ES2015 & 2016 Features meta-issue

[samwgoldman]

Feel free to open an issue for any unchecked feature that doesn't already have one, but please try to keep discussion in a single issue if possible. Use cases (and test cases) are very welcome!

To report a bug with a completed feature, please open a new issue for the bug.

ES2015+

• let/const declarations (let/const #784)
• Regular expression u flag (Allow u flag on RegExp #2510)
• Regular expression y flag
• Binary literals (Support ES6 binary and octal literals #566)
• New octal literals (Support ES6 binary and octal literals #566)
• Arrow functions
• Default parameters (mostly supported, except using this: Flow error on using this.x as a default param in a class #1234)
• Rest parameters
• Spread operator
• super references
• Classes
• for-of loops
• Generators (ES6 Generators #534)
• Object literal property shorthand
• Object literal method shorthand
• Object literal computed properties (Support computed property keys #252)
• Template strings
• Tagged templates
• Tagged template fn strings param has a raw prop
• Tagged template fn strings and raw arrays are frozen
• Destructuring
• Default values in destructuring (Add support for destructuring parameters + default values #183)
• Unicode code point escapes
• Allow duplicate object literal properties (except proto)
• Modules
• new.target (new.target support #1152)
• Symbols (Add support for ES6 symbol types #810, Support typeof x === "symbol" #1015)
• export Foo from 'foo' (Is export something from 'something supported? #2403)
• object spread
• async/await
• async iterators (Implement async iteration proposal (stage 3) #1824)
• Shared memory and atomics (Add Shared Memory definitions #7811)
• String.prototype.matchAll (Add String.prototype.matchAll #7812)
• Promise.allSettled
• Lifting template literal restriction

Proposals

Stage 4

• String.prototype.replaceAll

Stage 3

• optional chaining (Support optional chaining #4303)
• BigInt (BigInt typechecking #7602)
• globalThis (Support globalThis proposal #7704)
• import.meta (Flow parser should support import.meta #6913)
• private instance methods and accessors
• nullish coalescing (Support 'Nullish coalescing' proposal #6257)
• class property initializers (Support ES7 class property initializers #850)
• numeric separators
• top-level await (Top-level await #7710)
• WeakRefs
• RegExp Match array offsets

Stage 2

• Promise.any
• function.sent
• throw expressions (Support for throw expressions #6605)
• Atomics.waitAsync
• new Set methods
• ArrayBuffer.prototype.transfer
• Temporal
• Array.isTemplateObject
• iterator helpers
• collection normalization
• class decorators
• method decorators

Stage 1

• do expressions (Support for Syntax Do Expression #2523)
• Emitter
• pattern matching (TC39 pattern matching syntax support #6353)
• pipeline operator (Support Pipeline Operator #5443)

Stage 0

• function bind operator (Support function bind syntax (stage 0) #1545)

[mroch]

this is super useful!

there are probably a bunch of new APIs that we need to add, too. for example, I just noticed that we don't have String.prototype.codePointAt (part of new unicode stuff)

[nmn]

Perhaps, add an ES7 section and mark async-await as done.

(BTW, I find it surprising that async-await is done before generators!)

[Globegitter]

It seems there is an issue with ES6 Class support. I am using atom/nuclide IDE and having the following code:

export default class Resource {
  constructor(req:string) {
    this.req = req * 2;
  }
}

assignment of property req Property req not found in Resource even though I am defining it here in the constructor.

[gabelevi]

@Globegitter, Flow wants you to declare class properties like so:

export default class Resource {
  req: number;
  constructor(req: number) {
    this.req = req * 2;
  }
}

Why do we require this? Well, if we don't make people declare properties then we can't detect things like

function getReq(resource: Resource): number {
  return resource.rep;
}

We do realize this is a little bit more work for the developer, but we're hoping the benefits outweigh the pain.

[Globegitter]

Aahh, I see, thanks @gabelevi For know I mostly want to use flow for its type inference with /* @flow weak */. The issue here is that I can not just put req; res; at the top, it does throw an 'Unexpected token;` issue.

Generally I do think this is an acceptable extra-effort, but why can flow not just get the information from the constructor itself? Is it that much more complex to see what this.req is assigned to and then look up the type in the constructor's parameter list?

As a side note kind of fitting to this, it would be great to see property initialiser support as mentioned here: http://babeljs.io/blog/2015/06/07/react-on-es6-plus/

[gabelevi]

I think at the moment that Flow requires type annotations for the class member properties you declare. So

req: any; res: any;

should work. I did recently change the parsing of that (7e5b2a9) in preparation for class property initializers. We absolutely are planning support for class property initializers. They solve so many problems! The one I hate the most is how you have to do

class Foo {
  constructor() {
    this.myCallBack = this.myCallBack.bind(this);
  }
  myCallBack() { ... }
}

but with class property initializers you can do

class Foo {
  myCallBack = () => { ... }
}

[Globegitter]

Thanks for the quick response that all sounds good :) And yep it really does solve a lot of issues.

