Fixes https://github.com/vercel/next.js/discussions/32233⚠️ If you're looking at this PR please read the complete description including the part about incremental adoption.
## TLDR:
Official support for `<Link href="/about">About</Link>` / `<Link href="/about"><CustomComponent /></Link>` / `<Link href="/about"><strong>About</strong></Link>` where `<Link>` always renders `<a>` without edge cases where it doesn’t render `<a>`. You'll no longer have to put an empty `<a>` in `<Link>` with this enabled.
## Full context
### Changes to `<Link>`
- Added an `legacyBehavior` prop that defaults to `true` to preserve the defaults we have today, this will allow to run a codemod on existing codebases to move them to the version where `legacyBehavior` becomes `false` by default
- When using the new behavior `<Link>` always renders an `<a>` instead of having `React.cloneElement` and passing props onto a child element
- When using the new behavior props that can be passed to `<a>` can be passed to `<Link>`. Previously you could do something like `<Link href="/somewhere"><a target="_blank">Download</a></Link>` but with `<Link>` rendering `<a>` it now allows these props to be set on link. E.g. `<Link href="/somewhere" target="_blank"></Link>` / `<Link href="/somewhere" className="link"></Link>`
### Incremental Adoption / Codemod
The main reason we haven't made these changes before is that it breaks pretty much all Next.js apps, which is why I've been hesitant to make this change in the past. I've spent a bunch of time figuring out what the right approach is to rolling this out and ended up with an approach that requires existing apps to run a codemod that automatically opts their `<Link>` usage into the old behavior in order to keep the app functioning.
This codemod will auto-fix the usage where possible. For example:
- When you have `<Link href="/about"><a>About</a></Link>` it'll auto-fix to `<Link href="/about">About</Link>`
- When you have `<Link href="/about"><a onClick={() => console.log('clicked')}>About</a></Link>` it'll auto-fix to `<Link href="/about" onClick={() => console.log('clicked')}>About</Link>`
- For cases where auto-fixing can't be applied the `legacyBehavior` prop is added. When you have `<Link href="/about"><Component /></Link>` it'll transform to `<Link href="/about" legacyBehavior><Component /></Link>` so that your app keeps functioning using the old behavior for that particular link. It's then up to the dev to move that case out of the `legacyBehavior` prop.
**This default will be changed in Next.js 13, it does not affect existing apps in Next.js 12 unless opted in via `experimental.newLinkBehavior` and running the codemod.**
Some code samples of what changed:
```jsx
const CustomComponent = () => <strong>Hello</strong>
// Legacy behavior: `<a>` has to be nested otherwise it's excluded
// Renders: <a href="/about">About</a>. `<a>` has to be nested.
<Link href="/about">
<a>About</a>
</Link>
// Renders: <strong onClick={nextLinkClickHandler}>Hello</strong>. No `<a>` is included.
<Link href="/about">
<strong>Hello</strong>
</Link>
// Renders: <strong onClick={nextLinkClickHandler}>Hello</strong>. No `<a>` is included.
<Link href="/about">
<CustomComponent />
</Link>
// --------------------------------------------------
// New behavior: `<Link>` always renders `<a>`
// Renders: <a href="/about">About</a>. `<a>` no longer has to be nested.
<Link href="/about">
About
</Link>
// Renders: <a href="/about"><strong>Hello</strong></a>. `<a>` is included.
<Link href="/about">
<strong>Hello</strong>
</Link>
// Renders: <a href="/about"><strong>Hello</strong></a>. `<a>` is included.
<Link href="/about">
<CustomComponent />
</Link>
```
---
## Feature
- [x] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
Remove the webpack breaking change message per error since it can be contained in all errors and the current implementation causes to truncate other error messages
Follow up for https://github.com/vercel/next.js/pull/36190
## Bug
- [x] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## History
In PR #35889, `onError` prop was added explicitly on the `Image` component, but it was ommitted from the `ImageElement` component. In result, the callback didn't work at all.
## Now
This PR deletes the `onError` prop from explicitly defined props. Explicitly adding `onError` on the `const imgElementArgs` is considered, but deleting the defined prop from `Image` is more reasonable because the prop isn't used in `Image` component.
## Bug
- [x] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
Co-authored-by: Steven <229881+styfle@users.noreply.github.com>
Changes to the beforeInteractive strategy to make it work for streaming
Splitting `beforeInteractive` into two strategies `beforeInteractive` at the _document level and `beforePageRender` for page level <Scripts>
* Fix build activity indicator position
`devIndicators.buildActivityPosition` introduced in
https://github.com/vercel/next.js/pull/30109 needed more tweaks to properly
position the build indicator container. The build indicator was being rendered
off screen when set to a non-default position.
* Refactor stuff for smaller diff
* add config validation
* Apply suggestions from code review
Co-authored-by: Martin Šťovíček <martin.stovicek@monitora.cz>
Even though we never documented it, the `onLoad` prop used to work (in most cases) in Next.js 12.1.4 and broke in 12.1.5 once the code was refactored in PR #35889.
