-
Notifications
You must be signed in to change notification settings - Fork 47k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Restrict effect return type to a function or nothing #14119
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why null
? It shouldn’t be allowed. Either you do return or you don’t.
Also the return value from the destroy function itself should always be void.
@sebmarkbage We allow useEffect(() => {
// An if statement is arguably clearer but this seems fine, too?
return source !== null ? source.subscribe() : null;
}, [source]); |
That's can be annoying if you use an arrow function. For example, sometimes "destroy" -type functions return a boolean: useEffect(() => {
mySet.add(thing);
return () => mySet.delete(thing);
}); |
The argument for arrows in destroy also applies to Might be better to just enforce a style that makes it clear that these are side-effectful void functions? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should be limited to only undefined return values both in the callback and the destroy function.
We ran into problems with that when we tried to check compatibility between |
c9074ed
to
a5d9d85
Compare
We already warn in dev if the wrong type is returned. This updates the Flow type.
a5d9d85
to
2cb5b8a
Compare
getStackByFiberInDevAndProd(finishedWork), | ||
); | ||
} | ||
effect.destroy = create(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@sebmarkbage Usually we coerce missing values to null before storing them in our internal data structures, but I think that's because we usually accept either, and null is preferred because it's less likely to be unintentional. But in this case, since we don't accept null, I can skip the type check in prod. Let me know if this doesn't make sense.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I believe there has been times where V8 has treated undefined as effectively a missing property in the hidden class rather than a reified value. So setting to undefined might mess with the hidden class. Not sure though.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok I'll leave it like this until we learn more, I suppose
warningWithoutStack( | ||
false, | ||
'useEffect function must return a cleanup function or ' + | ||
'nothing (undefined).%s%s', |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are there other common examples we can include here? So it doesn't sounds like a jargon/technical error message.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'll add a branch specifically for null
ReactDOM: size: -0.0%, gzip: 0.0% Details of bundled changes.Comparing: 70d4075...bec016d react-dom
react-art
react-test-renderer
react-reconciler
Generated by 🚫 dangerJS |
Ok the warning message now has three branches, based on the type of the return value. If you return null
If you return a promise
If you return anything else that's neither a function nor undefined
I'll merge and we can bikeshed the messages more if they still need work. |
* Add 16.8.0 changelog * Mention ESLint plugin * Remove experimental notices from the ESLint plugin README * Update CHANGELOG.md * Add more details for Hooks * fix * Set a date * Update CHANGELOG.md Co-Authored-By: gaearon <[email protected]> * Update CHANGELOG.md * useReducer in changelog * Add to changelog * Update date * Add facebook#14119 to changelog * Add facebook#14744 to changelog * Fix PR links * act() method was added to test utils, too * Updated release date to February 6th
* Add 16.8.0 changelog * Mention ESLint plugin * Remove experimental notices from the ESLint plugin README * Update CHANGELOG.md * Add more details for Hooks * fix * Set a date * Update CHANGELOG.md Co-Authored-By: gaearon <[email protected]> * Update CHANGELOG.md * useReducer in changelog * Add to changelog * Update date * Add facebook#14119 to changelog * Add facebook#14744 to changelog * Fix PR links * act() method was added to test utils, too * Updated release date to February 6th
* Add 16.8.0 changelog * Mention ESLint plugin * Remove experimental notices from the ESLint plugin README * Update CHANGELOG.md * Add more details for Hooks * fix * Set a date * Update CHANGELOG.md Co-Authored-By: gaearon <[email protected]> * Update CHANGELOG.md * useReducer in changelog * Add to changelog * Update date * Add facebook#14119 to changelog * Add facebook#14744 to changelog * Fix PR links * act() method was added to test utils, too * Updated release date to February 6th
…eturns nothing Summary: to match the changes in facebook/react#14119. ran `make` and the tests, seemed ok? Pull Request resolved: #7430 Reviewed By: bvaughn Differential Revision: D13927998 Pulled By: jbrown215 fbshipit-source-id: 1bc0a6673d5f1bc9a58baa7ec30fdd6aace34813
We already warn in dev if the wrong type is returned. This updates the Flow type.