Best effort Module Resolution Algorithm implementation based on Node.js 25.2.1

[imakira]

Best effort Module Resolution Algorithm implementation based on Node.js 25.2.1 as documented in https://nodejs.org/api/esm.html#resolution-and-loading-algorithm.

The following features are not implemented:

• DETECT_MODULE_SYNTAX(source) is not implemented
• In ESM_FILE_FORMAT, node.js will try to guess the format of a file with no extension. I kept the original behavoir of esmFileFormat which is throwing an error (Otherwise it will break some tests on some other places where I dont want to touch...)
• ".wasm" files are not handled by ESM_FILE_FORMAT

This patch breaks one unit test:

@Test
public void dontImportCommonJs() throws IOException {
	final String src = "import('with-package').then(x => {throw 'unexpected'}).catch(console.log);";
	final String out = "TypeError: Unsupported file extension: '" + getTestRootFolderUrl() + "node_modules/with-package/alternative-index.js'\n";
	runAndExpectOutput(src, out);
}

The previous behavoir with regard to using import with CommonJS modules was:

• import() can't import a commonjs module with .js extension, in a package not marked with "type"="module"
• import() can import a commonjs module with .cjs extension.
• import() can import a commonjs module with .js extension but not in a package

As import/import() can be used to load JavaScript CommonJS modules stated in packages.html.
I believe it is fair to expect import/import() to be able to import CommonJS modules.
In this patch, import can import a commonjs module with a cjs extension, or with a js extension and in a package not marked with "type"="module".

I couldn't get WebAssemblySimpleTestSuite working correctly (error: WebAssembly API enabled but "wasm" language cannot be accessed! Make sure the "wasm" l
anguage is found on the module path and one of the permitted languages when creating the Context). All other tests are running correctly.

Another thing to note:

In this patch, if esmResolve returns a result and esmFileFormat returns CommonJs, it will pass the resoluted path directly to tryLoadingAsCommonjsModule. Instead of the original specifier.
I don't know if this is the best way to do it, but at least we can import a CommonJs module exported in exports using import now (ideally we should make require also support it).

This patch is based on #904

[oracle-contributor-agreement]

Thank you for your pull request and welcome to our community! To contribute, please sign the Oracle Contributor Agreement (OCA).
The following contributors of this PR have not signed the OCA:

• PR author: me@coruscation.net (@imakira)

To sign the OCA, please create an Oracle account and sign the OCA in Oracle's Contributor Agreement Application.

When signing the OCA, please provide your GitHub username. After signing the OCA and getting an OCA approval from Oracle, this PR will be automatically updated.

If you are an Oracle employee, please make sure that you are a member of the main Oracle GitHub organization, and your membership in this organization is public.

[imakira]

Hello, any update on this PR?

[woess]

@imakira thank you for your contribution! We'll try to review this PR soon, but we'll need the OCA check to be cleared first.
You mentioned in #960 (comment) that you already signed the OCA but that it is under review? Because it still does not show up as signed.

[imakira]

@woess Thanks for you reply!
Last time my application got rejected for reason, I re-submitted one just a moment ago.

[oracle-contributor-agreement]

Thank you for signing the OCA.

[iamstolis]

I couldn't get WebAssemblySimpleTestSuite working correctly ...

WebAssemblySimpleTestSuite needs wasm. You can build graaljs with wasm using mx --dy /wasm build (in graal-js project directory) and then execute the test-suite using mx --dy /wasm unittest WebAssemblySimpleTestSuite

Speaking of useful mx commands, you may want to reformat the code using the formatter that we use (using mx eclipseformat --primary). Of course, we can reformat it ourselves before the integration of this PR, but doing so will steal some credit from you (in the sense of git blame).

[iamstolis]

There is no need to check for !specifier.startsWith("#"). packageImportsResolve() is always called with a specifier that begins with # (as the first step of this method says).

It seems that you can also remove the check for #/ that is not part of the algorithm (anymore?).

[iamstolis]

You can use Strings.DOT instead of constant(DOT) here (and two lines above).

[iamstolis]

This method is always used to check whether there is exactly one occurrence of c in s. This can be implemented more efficiently. For example:

    private static boolean containsOnlyOne(String s, char c) {
        int first = s.indexOf(c);
        return (first != -1) && (first == s.lastIndexOf(c));
    }

[iamstolis]

This condition should check for index property keys. These are keys that are strings but can be parsed as array indices. So, the current check is incorrect. JSRuntime.isArrayIndex(key) is more appropriate here.

[iamstolis]

There is no need to check for enumerability. You can use ownPropertyKeys() instead of enumerableOwnNames()

[iamstolis]

Thank you for your contribution. We will be happy to integrate it into graaljs. I pointed out few things that can be improved. Let me know if you want to do that yourself. We can always do these changes ourselves when integrating your PR.

[imakira]

@iamstolis Thanks for pointing out the parts I have missed!

I have made the changes you suggested, formatted the code, fixed one unit test I somehow missed last time (unsupportedUrl test, about error message), and rebased the PR against the master branch.

Let me know if there's some thing I'm still missing.

[iamstolis]

@imakira, thank you again for your contribution. The PR has been merged into master branch. So, it will be part of daily builds and will appear in the next release.