This serves to address a few problems as it relates to headers handling within the App Page render pipeline.
1. During static generation, we should not write to the response directly as the revalidation for static responses is performed out of band from the actual request/response as a stale version can be served early. This is partially addressed by modifying the `onHeaders` to only write when there it is not during static generation, but long term I'd hope to see that we can instead pass a immutable request and `null` for the response to help indicate to the render pipeline that it should instead persist the data into the `RenderResult` objects. Alternatively, the response could use a mocked version that would then supersede the `RenderResult` and be used instead.
2. The types for rendering App Pages were merged incorrectly, so this additionally restructures the way that we initialize those renderers such that only the options that are applicable to each renderer is passed down.
## Story Time
Metadata API introduced two exports `metadata` and `generateMetadata` to the pages and layouts under app router, as partial prerendering work is going on and people are desiring to render the metadata asynchronizly, this change will be the preparation for moving to the dynamic & asynchronized land. In short, if we can render the metadata asynchronizedly, it will benefit the performance of the initial page loading and client page transiation a lot. Any slow data fetching can be handled while the essential page "shell" is rendered.
For meta tags, there're few ones will visually affect your web page, such as `<meta name="viewport">`, `<meta name="theme color">` and `<meta name="color-scheme">`, rendering them lately after the page frame is ready might bring flickering to the page such as revreting whole page's theme color or shaking due to viewport updates. Those meta are not majorly the "metadata" for SEO, but more for user experience when opening the page. If we're rendering everything as async meta tags, it won't be ideal due to the flickering on your web pages.
## Solution Preparation
We'll want to render the meta tags separately to make sure the visual ones are rendered as blocking along with web page, and then the ones for SEO or bots can be flushed later by later, like a suspense boundary keeps emitting them into the head of html.
We optionally picked up 3 meta tag "viewport", "theme-color" and "color-scheme" to be render first into the web page with html "shell", to guarantee the layout viewport and basic styling are rendered first.
This PR introduced two module export in the page and layout files: `viewport` and `generateViewport`, in order to separate the visual meta tags from the SEO metadata.
### API
```ts
// page.js | layout.js
export const viewport = {
// viewport meta tag
width: 'device-width',
initialScale: 1,
maximumScale: 1,
interactiveWidget: 'resizes-visual',
// visual meta tags
colorScheme: 'dark',
themeColor: { color: 'cyan', media: '(prefers-color-scheme: dark)' },
}
```
There's also a dynamic API like what we did for metadata API
```ts
// page.js | layout.js
export function generateViewport() {
return { ... }
}
```
## Notice
This PR won't get SEO metadata rendered asyncronizedly, instead it's a preparation for the later work in partial prerendering and async metadata. We'll encourage the Next.js community moving to the new metadata viewport API if you're customzing those 3 meta tags. Usually you don't have to change viewport itself, so mostly like only theme-color and color-scheme could relate to it.
This PR introduces a build optimization to create a "partial prerender" of the page.
1. During compilation, we create a static shell for the page using your existing Suspense boundaries. Components that can be static will be included in this static shell, leaving holes for the dynamic components.
1. Using `<Suspense />`, we can define fallbacks to be included in the partial prerender, as well as the holes for the dynamic components to stream into.
This means Next.js can initially serve a static loading skeleton, kicking off the dynamic parts in parallel. Then, the dynamic components stream in on demand. Dynamic components can use `cookies()`, `headers()`, `'cache': 'no-store'`, or `unstable_noStore()` to opt-into dynamic rendering.
Co-authored-by: Zack Tanner <1939140+ztanner@users.noreply.github.com>
This implements support for properly tracing sourcemaps when presenting
error stacks to the user. It also adds code frames when possible.
Closes WEB-1764
---------
Co-authored-by: kodiakhq[bot] <49736102+kodiakhq[bot]@users.noreply.github.com>
### What?
Note: This is not a breaking change, just removing some unused code.
### Why?
Since #56896 we don't need this, as Node.js 18+ has `fetch` exposed by default.
### How?
Depends on #56896, #56909
We already didn't load `fetch` if `globalThis` had it (ie. Node.js 18+ environments), and since we are dropping support for Node.js 16, these code paths should have no effect on runtime behavior.
In an attempt to avoid introducing experimental features into the Next.js userland space, we're reverting the `Promise.withResolvers` polyfill and preferring an internal `DetachedPromise` interface.
The PR supersedes the https://github.com/vercel/next.js/pull/53150, which is way too outdated, has way too many conflicts, and also heavily relies on GitHub Copilot (which makes the progress slow and tedious).
The PR uses [`json-schema-to-zod`](https://github.com/StefanTerdell/json-schema-to-zod) (instead of the GitHub Copilot) to generate the zod schema, and manually replaces all generated `z.customRefine` with my hand-written zod schema.
