Add support for computed class property names (#120)

[motiz88]

First time contributor here - someone please check this for sanity 😄

I think this implements #120 quite completely. I'm working on a PR for the corresponding changes required in Babel, tracked by babel/babel#4499.

[motiz88]

The CI failures (on Node 5 only, out of the whole build matrix) appear to be unrelated to the PR's content. (Right?)

[danez]

Thanks for contributing.

This looks really good, the only thing that I'm worried about is how the ambiguous cases are handled. For example this class:

class Bar {
   [x] = blub
   ['y'];
}

Does it contain two or one properties? I would assume one with a value of blub['y']. But I'm not sure what the spec says about this if it does.
What would your changes produce for this example?

[codecov-io]

Current coverage is 94.03% (diff: 100%)

No coverage report found for master at 015035c.

Powered by Codecov. Last update 015035c...f09ab3f

[motiz88]

My immediate answer is that ASI applies like everywhere else (and there are existing tests for literal-named properties reflecting this). Need to do a little bit of research before I can add more info.

[danez]

I found this
tc39/proposal-class-public-fields#25

Not sure what the current status is, but semicolons might be required in the case i posted.

[danharper]

There's some existing tests, which cover the ASI rules in this case for computed methods.

[danez]

Good catch. So it probably also throw with properties.

[motiz88]

There's also tc39/proposal-class-public-fields/pull/45 (an open PR at the moment) which suggests the semicolon is not in the spec, but IMHO it's safe to assume it will be (given past discussions and the nontrivial ambiguity otherwise).

[motiz88]

@danez your class Bar parses as having one property.

[hzoo]

Yeah ref babel/babel#3225 and revert pr babel/babel#3332

Can cc @jeffmo

[hzoo]

Looks fine to me? We made semicolons necessary before but it was reverted

[jeffmo]

The tests look right to me

[danez]

@jeffmo what will the spec say about this case? Or is it not yet decided?

class Bar {
   [x] = foo
   ['y'];
}

[jeffmo]

@danez: Grammatically, that would be the same as:

class Bar {
  [x] = foo['y'];
}

This is all just the practical fallout of supporting ASI in the language.