fix(transformer/decorator): transformed decorators should be injected after class-properties has run

[Dunqing]

• close [Bug]: Incorrect ordering of static class members and decorators rolldown/rolldown#5374

Example:

Input:

function myPropertyDecorator() {
  return (target: any, propertyKey: number) => {
    console.log(target.constructor.modelName, propertyKey);
  };
}

class Test {
  public static modelName = 'Test';

  @myPropertyDecorator()
  public myProperty = 1;
}

const instance = new Test();

export default instance;

Output diff:

function myPropertyDecorator() {
        return (target, propertyKey) => {
                console.log(target.constructor.modelName, propertyKey);
        };
}
class Test {
        constructor() {
                babelHelpers.defineProperty(this, "myProperty", 1);
        }
}
-babelHelpers.decorate([myPropertyDecorator()], Test.prototype, "myProperty", void 0);
-babelHelpers.defineProperty(Test, "modelName", "Test");
+babelHelpers.defineProperty(Test, "modelName", "Test");
+babelHelpers.decorate([myPropertyDecorator()], Test.prototype, "myProperty", void 0);

const instance = new Test();
export default instance;

The cause is that we transform class fields and decorators in two different plugins, and the decorator plugin should be transformed first, and then run the class-properties plugin, because class-properties might erase some decorator information. However, this cause transformed decorators injected before transformed class fields.

This PR fixes this problem by collecting all transformed decorators that wait for injection, and injecting them in batch after the class-properties plugin has run.

[Dunqing]

• fix(transformer/decorator): transformed decorators should be injected after class-properties has run #12418 👈 (View in Graphite)
• fix(transformer/typescript): should remove abstract field early #12417
• main

---

How to use the Graphite Merge Queue

Add either label to this PR to merge it via the merge queue:

• 0-merge - adds this PR to the back of the merge queue
• hotfix - for urgent hot fixes, skip the queue and merge this PR next

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

An organization admin has enabled the Graphite Merge Queue in this repository.

Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue.

This stack of pull requests is managed by Graphite. Learn more about stacking.

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #12418 will not alter performance

_{Comparing 07-21-fix_transformer_decorator_transformed_decorators_should_be_injected_after_class-properties_has_run (986c48e) with main (6838948)}

Summary

✅ 34 untouched benchmarks

[overlookmotel]

@Dunqing I've amended the diff in PR description. Before the diff showed no change. I think what I've changed it to is what you were intending to say, but can you please check my change and confirm?

[Dunqing]

@Dunqing I've amended the diff in PR description. Before the diff showed no change. I think what I've changed it to is what you were intending to say, but can you please check my change and confirm?

Thanks! Yes, it is good now.

[overlookmotel]

It feels a bit hacky to have 2 exit_class methods for the same transform, but I see why it's necessary.

The interaction between TS, decorators, and class props transforms feels a bit fragile, but I'm not sure how to improve it...

[overlookmotel]

Merge activity

• Jul 21, 3:13 PM UTC: The merge label '0-merge' was detected. This PR will be added to the Graphite merge queue once it meets the requirements.
• Jul 21, 3:13 PM UTC: overlookmotel added this pull request to the Graphite merge queue.
• Jul 21, 3:22 PM UTC: Merged by the Graphite merge queue.

[Dunqing]

It feels a bit hacky to have 2 exit_class methods for the same transform, but I see why it's necessary.

The interaction between TS, decorators, and class props transforms feels a bit fragile, but I'm not sure how to improve it...

Yes, I am either. Dealing with decorators around these three plugins is quite error-prone. I would consider your suggestion to add more tests that enable these three plugins.