TODO:
- [x] Convert schema
- [x] Reduce `z.any()` usage
- [x] Create human-readable errors from the `ZodError`
- [x] Update test cases to reflect the latest error message
-----
The benefit of using zod over ajv:
- Easier maintenance: zod schema is straightforward to compose.
- Better typescript support: config schema now strictly reflects the `NextConfig` type.
- Smaller installation size: by replacing `ajv` and `@segment/ajv-human-errors` w/ `zod`, I am able to reduce installation size by 114 KiB.
- Better Extension: the zod error message is easy to customize.
-----
In the previous PR https://github.com/vercel/next.js/pull/56083, @feedthejim replaces `zod` w/ `superstruct`. `superstruct` is lightweight and fast, which makes it perfect for creating simple schemas for RSC payload. But, this also means `superstruct` has its limitations compared to `zod`:
- `superstruct`'s syntax is different, and some utilities's usage is counter-intuitive:
- `z.array(z.string()).gt(1)` vs `s.size(s.array(s.string()), 1)`
- `z.numer().gt(1)` vs `s.size(s.number(), 1)`, `s.min(s.number(), 1)`
- `z.boolean().optional().nullable()` vs `s.nullable(s.optional(z.boolean()))`
- `superstruct` has weaker TypeScript support and worse DX compared to `zod` when composing huge schema:
- `zod.ZodType + z.object()` can provide a more detailed type mismatch message on which specific property is the culprit, while `Describe + s.object()` provides almost no information at all.
- `zod`'s schema is more powerful
- `z.function()` supports `z.args()` and `z.returns()`, while `superstruct` only has `s.func()`
- zod also has Promise type `z.promise()` and intersection type `z.and()`
- `superstruct`'s error is harder to parse compared to `zod`'s `ZodError`
So in the PR, I re-introduced `zod` for `next.config.js` validation.
This adds types for the component modules so that the modules no longer need to rely on `ComponentMod` being `any`. This allows stricter typing for those elements.
This also improves types around template files to allow their exports to form the `ComponentMod` types.
Today when we hydrate an SSR'd RSC response on the client we encounter import chunks which initiate code loading for client components. However we only start fetching these chunks after hydration has begun which is necessarily after the initial chunks for the entrypoint have loaded.
React has upstream changes that need to land which will preinitialize the rendered chunks for all client components used during the SSR pass. This will cause a `<script async="" src... />` tag to be emitted in the head for each chunk we need to load during hydration which allows the browser to start fetching these resources even before the entrypoint has started to execute.
Additionally the implementation for webpack and turbopack is different enough that there will be a new `react-server-dom-turbopack` package in the React repo which should be used when using Turbopack with Next.
This PR also removes a number of patches to React src that proxy loading (`__next_chunk_load__`) and bundler requires (`__next_require__`) through the `globalThis` object. Now the react packages can be fully responsible for implementing chunk loading and all Next needs to do is supply the necessary information such as chunk prefix and crossOrigin attributes necessary for this loading. This information is produced as part of the client-manifest by either a Webpack plugin or Turbopack.
Additionally any modifications to the chunk filename that were previously done at runtime need to be made in the manifest itself now. This means we need to encode the deployment id for skew protection and encode the filename to make it match our static path matching (and resolutions on s3) when using `[` and `]` segment characters.
There are a few followup items to consider in later PRs
1. we currently bundle a node and edge version of react-server-dom-webpack/client. The node version has an implementation for busboy whereas the edge version does not. Next is currently configured to use busboy when handling a fetch action sent as multipart with a node runtime. Ideally we'd only bundle the one platform we are buliding for but some additional refactoring to support better forking is possibly required here
This PR also updates react from 09285d5a7 to d900fadbf.
### React upstream changes
- https://github.com/facebook/react/pull/27439
- https://github.com/facebook/react/pull/26763
- https://github.com/facebook/react/pull/27434
- https://github.com/facebook/react/pull/27433
- https://github.com/facebook/react/pull/27424
- https://github.com/facebook/react/pull/27428
- https://github.com/facebook/react/pull/27427
- https://github.com/facebook/react/pull/27315
- https://github.com/facebook/react/pull/27314
- https://github.com/facebook/react/pull/27400
- https://github.com/facebook/react/pull/27421
- https://github.com/facebook/react/pull/27419
- https://github.com/facebook/react/pull/27418
### What?
Deduplicates our env var injection between the JS and rust side
Closes WEB-937
---------
Co-authored-by: Tim Neutkens <tim@timneutkens.nl>
Co-authored-by: Zack Tanner <zacktanner@gmail.com>
**low-priority chore change**
## What?
This PR removes the 'beta.' prefix from links beginning with it. These links are no longer in beta and are now automatically redirected to their non-beta versions. The change serves as a minor enhancement.
