feat(transformer): async-to-generator plugin.

[7086cmd]

Tests are still not passed. A lot need to do yet.

[graphite-app]

Your org has enabled the Graphite merge queue for merging into main

Add the label “0-merge” to the PR and Graphite will automatically add it to the merge queue when it’s ready to merge. Or use the label “hotfix” to add to the merge queue as a hot fix.

You must have a Graphite account and log in to Graphite in order to use the merge queue. Sign up using this link.

[codspeed-hq]

CodSpeed Performance Report

Merging #5590 will not alter performance

_{Comparing 7086cmd:feat/transform-async-to-generator (58863b5) with main (67ad08a)}

Summary

✅ 29 untouched benchmarks

[7086cmd]

It appears that the iteration of the ast tree incurs performance degradation. However, it necessitates the duplication of statements for replacement. Are there alternative approaches that could be considered?

[Boshen]

@7086cmd Can you hold onto this and let us take a look first? We may end up implementing this differently with architectural changes.

[7086cmd]

@7086cmd Can you hold onto this and let us take a look first? We may end up implementing this differently with architectural changes.

Ok. It does need some refactor, especially when it comes some polyfills here, we'd better do as what babel does: from js text to ast, which is visually more vibrant and more convenient. Replacing to multiple statements is also a handful problem.

[7086cmd]

For this PR, I would prefer what esbuild does: more visible and no need for refactor for multiple statements here: https://esbuild.github.io/try/#dAAwLjIzLjEALS10YXJnZXQ9Y2hyb21lNTAAYXN5bmMgZnVuY3Rpb24gYSgpIHsKICAgIGF3YWl0IDE7Cn0KCmFzeW5jIChkKSA9PiB7CiAgICBhd2FpdCAyOwogICAgYXdhaXQgZCgpCn0K.

[overlookmotel]

I've not dug into this in full, but at first glance, it looks like this is using ESBuild's technique, rather than Babel's.

I can see why you've chosen to do that, but I'm afraid we need to follow @babel/plugin-transform-async-to-generator exactly.

There's a couple of reasons for this:

1. Runtime performance

ESBuild's technique is easier to implement, and maybe faster, but the code it generates will be slower at runtime.

Input:

async function foo() {
  await bar();
}

Babel output:

var foo = _asyncToGenerator(function*() {
  yield bar();
});

ESBuild output:

function foo() {
  return __async(this, null, function*() {
    yield bar();
  });
}

Babel's version creates a single function foo. But ESBuild's version has the cost of generating a new function function*() every time you call foo.

Basically, Babel's version is better.

2. Testing

Our testing infra depends on passing Babel's tests. Sometimes it's annoying that we have to produce exactly same output as Babel, even in cosmetic things like variable naming. But the advantage is that we don't have to write loads of tests ourselves. I don't know about you, but I don't particularly enjoy writing tests!

Once we have the transformer passing all Babel's tests, our next step will be run the transformer on all source files for the top 3000 npm packages, like monitor-oxc does. We'll compare Oxc's output to Babel's output transforming the same files. Bombarding our transformer with so many test cases will no doubt reveal many bugs, and once we fix them we can have extremely high confidence that the transformer is behaving correctly. But for this to work, we need Oxc's transformer to emulate Babel exactly.

Only once we've got all that passing, then we'll bootstrap ourselves, and we can start making Oxc produce better output than Babel. But we haven't got to that stage yet.

[7086cmd]

some descs.

[Dunqing]

Thanks for your work! I am going to merge this PR and then complete this plugin later

[7086cmd]

Thanks for your work! I am going to merge this PR and then complete this plugin later

Thanks! Just a heads-up: this PR is from my personal forked repository, so we can't merge it using Graphite. If that's inconvenient, I'm willing to reopen it from the origin repository.

[Dunqing]

Thanks! Just a heads-up: this PR is from my personal forked repository, so we can't merge it using Graphite. If that's inconvenient, I'm willing to reopen it from the origin repository.

It doesn't matter, I have been rebasing the main branch

[Dunqing]

LGTM