We still shouldn't document it because `onLoad` doesn't work in all cases so [`onLoadingComplete`](https://nextjs.org/docs/api-reference/next/image#onloadingcomplete) is preferred, however tests were added for the cases that do work.
Use cases that don't work with `onLoad` are Data URLs as well as `loading="eager"`.
The `config` was changed from a global variable to a function parameter of the `loader()` function in PR https://github.com/vercel/next.js/pull/33559 as an implementation detail, not as a public API. It was not meant to be used by the end user which is why it was [undocumented](https://nextjs.org/docs/api-reference/next/image#loader).
This config is meant for the default Image Optimization API. Since the `loader` prop bypasses the default Image Optimization API in favor of a custom function, that config is no longer be relevant because the function can implement the optimization url however it desires.
- Fixes#35115
There are several places where the decoder should have `stream: true`. Also changed the `encodeText(extraChunk + decodeText(chunk))` pattern to just emit two chunks.
## Bug
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
## History
In PR #24153, `placeholder=blur` was added and it set `element.style.backgroundImage = 'none'` instead of using react state to re-render. Then in PR #25916, a delay was added to handle removing the blur placeholder. Then in PR #25994, `img.decode()` was utilized but we found this caused problems in Firefox in #26011.
## Today
This PR changes the the blur placeholder removal to use react state to re-render. This _should_ prevent Firefox from erroring although we should probably keep the catch() just in case. This was pointed out when React 18 caused subtle differences in Firefox in this comment https://github.com/vercel/next.js/issues/30128#issuecomment-1086754374
x-ref: #33149
RFCs:
- #30996
- #31506
## Feature
- [x] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
* fix(#35286): reset visible state when image src changed
* fix(#35286): reset visible state when link href changed
* test(image): image with new src should be re-lazyloaded
* test: apply code suggestions from @styfle's review
* test: incorrect condition of image load check
This fixes an issue with how the sizes prop was being applied to images with the experimental raw layout. The only effects were on the raw layout, so won't effect any non-experimental functionality.
Part of #31506 and https://github.com/vercel/next.js/discussions/34179. This PR ensures that in the `nodejs` runtime, the flight data is statically stored as a JSON file if possible. Most of the touched code is related to conditions of static/SSG/SSR when runtime and/or RSC is involved.
## Bug
- [ ] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
Firefox doesn't support `TransformStream` but `ReadableStream` at the moment, change the implmentation of hydration to make firefox work with RSC
## Bug
Fixes#35763
x-ref: https://github.com/vercel/next-react-server-components/issues/38
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
* Update rsc refresh typing and doc example
* Update docs/api-reference/next/streaming.md
Co-authored-by: Shu Ding <g@shud.in>
Co-authored-by: Shu Ding <g@shud.in>
.substr() is deprecated so we replace it with .slice() which works similarily but isn't deprecated
Signed-off-by: Tobias Speicher <rootcommander@gmail.com>
Co-authored-by: Steven <steven@ceriously.com>
* Update next/link error when no children are provided
* update manifest
* Apply suggestions from code review
Co-authored-by: Balázs Orbán <info@balazsorban.com>
Fix error of `Unexpected server data: missing bootstrap script` when the initial rendering is streaming rsc.
Introduced in #35344, the buffer might still keep flushing after on mount.
Co-authored-by: Shu Ding <3676859+shuding@users.noreply.github.com>
### Bugfix
Made some changes for the data register buffer flushing in #34631 and #34475 that tried to delete the buffer or flush them only once. But turns out it will break the navigation between RSC pages.
### Enhancements
Simplify the inline response writer and inline response data for the initial render. Since they're only for the initial render, navigations will leverage the serialized data fetched from router and construct the react tree.
## Bug
Fixes#35135
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
This PR adds a new layout mode for images called `raw`, as discussed with the core team a while back. This mode has the following characteristics:
- No wrapper `span` around the `img` element
- No sizer svg
- Almost no styles automatically added to the `img` element
- `style` parameter is allowed and is passed through to the underlying `img` element
This also adds documentation changes to describe the new component.
There are a few tradeoffs and DX decisions that may warrant discussion/revision before merging. I'll add a few comments to highlight those issues.
- Related to #18637
## Summary
This PR adds a new `worker` strategy to the `<Script />` component that automatically relocates and executes the script in a web worker.
```jsx
<Script
strategy="worker"
...
/>
```
[Partytown](https://partytown.builder.io/) is used under the hood to provide this functionality.
## Behavior
- This will land as an experimental feature and will only work behind an opt-in flag in `next.config.js`:
```js
experimental: {
nextScriptWorkers: true
}
```
- This setup use a similar approach to how ESLint and Typescript is used in Next.js by showing an error to the user to install the dependency locally themselves if they've enabled the experimental `nextScriptWorkers` flag.
<img width="1068" alt="Screen Shot 2022-03-03 at 2 33 13 PM" src="https://user-images.githubusercontent.com/12476932/156639227-42af5353-a2a6-4126-936e-269112809651.png">
- For Partytown to work, a number of static files must be served directly from the site (see [docs](https://partytown.builder.io/copy-library-files)). In this PR, these files are automatically copied to a `~partytown` directory in `.next/static` during `next build` and `next dev` if the `nextScriptWorkers` flag is set to true.
