Allow special element access assignments to create declarations

[andrewbranch]

Fixes #32153

Allows expando properties to be created via element access:

function Paragraph(props) {
  return <p>{props.children}</p>;
}

Paragraph.displayName = "Paragraph"; // This has worked for a while
Paragraph["defaultProps"] = {}; // This PR enables this

[weswigham]

@andrewbranch I suspect this overlaps with #33220 quite a bit, if not is wholly consumed by it

[andrewbranch]

@weswigham I just checked out your PR, and the only place I can observe overlap is on this["str"] = something I actually observe no overlap, because I have a mistake on this["str"] = something

In my branch, all of these work (though the this assignments trigger implicit any errors, as they do with property access assignments) all of these work except for the this assignments, but they’re supposed to.

[andrewbranch]

Ok, now my PR works with this, and that’s the place we overlap.

[sandersn]

You might look into the spacing changes before merging this.

[weswigham]

You wanna merge with master and see what some of these (class members, especially) emit with declarations on? Since they're not latebound, I imagine they're just emitted as their names, but tests are nice (especially of names like "prop-with-dashes").

[andrewbranch]

Oh dear

[weswigham]

Interestingly enough, the late bound variant in the other PR is already in a good place wrt declarations, looks like: https://github.com/microsoft/TypeScript/pull/33220/files#diff-2157e3c82025f7ab3e125ece2a9007eb

I'll try to get that merged so we can reconcile and work on the changes to the declaration emitter -

[andrewbranch]

I think this will be for 3.8, as #33220 is still open—I won’t have time to fix conflicts, much less fix the JS declaration emit.

[weswigham]

Naw, I think this'll maybe still be in 3.7 - #33220 will be merged once the new test shows as passing and because of other things, the 3.7 beta snap is being delayed a bit.

[andrewbranch]

@weswigham this is fixed up with your PR now, just need to work on the declaration emit Monday morning 👍

[weswigham]

A nav bar test like tests/cases/fourslash/navigationBarItemsSymbols4.ts is probably pertinent. If only to comment that none of these show up in the navbar because they're all non-syntactic members.

Otherwise, just some nits that could be cleaned up.

[andrewbranch]

A nav bar test like tests/cases/fourslash/navigationBarItemsSymbols4.ts is probably pertinent. If only to comment that none of these show up in the navbar because they're all non-syntactic members.

They actually do show up. I updated that test and it exposed one more place in navigationBar that was checking exclusively for PropertyAccessExpression.

Ready to merge once tests run 👍