Tests for the RSC shim

[lubieowoce]

Just sketching some stuff out. Not sure if i like the conditional bits... but it's hard to do that otherwise w/o a ton of duplication. for example, useContext seems to exist in the RSC world in the React version we're installing! Lemme know what you think or if you have ideas on how to structure this better.

[phryneas]

Hmm, just a thought: might it make sense to directly import the react dist files for the different environments and compare those with the rsc export?

Generally: A little bit of duplication is not a bad thing - too much abstraction makes tests harder to read :)

It's already late, so I'll take a closer look tomorrow!

[lubieowoce]

A little bit of duplication is not a bad thing - too much abstraction makes tests harder to read :)

Oh don't worry, the last commit is already slightly de-abstractionized :D the original was worse lol. But lmk which bits you think we should de-abstractionize further, i have a penchant for deduplicating my tests to death lol

might it make sense to directly import the react dist files for the different environments and compare those with the rsc export?

well, the issue is that rsc.js imports react directly. so we can import anything we want in the test, but it might not be what the shim imported, so there's no point. we could use module mocking, but node:test doesn't have that, so we'd need to pull in something that does.

...but also i think it'd be better to stick to plain --conditions, no mocking/resolution tricks. that way we can kinda ensure we're testing the real behavior. we can just have two test scripts, one for RSC, one for vanilla

[lubieowoce]

TODO - tests for classes

[phryneas]

well, the issue is that rsc.js imports react directly. so we can import anything we want in the test, but it might not be what the shim imported, so there's no point. we could use module mocking, but node:test doesn't have that, so we'd need to pull in something that does.

...but also i think it'd be better to stick to plain --conditions, no mocking/resolution tricks. that way we can kinda ensure we're testing the real behavior. we can just have two test scripts, one for RSC, one for vanilla

My idea was somewhere along the lines of

const browserReact = require('react/index.js')
const rscReact = require('react/react.react-server.js')

const rehackt = require('rehackt')

for (const key of Object.keys(browserReact)) {
  assert.ok(key in rehackt)
  assert.equal(typeof rehackt[key], typeof browserReact[key])
  assert.notEqual(rehackt[key], browser[key])
}
for (const key of Object.keys(rscReact)) {
  assert.equal(rehackt[key], rscReact[key])
}