JS2 HIR Implementation

[ambiguousname]

Per #420, this is my ongoing effort to implement the HIR backend for Javascript and Typescript.

There's still quite a bit to be done (lots of backend functions with TODOs, entry-level matches for hir types, etc.), but this is for anyone to get a good look to see if there's any issues with the way that I'm approaching things.

This is mostly based on the Dart backend, with some added comments that were for my own understanding during development.

Let me know if there's anything I can clarify!

[ambiguousname]

Updated, all feature tests are now passing.

There are still quite a lot of TODOs and FIXMEs that involve things not covered in the feature tests, so I will be coming back to this later to try and fill in the gaps.

[sffc]

We should land something even if it's not complete because it allows other people to start contributing any missing pieces

[ambiguousname]

The notes doc I just removed had things that were mostly out of date. The only thing that I'd like to re-iterate here is that I make sure to test the build scripts on Node v20

[ambiguousname]

Note: test coverage for NestedBorrowFields would be good for a lot of features. I'll try to add that

[ambiguousname]

Since we're passing all checks, here's a list of everything that isn't done that was brought up during review.

Most of these I think can be iterated on in separate PRs:

• Is BackendAttrSupport::memory_sharing dead? Let's remove it if so. (#531)
• Lifetimes and edges (#525)
  • If we don't want to retain the selfEdge argument itself, we could this.#selfEdge.push(...selfEdge) (see example/ICU4XDataProvider for an example of this)
  • Are edges for structs, opaques, and slices implemented correctly? Do we clean all of these up correctly?
  • See Foo.mjs, Bar.mjs, and anything using BorrowedFields as an example.
  • Code passes tests, but garbage collection is not thoroughly tested.
• Helper functions in diplomatRuntime to minimize the amount of direct reference to wasm. (#522)
  • For instance, functions that return a DiplomatWrite buffer should do this.
• == vs. === for reading from buffers in JS. Is this desired behavior? (#523)
• Could we use valueOf and [Symbol.toPrimitive] for enumerators instead of enum.value or enum.ffiValue? (#524)
• Typescript tests for Javascript. Somewhat working, but require some fixes to the original JS backend. See #509
• Separate .d.ts templates with shared structures like MethodInfo (to avoid generating the same information twice for each type) (#532)
• New typescript and javascript tests for additional functionality (#526)
  • Tests for borrowing types like NestedBorrowedFields and their lifetimes.
  • Tests for reading slices (a slice of strings especially) from function output
  • Tests for iterators and iterables (see next bullet point)
• Disabling all special method support statements in lib.rs, re-enable in separate PRs (#527)
  • I've removed explicit support for iterators/iterables (though the code is still in there), we could test this in a separate PR later
• Is there a way we can verify the argument positions of our functions? (#528)
  • I found that we can just read the LLVM IR output as a possible place to test outputs, but I think it'd be nice to find a place where we can explicitly see how a function in Rust corresponds to a function in WASM. If anyone wants to go digging with me into the Rust and LLVM compilers, let me know.
• Are we supporting 'static lifetimes for items like slices? (#529)
• Abundant whitespace issues in templates (#530)

[ambiguousname]

If there's any changes we'd like to make to this PR let me know. I think I'd like to make a push to get this ready to merge into main and branching off these items into separate issues and PRs. But if there's anything we'd like to address in this PR, let me know!

[sffc]

Some follow-ups were filed; please open the remaining ones.

[ambiguousname]

Done!