## Checklist
- [X] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [X] Related issues linked using `fixes #number`
- [X] Integration tests added
- [X] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
This PR fixes#31517.
* fix: add try/catch to gracefull handle performance observer exception on older browsers
* fix: run prettier
Co-authored-by: Steven <steven@ceriously.com>
For streaming, we want the main bundle to be loaded with `async` rather than `defer`. As the first step to doing that, we need an explicit initialization step, rather than occurring as a side effect of being loaded.
For now, I've just split the logic into a `initialization` and `hydrate` (previously `initNext`) functions and called them as a side effect of `next/client/next` and `next/client/next-dev` as before. A follow-up PR will move that to a separate script tag [as recommended](https://github.com/reactwg/react-18/discussions/114).
### Changes
* Remove top-level suspense boundary
* Pipe stream resolved from returned promimse of `renderToReadableStream`
* Remove jsx-runtime alias hack
### Test Changes
Since top level suspense boundary is removed, now content are filled in 1st SSR
x-ref: #34021 , #34004
Only log each warning once and only trigger in concurrent mode
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
## Bug
If there's upcoming streaming data from server components, should safely skip the bootstrap process. Previously we deleted the buffer then it will cause the buffer is missing in the later re-renders. Now we mark it as empty array, so it can safely skip the boostrap phase
x-ref: #34475
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
* Move type to image component
* Add types/global.d.ts to excludes too
* Undo global exclude as it's using internally
* Don't add root imports for module augmentations
Currently, we print warnings during `next dev` for every render of `next/image`, which can quickly fill the console making it really unfriendly to developers trying to read the logs.
This PR changes the behavior so that each unique warning prints at most once.
- Related to #33007
- Related to #31340
Implements https://github.com/vercel/next.js/issues/30997 with some minor tweaks to the design:
* The hook is moved to Client Components (e.g. `pages/_app` instead of `pages/_document`). This was a silly oversight in the original design: the hook needs to be called during server prerendering.
* `useFlushEffects` instead of `useFlushEffect` as there isn't a particularly safe way to implement the singular semantics as a Client Component hook given the current implementation of server rendering.
---
Fixes#30997
## Bug
Fixes#34464
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
Fixes: #34342
Visiting the following page will call gSSP indefinitely in a loop and logs errors from `on-demand-entries-client`:
```js
const Home = () => null
export default Home
export function getServerSideProps() {
console.log("gssp called")
return { notFound: true }
}
```
We should not keep fetching the page if it returns 404 as it can introduce unnecessary data requests.
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
x-ref: #31827
x-ref: #34212
## Bug
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
As the comment mentioned, React 18 with Strict Mode enabled might cause double invocation of lifecycle methods. This makes the `<RouteAnnouncer/>` being non-empty for the initial page, which is a bug (it should only announce page change whenever a navigation happens).
## Bug
- [ ] Related issues linked using `fixes #number`
- [x] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes by running `yarn lint`
### Fixes
* We need give value for `__flight__` query or it will lost in node runtime
* Match the cache key of flight response (`path + search`, `route, `id`)
* Flight response should return `200` instead of `404` for 404 page
### Tests
* Run rsc for node runtime test suite
* Add more cases for 404
Fixed lazyRoot functionality (#33290). Changed the unique id for Intersection Observers discrimination since previously they were only identified by the different rootMargins, now each being identified by the rootMargin and the root element as well
Added more Images to the test with different margins and with/without lazyRoot prop. Browser correctly initially loading two of the four Images according to the props' specifications.
Co-authored-by: Steven <steven@ceriously.com>
* Allow scroll prevention on hash change
Currently, `scrollToHash` is performed on every hash change, even when this change is caused by `<Link scroll={false} {...props} />`.
This change prevents scrolling in this case and allows users to specify the desired scrolling behavior in the router's `hashChangeComplete` event.
* Add test case and apply fixes
Co-authored-by: JJ Kasper <jj@jjsweb.site>
Currently, if you render `next/link` like
```ts
<Link>
<CustomComponent />
</Link>
```
and have
```tsx
interface Props {
onClick?: () => void; // <— note we're not passing event as an argument
}
function CustomComponent({ onClick }: Props) {
return <div onClick={() => onClick?.()}>Hello</div>
}
```
It'll result in error
```
link.js?f421:21 Uncaught TypeError: Cannot read property 'defaultPrevented' of undefined
```
<!--
Thanks for opening a PR! Your contribution is much appreciated.
In order to make sure your PR is handled as smoothly as possible we request that you follow the checklist sections below.
Choose the right checklist for the change that you're making:
-->
## Bug
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
## Feature
- [ ] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR.
- [ ] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Documentation added
- [ ] Telemetry added. In case of a feature if it's used or not.
- [ ] Errors have helpful link attached, see `contributing.md`
## Documentation / Examples
- [ ] Make sure the linting passes
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