### What
#51394 introduced a pretty strict type of return value of route type
that causing failure with `next build`.
There're few ways of writing a app route, it could contain few return
values based on the usage:
* return a `Response` or promise of it
* return `NextResponse` of promise of it, since it's extended from
`Response`, same type
* use `redirect()` or `notFound(), since it returns `never`, and the
below code is not reached, the handler itself could still return void.
e.g. using `redirect` in a `GET` route
We loosed the type so `redirect()` can be still allowed without
specifying the return value there.
Related typescript issue:
https://github.com/microsoft/TypeScript/issues/16608#issuecomment-309327984
### How
* Re-enable the bail on types / build error in the app-routes tests
* Separate the tests, move runtime erroring ones to
`test/e2e/app-dir/app-routes-errors`
* Add new case to app-routes tests of mixed return value
Closes#55623
Related #55604
In JS, Promise's are used to help manage async tasks and control flows. When code calls methods on a promise like `.then()`, `.catch()`, or `.finally()` the results of the promise are forwarded to the callback as soon as they're resolved. This serves to make a change to the promise creation such that we do not await on the promise until we're within the try/finally block. This will ensure that the promise will always be added to the map before it's resolved or rejected and it's cleanup (removing it from the active promises) is also completed.
This additionally introduces a new `scheduleOnNextTick` method and polyfill for `Promise.withResolvers()`.
`scheduleOnNextTick` is based on the scheduling algorithm used by https://github.com/graphql/dataloader which utilizes a `Promise.resolve()` combined with `process.nextTick` in order to schedule an operation to occur after the promises have resolved (see [graphql/dataloader](d336bd1528/src/index.js (L213-L255)))
The `Promise.withResolvers()` polyfill is an implementation of a soon-to-be-landed spec for inside-out promises. [Read the spec](https://tc39.es/proposal-promise-with-resolvers/)
## What
This PR changes Next.js to bundle its vendored React libraries so that the App Router pages can use those built-in versions.
## Why
Next.js supports both Pages and App Router and we've gone through a lot of iteration to make sure that Next.js stays flexible wrt to the version of React used: in Pages, we want to use the React provided by the user and in the App Router, to be able to use it, we need to use the canary version of React, which we've built into Next.js for convenience.
The problem stems from the fact that you can't run two different instances of React (by design).
Previously we have a dual worker setup, where we would separate completely each Next.js versions (App and Pages) so that they would not overlap with each other, however this approach was not great performance and memory wise.
We've recently tried using an ESM loader and a single process, but this change would still opt you into the React canary version if you had an app page, which breaks some assumptions.
## How
A list of the changes in this PR:
### New versions of the Next.js runtime
Since we now compile a runtime per type of page (app/route/api/pages), in order to bundle the two versions of React that we vendored, we introduced a new type of bundle suffixed by `-experimental`. This bundle will have the bleeding edge React needed for Server Actions and Next.js will opt you in into that runtime automatically.
For internal contributors, it means that we now run a compiler for 10 subparts of Next.js:
- next_bundle_server
- next_bundle_pages_prod
- next_bundle_pages_turbo
- next_bundle_pages_dev
- next_bundle_app_turbo_experimental
- next_bundle_app_prod
- next_bundle_app_prod_experimental
- next_bundle_app_turbo
- next_bundle_app_dev_experimental
- next_bundle_app_dev
### Simplified require-hook
Since the versions of React are correctly re-routed at build time for app pages, we don't need the require hook anymore
### Turbopack changes
The bundling logic in Turbopack has been addressed to properly follow the new logic
### Changes to the shared contexts system
Some context files need to have a shared instance between the rendering runtime and the user code, like the one that powers the `next/image` component. In general, the aliasing setup takes care of that but we need the require hook for code that is not compiled to reroute to the correct runtime. This only happens for pages node_modules.
A new Turbopack resolving plugin has been added to handle that logic in Turbopack.
### Misc changes
- `runtime-config` (that powers `next/config`) has been converted to an `.external` file, as it should have been
- there are some rules that have been added to the aliases to support the usage of `react-dom/server` in a server-components. We can do that now since the runtime takes care of separating the versions of React.
Co-authored-by: JJ Kasper <22380829+ijjk@users.noreply.github.com>
This bumps `undici` to the latest version to match the one used in newer Node.js versions', potentially fixing compatibility issues.
I also cleaned up some related `global as any` usage.
This PR removes our client module reference proxy implementation to directly use the one from the upstream Flight server, as it's added here: https://github.com/facebook/react/pull/27033.
Also updated the server reference creation code a bit to use `defineProperties` - we can't switch to the upstream `registerServerReference` API yet as our Server Actions compiler needs to change a bit to adapt that API since we might have existing `bound` and/or `originalAction` provided.
