feat(snapshot): support snapshotResolver and snapshotSerializers written in ESM

[chentsulin]

Summary

Part of #11167.

See the discussion on the snapshotResolver and snapshotSerializers config: #11167 (comment) #11167 (comment)

Test plan

Integration test added.

[chentsulin]

This runJest function actually doesn't return a promise, so we should find a way to test this kind of stuff.
This test got failed after I put runtime.requireModule into a async function.

From:

const esm = runtime.unstable_shouldLoadAsEsm(testPath);

if (esm) {
  await runtime.unstable_importModule(testPath);
} else {
    runtime.requireModule(testPath);
}

To:

const localRequire = async <T = unknown>(path: string): Promise<T> => {
  const esm = runtime.unstable_shouldLoadAsEsm(path);

  if (esm) {
    return runtime.unstable_importModule(path) as any;
  } else {
    return runtime.requireModule(path);
  }
};

await localRequire(testPath);

[SimenB]

runJest spawns jest in a shell, so it should not be affected by any internal changes in Jest

[chentsulin]

I have no clue what to do here, but I found this method sometimes returns vm.SourceTextModule as result.

[SimenB]

There is no "result" of executing a module. Do we need access to evaluated module? We haven't needed that before... We might need an unstable_importInternalModule which essentially does what you do here (wait for evaluatedModule.status === 'evaluated', then fetch its default export).

Essentially faking an import() call

[chentsulin]

If i didn't get it wrong, the result is used by this localRequire call to support ESM resolvers:

const custom: SnapshotResolver = await localRequire(
  snapshotResolverPath,
  true,
);

[chentsulin]

Add a boolean parameter to localRequire to support interopRequireDefault here.

[Maxim-Mazurok]

@chentsulin I wonder if this could be finalized or if it is waiting for a re-review from @SimenB ?...

We need this here: https://github.com/Maxim-Mazurok/google-api-typings-generator/blob/062854a2021be2dd244855261a73f5172bf597c2/custom-resolver.cjs for the use-case of using one snapshot file per test.

Thank you!

[SimenB]

Main issue is that we load the resolver within the tests (#12014 (comment)), which means it's pretty much blocked by full ESM support

[Maxim-Mazurok]

Ah, I see, thank you for the update, we are re-considering our approach now. Cheers!

[SimenB]

It might make sense to not load it within, but that means people who have written it in TS or something that requires transpilation will fail. So not sure which tradeoff is the best here

[Anutrix]

It's been a year. Any updates?

[github-actions]

This PR is stale because it has been open 1 year with no activity. Remove stale label or comment or this will be closed in 30 days.

[Anutrix]

It's been another year. Any updates?