[samwgoldman]

Generally I do think this is an acceptable extra-effort, but why can flow not just get the information from the constructor itself? Is it that much more complex to see what this.req is assigned to and then look up the type in the constructor's parameter list?

@Globegitter, good insight! Actually, flow used to do this, but the functionality was removed. I'm not sure about the details surrounding this, but there is some explanation in a comment in the source code.

Unfortunately, this change was from the days of "sync to upstream" so the diff is big and it's hard to see the implications in terms of the code changes themselves.

[oztune]

^ What happened with let support?

I think it'd help if this was taken care of if only because at a glance it makes the whole project seem to be behind. For example it's being used as a con on this list, which happens to be a top result for googling "flow vs TypeScript"

http://www.reddit.com/r/javascript/comments/39cere/typescript_vs_flow_results_from_our_investigation/

Also in this comment:

http://www.reddit.com/r/javascript/comments/39cere/typescript_vs_flow_results_from_our_investigation/ct6xagb

Flow looks really promising, keep up the great work!

[samwgoldman]

What happened with let support?

We're actively working on it. Please be patient.

[nmn]

@oztune by all means use my fork in the mean time: github.com/nmn/flow

• Up to date with flow itself
• Let/const support added by making flow treat them like var. (if you use Babel, the block scoping and const checking is already done for you)

[gregwebs]

classes are checked, but class property initializers are not supported: can we have a check box for that? Making it hard for me to swallow ES6 classes right now.

An alternative approach to much of this work is to leverage babel more

[samwgoldman]

@gregwebs class property initializers are not ES6/ES2015, but it's definitely on our radar. I think, instead of creating a separate meta issue for ES2016 features, we should just track that here. I'll make the necessary changes.

[gregwebs]

Ah, in that case I will take any hack I can get. Seems to work to put this in the constructor:

      // fix broken es6 classes
      _.each(Object.getOwnPropertyNames( Object.getPrototypeOf(this) ), prop => {
        this[prop] = this[prop].bind(this)
      })

[Gozala]

@samwgoldman I think it would make sense to update #431 to #784 given that's where the ongoing work is now.

[samwgoldman]

@Gozala You're right! Thanks & done.

[cesarandreu]

I'd vote for adding export extensions to the ES2016 list. (Not sure if they're supported?)

[samwgoldman]

Is there an maintained list of the various proposals, along with the stage? Official would be ideal, but anything would help.

[cesarandreu]

@samwgoldman there's a list of the ones that are available in babel.

[DimitryDushkin]

export default from './some-component';

throws "Unexpected string" error.

Issue is resolved via some "hack":

export { default } from './some-component';

[Jessidhia]

That is not some hack, that is the correct syntax (reexport the "default" export as an export named "default"). export default from is not valid syntax, but maybe the error message could be better.

[Pauan]

There is a proposal to add in this feature: [link]

And a related proposal: [link]

[DimitryDushkin]

Thank you. I thought it's already in specification, because webpack 2 understands export default from ....

[langri-sha]

do expressions are now stage 0. Here's is the list of active ECMAScript proposals for the curious.

[lyleunderwood]

do expressions are now stage 1.

[leoselig]

@samwgoldman I believe this

• Default parameters (mostly supported, except using this: Flow error on using this.x as a default param in a class #1234)

is already done as of d3887c2.

Is this list still maintained?

[grundiss]

Still no support for do-expressions? :C

[PinkaminaDianePie]

Any news about do-expressions? They are extremely useful for FP-style, but I can't use flow together with them.

[verekia]

Throw expressions: #6605

[bradennapier]

do expressions ftw!

[bradennapier]

Came here to say "do expressions ftw!" but I see I already did this in July :-\ heh

[Xunnamius]

Not seeing flow errors for throw expressions would be incredibly nice!

[SPAHI4]

still waiting for do expressions

[goodmind]

@SPAHI4 it is not standard

[goodmind]

Updated list in post to reflect latest proposals

[fredericgermain]

I'd be very nice to have a status on import.meta (#6913)
@wchargin made a pull request a year ago and again two weeks ago, but @mroch in charge of it never commented. #6958

[gkz]

This issue is old and no longer up to date. We can track the remaining features left in individual issues.

[noahehall]

@gkz the top-level await issue is closed :(

blah nvm found it ;)