Closes#53732. Closes#52929.
When using the statically typed routes feature, we might have code like:
```ts
export function Card<T extends string>({ href }: { href: Route<T> | URL })...
export function Card<T extends string>({ href }: LinkProps<T>)...
```
To statically check `<Card href={...}>` and make sure it's `href` is an existing route. However, in certain cases these route types are not generated (e.g. running `tsc` directly w/o a `next dev` or `next build`), which results in TS errors.
This PR adds stub generics to `Route` and `LinkProps` so even if that plugin isn't executed, these types will not block type checking.
Reverts vercel/next.js#53578
This PR (#53578) will break client components test, revert it for now.
Can repro by adding `"use client"` to `app/page.tsx` in `test/production/jest/server-only.test.ts`
```
FAIL app/page.test.jsx
● Test suite failed to run·
Cannot find module 'private-next-rsc-mod-ref-proxy' from 'app/page.jsx'·
Require stack:
```
### 🧐 What's in there?
This is another attempt to allow testing server-only code with Jest.
### 🧪 How to test?
There's an integration tests which can be triggered with `pnpm testheadless server-only`
Here is a more comprehensive setup:
<details>
<summary><code>app/lib/index.ts</code></summary>
```ts
import 'server-only'
export function add(num1: number, num2: number) {
return num1 + num2
}
```
</details>
<details>
<summary><code>app/lib/index.test.ts</code></summary>
```ts
import { add } from '.'
it('adds two numbers', () => {
expect(add(1, 3)).toEqual(4)
})
```
</details>
<details>
<summary><code>app/client-component.tsx</code></summary>
```ts
'use client'
import { useState } from 'react'
export default function ClientComponent() {
const [text, setText] = useState('not clicked yet')
return <button onClick={() => setText('clicked!')}>{text}</button>
}
```
</details>
<details>
<summary><code>app/client-component.test.tsx</code></summary>
```ts
import { fireEvent, render, screen } from '@testing-library/react'
import ClientComponent from './client-component'
it('can be clicked', async () => {
render(<ClientComponent />)
const button = screen.getByRole('button')
expect(button).toHaveTextContent('not clicked yet')
await fireEvent.click(button)
expect(button).toHaveTextContent('clicked!')
})
```
</details>
<details>
<summary><code>app/server-component.tsx</code></summary>
```ts
import { add } from '@/lib'
export default function ServerComponent({ a, b }: { a: number; b: number }) {
return (
<code role="comment">
{a} + {b} = {add(a, b)}
</code>
)
}
```
</details>
<details>
<summary><code>app/server-component.test.tsx</code></summary>
```ts
import { render, screen } from '@testing-library/react'
import ServerComponent from './server-component'
it('renders', () => {
render(<ServerComponent a={2} b={3} />)
expect(screen.getByRole('comment')).toHaveTextContent('2 + 3 = 5')
})
```
</details>
<details>
<summary><code>app/page.tsx</code></summary>
```ts
import Link from 'next/link'
import ClientComponent from './client-component'
import ServerComponent from './server-component'
export default function Page() {
return (
<>
<h1>Hello World</h1>
<Link href="/dave">Dave?</Link>
<p>
<ClientComponent />
</p>
<p>
<ServerComponent a={5} b={2} />
</p>
</>
)
}
```
</details>
<details>
<summary><code>app/page.test.tsx</code></summary>
```ts
import { render, screen } from '@testing-library/react'
import Page from './page'
it('greets', () => {
render(<Page />)
expect(screen.getByRole('link')).toHaveTextContent('Dave?')
expect(screen.getByRole('heading')).toHaveTextContent('Hello World')
expect(screen.getByRole('button')).toHaveTextContent('not clicked yet')
expect(screen.getByRole('comment')).toHaveTextContent('5 + 2 = 7')
})
```
</details>
<details>
<summary><code>app/[blog]/page.tsx</code></summary>
```ts
import { Metadata } from 'next'
import Link from 'next/link'
type Props = {
params: { blog: string }
}
export async function generateMetadata({
params: { blog: title },
}: Props): Promise<Metadata> {
return { title, description: `A blog post about ${title}` }
}
export default function Page({ params }: Props) {
return (
<>
<div>
<Link href="/">Back</Link>
</div>
<h1>All about {params.blog}</h1>
</>
)
}
```
</details>
<details>
<summary><code>app/[blog]/page.test.tsx</code></summary>
```ts
import { render, screen } from '@testing-library/react'
import Page from './page'
it('has the appropriate title', () => {
const title = 'Jane'
render(<Page params={{ blog: title }} />)
expect(screen.getByRole('heading')).toHaveTextContent(`All about ${title}`)
expect(screen.getByRole('link')).toHaveTextContent('Back')
})
```
</details>
<details>
<summary><code>app/layout.tsx</code></summary>
```ts
export default function RootLayout({ children }) {
return (
<html lang="en">
<body>{children}</body>
</html>
)
}
```
</details>
<details>
<summary><code>jest.config.js</code></summary>
```ts
const nextJest = require('next/jest')
const createJestConfig = nextJest({ dir: './' })
module.exports = createJestConfig({
testEnvironment: 'jsdom',
setupFilesAfterEnv: ['<rootDir>/test-setup.ts'],
})
```
</details>
<details>
<summary><code>package.json</code></summary>
```ts
{
"name": "rsc-test",
"version": "0.0.0",
"private": true,
"scripts": {
"test": "jest"
},
"devDependencies": {
"@testing-library/jest-dom": "latest"
}
}
```
</details>
<details>
<summary><code>test-setup.ts</code></summary>
```ts
import '@testing-library/jest-dom'
```
</details>
The app should run and all test should pass.
