likely to contribute if it's ES6 instead of CoffeeScript.

For me this argument is sufficiently to switch. 😄
And I imagine you are not the only one.

@n-riesco Node 4.4.5 doesn't support all ES6 features so this could lead to strange errors. With babel we would be definitely on the save side.

One sad thing, we would lose our position in the github developers ranking for coffeescript 😜

[n-riesco]

On 11/11/16 16:44, Lukas Geiger wrote:

@n-riesco https://github.com/n-riesco Node 4.4.5 doesn't support all ES6 features http://node.green/ so this could lead to strange errors. With babel we would be definitely on the save side.

I chose my words carefully for that reason. A few more thoughts:

• I think the most compelling argument to start using Javascript is to lower the hurdle our contributors have to jump.
• Using all the ES6 features may put off some developers that work mostly with browsers.
• Since version 1.0.0 is coming soon, this is a perfect time to plan this transition.
• Do you have anything in mind about what this transition to JS would look like?

[lgeiger]

I think the most compelling argument to start using Javascript is to lower the hurdle our contributors have to jump.

That's the reason why I raised this discussion.

Using all the ES6 features may put off some developers that work mostly with browsers.

I agree. I'm not familiar with them either but we could use them if we need to.

Do you have anything in mind about what this transition to JS would look like?

If we really wan't to move to JS I would propose this short term plan (post 1.0.0):

• I'll be traveling to NY on monday (wooo 🎉 ) and have some time during the flight to do boring tasks, so I volunteer to move the entire codebase to JS (including eslint). Probably with the help of decaffeinate. The only thing I fear and will probably miss is the existential operator in coffeescript.
• Then we should discuss (probably in a dev meeting) how we'd like to move forward. I really want to make use of more code from nteract (kernelspecs, enchannel, rx-jupyter, ...).

I don't have the background in JS you have. So if you say moving to JS just for the sake of change doesn't make sense, I'm absolutely OK with it.

[nikitakit]

No opinion on language change.

But if you do end up switching to JS, please make sure that all the files use strict mode. ('use babel' should include this, but otherwise you may need 'use strict' at top of each file)

[rgbkrk]

I'm happy even if I'm writing vanilla ES5, we probably aren't going to use most ES6 features. Even node 4 was a dramatic improvement.

[rgbkrk]

The only thing I fear and will probably miss is the existential operator in coffeescript.

That's my favorite feature from coffeescript that didn't end up as part of ES6.

Our best bet here is to rely on lodash heavily with _.has, _.at and _.get. Lodash has excellent deep path support. We already have it as a dependency if I recall correctly.

[rgbkrk]

It would be neat to see if you can convert single files within the code base to JS (with use babel; at the top) yet still require them from coffeescript in Atom.

[lgeiger]

It would be neat to see if you can convert single files within the code base to JS (with use babel; at the top) yet still require them from coffeescript in Atom.

That totally works!

[n-riesco]

The only thing I fear and will probably miss is the existential operator in coffeescript.

I never tried to convert a project from CS to JS, so this is going to be a good opportunity to gain some experience. At first, I'd try to decaffeinate only one module.

As Kyle suggests above, taking the route of _.has et al. seems the most straight forward replacement for ?. This route will raise the hurdle for new contributors and make more difficult to remove the dependency on lodash (if we ever wanted to). As a guideline, mostly for the sake of a lower hurdle for new contributors, I'd try to limit the use of lodash to the necessary cases.

I don't have the background in JS you have. So if you say moving to JS just for the sake of change doesn't make sense, I'm absolutely OK with it.

This was Slava's thought, but in this case there is a reason for changing to JS: gaining new contributors.