fix(ESM): Add operators/package.json for ESM support, fixes #3227

[jayphelps]

See #3227

I didn't do anything for @reactivex/rxjs package because I'm hopeful for its elimination (#2916) and I'm not entirely sure how it works lol.

[rxjs-bot]

	Messages
📖

CJS: 1358.5KB, global: 726.7KB (gzipped: 117.3KB), min: 140.3KB (gzipped: 30.6KB)

Generated by 🚫 dangerJS

[jayphelps]

It's hard to keep everything straight, but AFAICT this also makes the whole path-mappings.js thing unnecessary if you're using a newish version of webpack or rollup.

Would be cool to still include instructions (and maybe path-mappings still) for people who are stuck on older build pipelines.

[simonbuchan]

Yes, if users only import from "rxjs" and "rxjs/operators", importing "rxjs/index" or (more realistically) "rxjs/Rx" would not be mapped.

[coveralls]

Coverage remained the same at 97.56% when pulling c6d9e8e on jayphelps:esm into 52cdfe8 on ReactiveX:master.

[simonbuchan]

This says "fixes #3227", but it doesn't handle webpack native tree-shaking? I'd have to check the quality difference between having "sideEffects": false and not in webpack 4, I think with the new uglifyjs in there too it should be able to optimize without it, but it might be slower or something.

[jayphelps]

it doesn't handle webpack native tree-shaking?

Can you clarify why it does not? (I tested this btw)

[simonbuchan]

See the example
It's why I called out "webpack native" tree shaking and mentioned uglify should still work - but presumably the option exists for a reason.

[jayphelps]

@simonbuchan oh yeah sideEffects: false might be able to be supported but that's outside the scope of the work I'm doing for this PR and it's being tracked in #3212 not #3227. Perhaps confusing the two? 😄

[jasonaden]

This looks correct.

I think we also need a similar package.json file in ajax, testing, and websocket, right?

[jasonaden]

Agreed on #3212 vs #3227. This PR should close #3227 (as long as a package.json gets added for other secondary entry points such as ajax, websocket, and testing).

#3212 will have another PR shortly (I've got that one).

[simonbuchan]

@jayphelps Probably! But #3227 reads to me an umbrella issue of "make it just work™".

@jasonaden Is "rxjs/Rx" going away too?

[jayphelps]

@jasonaden

I think we also need a similar package.json file in ajax, testing, and websocket, right?

Ohhh right! Thanks for catching that!

@simonbuchan sorry for the confusion! My primary intention with that ticket was how to prevent two copies of RxJS. 🕺

[jayphelps]

@jasonaden I've updated to include package.json's for ajax/websocket/testing

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.