### ❗ Notes to reviewers
#### The problem:
1. next/jest configures jest with a transformer ([jest-transformer](https://github.com/vercel/next.js/blob/canary/packages/next/src/build/swc/jest-transformer.ts)) to compile react code with next -swc
2. the transformers configures next -swc for a given environment: Server or Client, based on jest global environment
3. Based on the environment, next -swc checks for invalid usage of `import('server-only')` `“use client”`, `export const metadata` or `export async function generateMetadata`
4. Because the global test environment is either jsdom or node, the same test suite can not include both client and server components
#### Possible mitigations
*A. Using jest projects*
When configured with [multiple projects](https://jestjs.io/docs/next/configuration/#projects-arraystring--projectconfig), Jest can launch different runners with different environment. This would allow running server-only code in node and client-only code in jsdom.
However, it requires user to completely change their jest configuration. It would also require a different setup when scaffolding new app-directory project with create-next.
*B. Using doc blocks*
Jest allows changing the environment per test file [with docBlock](https://jestjs.io/docs/configuration#testenvironment-string).
However, by the time jest is invoking next -swc on a source file to transform it, this information is gone, and next -swc is still invoked with the (wrong) global environment.
The PR #52393 provides a workaround for files with `import('server-only')`, but does not allow testing pages with metadata.
*C. Always compile for node*
Our jest-transformer could always configure next -swc for server:
- pass Server-specific validations `import('server-only')` `export const metadata` or `export async function generateMetadata`
- does not complain about `"use client"`
This is what this PR is about!
Fixes#47299
Co-authored-by: Jiachi Liu <4800338+huozhi@users.noreply.github.com>
Followup on the https://github.com/vercel/next.js/pull/52520. The enhancements:
1. Both `fetch` and `http(s)` APIs are intercepted in the testmode.
2. The loopback mode that would allow the test-side to use any fetch-mocking library of their choice.
## Vendoring
Updates all module resolvers (node, webpack, nft for entrypoints, and nft for next-server) to consider whether vendored packages are suitable for a given resolve request and resolves them in an import semantics preserving way.
### Problem
Prior to the proposed change, vendoring has been accomplished but aliasing module requests from one specifier to a different specifier. For instance if we are using the built-in react packages for a build/runtime we might replace `require('react')` with `require('next/dist/compiled/react')`.
However this aliasing introduces a subtle bug. The React package has an export map that considers the condition `react-server` and when you require/import `'react'` the conditions should be considered and the underlying implementation of react may differ from one environment to another. In particular if you are resolving with the `react-server` condition you will be resolving the `react.shared-subset.js` implementation of React. This aliasing however breaks these semantics because it turns a bare specifier resolution of `react` with path `'.'` into a resolution with bare specifier `next` with path `'/dist/compiled/react'`. Module resolvers consider export maps of the package being imported from but in the case of `next` there is no consideration for the condition `react-server` and this resolution ends up pulling in the `index.js` implementation inside the React package by doing a simple path resolution to that package folder.
To work around this bug there is a prevalence of encoding the "right" resolution into the import itself. We for instance directly alias `react` to `next/dist/compiled/react/react.shared-subset.js` in certain cases. Other times we directly specify the runtime variant for instance `react-server-dom-webpack/server.edge` rather than `react-server-dom-wegbpack/server`, bypassing the export map altogether by selecting the runtime specific variant. However some code is meant to run in more than one runtime, for instance anything that is part of the client bundle which executes on the server during SSR and in the browser. There are workaround like using `require` conditionally or `import(...)` dynamically but these all have consequences for bundling and treeshaking and they still require careful consideration of the environment you are running in and which variant needs to load.
The result is that there is a large amount of manual pinning of aliases and additional complexity in the code and an inability to trust the package to specify the right resolution potentially causing conflicts in future versions as packages are updated.
