Fix splitbrain between .js being ESM & CJS inconsistently

[t3chguy]

Fixes #188

no actual changes to the js/cjs/mjs files barring some ts annotations

Switches the package type to ESM as that is what wasm-pack is outputting
Convinces the wasm bindgen test to use an ESM runner
Uses explicit extensions on the entrypoints between CJS/ESM

[richvdh]

would it be better to stop using .js altogether, and just call everything .mjs / .cjs, to avoid this sort of confusion?

[t3chguy]

That would also be possible but I somehow hate the idea of having to rename the output files of wasm-pack s/js/mjs - "type": "module" is also a good thing going forward. I can imagine also imagine issues going forward if the tests/entrypoints were ts-ified and needing mts/cts and the issues typescript has around those

Edit: ended up being explicit on the checked in files, still added "type": "module" for wasm-pack generated files and for ts downstream to be happier

[andybalaam]

@richvdh I assigned this to you since you seem to understand it more than me. Feel free to assign another one to me :-)

[richvdh]

no actual changes to the js/cjs/mjs files

This doesn't seem quite right? Could you give the description of the PR an update, please, to give an overview of the approach? (Looks like we're now standardising on .js meaning ESM, and .cjs meaning CommonJS?)

Also, please can you add an entry to the changelog?

[richvdh]

Can you help me understand what exactly this is doing? I get that it's to do with https://www.typescriptlang.org/docs/handbook/release-notes/typescript-5-3.html?#stable-support-resolution-mode-in-import-types, but what's the actual outcome/why do we need this?

[t3chguy]

Its what the error told me to do :D

[richvdh]

Hrm, fair enough. I assume the commandline yarn lint:types/tsc --noEmit showed the same error? Or was this specific to your IDE?

I'd be interested to understand why this is a problem now when it wasn't before.

tsc is now detecting node.cjs as CommonJs where it wasn't before? (which seems odd, why would it have done that?)

Or tsc now considers ./pkg/matrix_sdk_crypto_wasm_bg.wasm.d.ts as ESMish where it wasn't before? (seems odd, it's typescript)

[t3chguy]

Hrm, fair enough. I assume the commandline yarn lint:types/tsc --noEmit showed the same error? Or was this specific to your IDE?

Same error

/opt/homebrew/bin/yarn run lint:types
yarn run v1.22.22
$ tsc --noEmit
node.cjs:83:36 - error TS1542: Type import of an ECMAScript module from a CommonJS module must have a 'resolution-mode' attribute.

83  * @returns {Promise<typeof import("./pkg/matrix_sdk_crypto_wasm_bg.wasm.d")>}
                                      ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


Found 1 error in node.cjs:83

error Command failed with exit code 2.
info Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command.

Process finished with exit code 2

Or tsc now considers ./pkg/matrix_sdk_crypto_wasm_bg.wasm.d.ts as ESMish where it wasn't before? (seems odd, it's typescript)

Yes exactly that, because .mts and .cts exist, .ts esm/cjs-ness also depends on package.json type field.

[t3chguy]

This doesn't seem quite right? Could you give the description of the PR an update, please, to give an overview of the approach? (Looks like we're now standardising on .js meaning ESM, and .cjs meaning CommonJS?)

Have updated the description

Also, please can you add an entry to the changelog?

Could you please point me to any guidance on doing so? Not seeing anything in the contributing doc

[richvdh]

Also, please can you add an entry to the changelog?
Could you please point me to any guidance on doing so? Not seeing anything in the contributing doc

https://github.com/matrix-org/matrix-rust-sdk/blob/main/CONTRIBUTING.md#writing-changelog-entries is the most relevant guide.

TL;DR though: if you could just add an entry under the "UNRELEASED" heading, following the same style as previous entries, that would be be great.