Try assigning fetch to globalThis if global assignment fails

[sebmarkbage]

In case it's a more modern yet rigid environment.

[sizebot]

Comparing: 1720405...7461214

Critical size changes

Includes critical production bundles, as well as any change greater than 2%:

Name	+/-	Base	Current	+/- gzip	Base gzip	Current gzip
oss-stable/react-dom/cjs/react-dom.production.min.js	=	152.77 kB	152.77 kB	=	48.73 kB	48.73 kB
oss-experimental/react-dom/cjs/react-dom.production.min.js	=	154.69 kB	154.69 kB	=	49.34 kB	49.34 kB
facebook-www/ReactDOM-prod.classic.js	=	526.93 kB	526.93 kB	=	94.15 kB	94.15 kB
facebook-www/ReactDOM-prod.modern.js	=	512.19 kB	512.19 kB	=	91.98 kB	91.98 kB
facebook-www/ReactDOMForked-prod.classic.js	=	526.93 kB	526.93 kB	=	94.15 kB	94.15 kB

Significant size changes

Includes any change greater than 0.2%:

Expand to show

Name	+/-	Base	Current	+/- gzip	Base gzip	Current gzip
oss-stable-semver/react/cjs/react.shared-subset.production.min.js	+0.45%	7.84 kB	7.88 kB	+0.40%	3.24 kB	3.26 kB
oss-stable/react/cjs/react.shared-subset.production.min.js	+0.44%	7.87 kB	7.90 kB	+0.40%	3.27 kB	3.28 kB
oss-experimental/react/cjs/react.shared-subset.production.min.js	+0.42%	8.27 kB	8.31 kB	+0.38%	3.39 kB	3.40 kB

Generated by 🚫 dangerJS against 7461214

[sophiebits]

Why not try globalThis first?

[gnoff]

left a nit/question

[gnoff]

Isn't this like typing it as any when it is quite unsafe to do so? why not just suppress the flow warning where you attempt to reference it if you don't want to existence check it?

[jdalton]

Why not replace user references to fetch with the React one, during build/transform, without overwriting the one on the global object?

[sebmarkbage]

This is specifically for all kinds of code/libraries that’s already written for fetch and not specifically for React necessarily. Lots of code gets it from the global object (eg window.fetch()) and get to the global object in different ways.

[sebmarkbage]

Also React itself doesn’t have a compiler anyway so no way to make it universal that way anyway.

[jdalton]

This is a React package though (so React centric). At glance this looks like something that could continue to be a package and then shipped with a companion babel (and friends) transform in much the same way JSX is translated. Folks can always bolt on the global if they really want to but doing it by default seems less than great.

[sebmarkbage]

It's really worth talking about when it actually kicks in, what it actually does even if it does kick in, that it doesn't add any new methods, options or anything to the prototype and that it gracefully backs off if you're in a frozen environment.

If it was a separate package, everyone mistaking installing it in their custom runtimes would miss out and not know why. If it was highly recommended everywhere and everyone used it anyway, what is it protecting against?

Other libraries already do this less gracefully so whatever changes the platform would do is already busted. We're not the first by far.

Also, this is not done yet and there's a larger RFC to describe the whole story before people jump to conclusions.

[jdalton]

It's really worth talking about when it actually kicks in, what it actually does even if it does kick in, that it doesn't add any new methods, options or anything to the prototype and that it gracefully backs off if you're in a frozen environment.

In frozen environments folks trying to rely on it couldn't where as if user code was transpiled with fetch() references replaced with the React flavor then they could still use this cacheable fetch.

If it was highly recommended everywhere and everyone used it anyway, what is it protecting against?

Highly recommended in the React ecosystem is different than the entire platform.

Other libraries already do this less gracefully so whatever changes the platform would do is already busted. We're not the first by far.

I mean yes, libraries have done this but it's not been great for the language/ecosystem.

Also, this is not done yet and there's a larger RFC to describe the whole story before people jump to conclusions.

Okay.

[sebmarkbage]

Highly recommended in the React ecosystem is different than the entire platform.

Sorry I meant if we recommended the explicit patching in the entire React ecosystem vs implicit. If it was always separate import or compiled by scope it would be a different story, but that's not enough.

libraries have done this but it's not been great for the language/ecosystem.

I mean specifically for fetch(). Hydration based libraries patch it to restore fetches that happened on the server.

[jdalton]

@sebmarkbage Do you have a reference to the larger RFC?

[iRunner]

https://github.com/acdlite/rfcs/blob/first-class-promises/text/0000-first-class-support-for-promises.md

Is this it?

[gaearon]

No, the relevant RFC isn’t finished or published yet.