It should be noted that aliasing is not in and of itself problematic when we are trying to implement a sort of lightweight forking based on build or runtime conditions. We have good examples of this for instance with the `next/head` package which within App Router should export a noop function. The problem is when we are trying to vendor an entire package and have the package behave semantically the same as if you had installed it yourself via node_modules
### Solution
The fix is seemingly straight forward. We need to stop aliasing these module specifiers and instead customize the resolution process to resolve from a location that will contain the desired vendored packages. We can then start simplifying our imports to use top level package resources and generally and let import conditions control the process of providing the right variant in the right context.
It should be said that vendoring is conditional. Currently we only vendor react pacakges for App Router runtimes. The implementation needs to be able to conditionally determine where a package resolves based on whether we're in an App Router context vs a Page Router one.
Additionally the implementation needs to support alternate packages such as supporting the experimental channel for React when using features that require this version.
### Implementation
The first step is to put the vendored packages inside a node_modules folder. This is essential to the correct resolving of packages by most tools that implement module resolution. For packages that are meant to be vendored, meaning whole package substitution we move the from `next/(src|dist)/compiled/...` to `next/(src|dist)/vendored/node_modules`. The purpose of this move is to clarify that vendored packages operate with a different implementation. This initial PR moves the react dependencies for App Router and `client-only` and `server-only` packages into this folder. In the future we can decide which other precompiled dependencies are best implemented as vendored packages and move them over.
It should be noted that because of our use of `JestWorker` we can get warnings for duplicate package names so we modify the vendored pacakges for react adding either `-vendored` or `-experimental-vendored` depending on which release channel the package came from. While this will require us to alter the request string for a module specifier it will still be treating the react package as the bare specifier and thus use the export map as required.
#### module resolvers
The next thing we need to do is have all systems that do module resolution implement an custom module resolution step. There are five different resolvers that need to be considered
##### node runtime
Updated the require-hook to resolve from the vendored directory without rewriting the request string to alter which package is identified in the bare specifier. For react packages we only do this vendoring if the `process.env.__NEXT_PRIVATE_PREBUNDLED_REACT` envvar is set indicating the runtime is server App Router builds. If we need a single node runtime to be able to conditionally resolve to both vendored and non vendored versions we will need to combine this with aliasing and encode whether the request is for the vendored version in the request string. Our current architecture does not require this though so we will just rely on the envvar for now
##### webpack runtime
Removed all aliases configured for react packages. Rely on the node-runtime to properly alias external react dependencies. Add a resolver plugin `NextAppResolverPlugin` to preempt perform resolution from the context of the vendored directory when encountering a vendored eligible package.
##### turbopack runtime
updated the aliasing rules for react packages to resolve from the vendored directory when in an App Router context. This implementation is all essentially config b/c the capability of doing the resolve from any position (i.e. the vendored directory) already exists
##### nft entrypoints runtime
track chunks to trace for App Router separate from Pages Router. For the trace for App Router chunks use a custom resolve hook in nft which performs the resolution from the vendored directory when appropriate.
##### nft next-server runtime
The current implementation for next-server traces both node_modules and vendored version of packages so all versions are included. This is necessary because the next server can run in either context (App vs Page router) and may depend on any possible variants. We could in theory make two traces rather than a combined one but this will require additional downstream changes so for now it is the most conservative thing to do and is correct
Once we have the correct resolution semantics for all resolvers we can start to remove instances targeting our precompiled instances for instance making `import ... from "next/dist/compiled/react-server-dom-webpack/client"` and replacing with `import ... from "react-server-dom-webpack/client"`
We can also stop requiring runtime specific variants like `import ... from "react-server-dom-webpack/client.edge"` replacing it with the generic export `"react-server-dom-webpack/client"`
There are still two special case aliases related to react
1. In profiling mode (browser only) we rewrite `react-dom` to `react-dom/profiling` and `scheduler/tracing` to `scheduler/tracing-profiling`. This can be moved to using export maps and conditions once react publishses updates that implement this on the package side.
2. When resolving `react-dom` on the server we rewrite this to `react-dom/server-rendering-stub`. This is to avoid loading the entire react-dom client bundle on the server when most of it goes unused. In the next major react will update this top level export to only contain the parts that are usable in any runtime and this alias can be dropped entirely
There are two non-react packages currently be vendored that I have maintained but think we ought to discuss the validity of vendoring. The `client-only` and `server-only` packages are vendored so you can use them without having to remember to install them into your project. This is convenient but does perhaps become surprising if you don't realize what is happening. We should consider not doing this but we can make that decision in another discussion/PR.
#### Webpack Layers
One of the things our webpack config implements for App Router is layers which allow us to have separate instances of packages for the server components graph and the client (ssr) graph. The way we were managing layer selection was a but arbitrary so in addition to the other webpack changes the way you cause a file to always end up in a specific layer is to end it with `.serverlayer`, `.clientlayer` or `.sharedlayer`. These act as layer portals so something in the server layer can import `foo.clientlayer` and that module will in fact be bundled in the client layer.
#### Packaging Changes
Most package managers are fine with this resolution redirect however yarn berry (yarn 2+ with PnP) will not resolve packages that are not defined in a package.json as a dependency. This was not a problem with the prior strategy because it was never resolving these vendored packages it was always resolving the next package and then just pointing to a file within it that happened to be from react or a related package.
To get around this issue vendored packages are both committed in src and packed as a tgz file. Then in the next package.json we define these vendored packages as `optionalDependency` pointing to these tarballs. For yarn PnP these packed versions will get used and resolved rather than the locally commited src files. For other package managers the optional dependencies may or may not get installed but the resolution will still resolve to the checked in src files. This isn't a particularly satisfying implemenation and if pnpm were to be updated to have consistent behavior installing from tarballs we could actually move the vendoring entirely to dependencies and simplify our resolvers a fair bit. But this will require an upstream chagne in pnpm and would take time to propogate in the community since many use older versions
#### Upstream Changes
As part of this work I landed some other changes upstream that were necessary. One was to make our packing use `npm` to match our publishing step. This also allows us to pack `node_modules` folders which is normally not supported but is possible if you define the folder in the package.json's files property.
See: #52563
Additionally nft did not provide a way to use the internal resolver if you were going to use the resolve hook so that is now exposed
See: https://github.com/vercel/nft/pull/354
#### additional PR trivia
* When we prepare to make an isolated next install for integration tests we exclude node_modules by default so we have a special case to allow `/vendored/node_modules`
* The webpack module rules were refactored to be a little easier to reason about and while they do work as is it would be better for some of them to be wrapped in a `oneOf` rule however there is a bug in our css loader implementation that causes these oneOf rules to get deleted. We should fix this up in a followup to make the rules a little more robuts.
## Edits
* I removed `.sharedlayer` since this concept is leaky (not really related to the client/server boundary split) and it is getting refactored anyway soon into a precompiled runtime.
### What?
Fix documentation about `fetch` polyfilling, and drop `node-fetch` references.
### Why?
Since we stopped using `node-fetch` in `next` in favor of `undici` there were some inconsistencies in the docs, and unused references to the `node-fetch` package inside `packages/next`-
Noted this while answering this [Slack thread](https://vercel.slack.com/archives/C03S8ED1DKM/p1691089801007129)
### 🧐 What's in there?
At the moment, it is not possible to test code with `import 'server-only'` in app directory.
When trying to load such file in jest (even with `testEnvironment: node`), the error will be:
```
● Test suite failed to run··
x NEXT_RSC_ERR_CLIENT_IMPORT: server-only
,-[lib/util.js:1:1]
1 | /** @jest-environment node */·
2 | import 'server-only'
: ^^^^^^^^^^^^^^^^^^^^
3 | export const PI = 3.14;
`----·
at Object.transformSync (node_modules/next/src/build/swc/index.ts:443:25)
at transformSync (node_modules/next/src/build/swc/index.ts:629:19)
at Object.process (node_modules/next/src/build/swc/jest-transformer.ts:117:25)
at ScriptTransformer.transformSource (node_modules/@jest/transform/build/ScriptTransformer.js:619:31)
at ScriptTransformer._transformAndBuildScript (node_modules/@jest/transform/build/ScriptTransformer.js:765:40)
at ScriptTransformer.transform (node_modules/@jest/transform/build/ScriptTransformer.js:822:19)·
```
In a nutshell:
- next/swc is looking for ‘server-only’ [text in the source](https://github.com/vercel/next.js/blob/canary/packages/next-swc/crates/core/src/react_server_components.rs#L576), and throw if not configured for server
- next's jest-transformer will only configure next/swc for server [if the environment is node](https://github.com/vercel/next.js/blob/canary/packages/next/src/build/swc/jest-transformer.ts#L88)
- when testing Next.js apps, your jest testEnvironment is most likely jsdom. But you can configure it [per file with docBlock](https://jestjs.io/docs/configuration#testenvironment-string), which jest-transformer ignores because it only reads the top-level configuration.
This PR fixes this, by
1. reading the docblock to configure next/swc accordingly and bypass its hardcoded guard
2. mocking `server-only` so [it does not throw](https://github.com/vercel/next.js/blob/canary/packages/next/src/compiled/server-only/index.js) when loaded (jest does not read `react-server` export from package.json)
Users would still have to annotate their `server-only` files with `/** @jest-environment node */` in order to test them.
### 🧪 How to test?
There's a full test available: `pnpm testheadless --testPathPattern jest/server-only`
Here is also a minimal reproduction:
<details>
<summary>app/layout.tsx</summary>
```typescript
export default function RootLayout({ children }: { children: React.ReactNode }) {
return (<html lang="en"><body>{children}</body></html>)
}
```
</details>
<details>
<summary>app/page.tsx</summary>
```typescript
import { PI } from '@/lib/utils'
export default function Home() {
return <h1>{PI}</h1>
}
```
</details>
<details>
<summary>lib/utils.ts</summary>
```typescript
import 'server-only'
export const PI = 3.14
```
</details>
<details>
<summary>lib/utils.test.ts</summary>
```typescript
import { PI } from './utils'
it('works', () => expect(PI).toEqual(3.14))
```
</details>
<details>
<summary>jest.config.js</summary>
```typescript
const nextJest = require('next/jest')
module.exports = nextJest({ dir: './' })({ testEnvironment: 'jsdom' })
```
</details>
### ❗ Notes to reviewers
[jest-docblock](https://packagephobia.com/result?p=jest-docblock) with dependencies is only 12.5 kB.
Fixes#47448
### What?
Node.js sets stdout to non-blocking by default, rust expects it to be blocking
so when logging a lot of data, it can happen that rust receives an error while writing causing it to panic, the error is just that the resource is temporarily unavailable, but rust doesn't handle this case
See https://github.com/napi-rs/napi-rs/issues/1630
## What?
While investigating slow compilation for a page on vercel.com in
development I found that there was close to 10 seconds of time
unaccounted for in `.next/trace`. Ran a profile and found that time was
spent in watchpack `batch`, specifically calling `close` many times.
When I tried to debug this further by running unbundled webpack I
noticed the same issue didn't exists.
### Before
<img width="1329" alt="Before"
src="https://github.com/vercel/next.js/assets/6324199/9ace4628-db04-4de7-993f-65aef9dffc55">
### After
<img width="1278" alt="After"
src="https://github.com/vercel/next.js/assets/6324199/55d5e58b-4a27-4235-8dea-723a7a78c117">
## Raw numbers
<table>
<tr>
<td>Before</td>
<td>After</td>
<td>Delta</td>
<td>Delta (percent)</td>
</tr>
<tr>
<td>13840 ms</td>
<td>3580 ms</td>
<td>-10260 ms</td>
<td>-74.13%</td>
</tr>
</table>
## How?
Investigated further and found that specifically not minifying watchpack
solved the issue.
<!-- Thanks for opening a PR! Your contribution is much appreciated.
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(s) that you're making:
## For Contributors
### Improving Documentation or adding/fixing Examples
- The "examples guidelines" are followed from our contributing doc
https://github.com/vercel/next.js/blob/canary/contributing/examples/adding-examples.md
- Make sure the linting passes by running `pnpm build && pnpm lint`. See
https://github.com/vercel/next.js/blob/canary/contributing/repository/linting.md
### Fixing a bug
- Related issues linked using `fixes #number`
- Tests added. See:
https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
### Adding a feature
- Implements an existing feature request or RFC. Make sure the feature
request has been accepted for implementation before opening a PR. (A
discussion must be opened, see
https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added
(https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see
https://github.com/vercel/next.js/blob/canary/contributing.md
## For Maintainers
- Minimal description (aim for explaining to someone not on the team to
understand the PR)
- When linking to a Slack thread, you might want to share details of the
conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic
behind a change
### What?
### Why?
### How?
Closes NEXT-
Fixes #
-->
This `draftMode` prop was missing from the previous PR:
- https://github.com/vercel/next.js/pull/48669
It was added on `GetStaticPropsContext` but not on
`GetServerSidePropsContext` so this PR fixes it.
---------
This PR does three things:
- Vendors the package `react-server-dom-webpack@experimental` as
`react-server-dom-webpack-experimental` similar to React and React DOM
- Upgrades all vendored React packages
- Re-lands the `experimentalReact` flag to switch between `@next` and
`@experimental` channels of React for app dir.
Fix NEXT-926.
Mostly mirrors the changed made in
https://github.com/facebook/react/pull/26632 to our SWC transform. The
implementation difference is that the AST transformer only adds a
general purpose wrapper call `createServerReference(id)` from an aliased
import, so we can easily change the underlying function in the bundler.
This change only affects the client layer (when `self.config.is_server
=== false`).
Needs to be landed after another React upgrade:
https://github.com/vercel/next.js/pull/48697.
cc @sebmarkbage.
This PR implements preloading of CSS from RSC.
1. The underlying Flight protocol was extended in
https://github.com/facebook/react/pull/26502 to allow sending hints from
RSC to SSR and Client runtimes. React was updated to include these
changes.
2. We now call `ReactDOM.preload()` for each stylesheet used in a
layout/page layer
There are a few implementation details to take note of
1. we were previously using the `.browser` variant of a few React
packages. This was a holdover from when there was just browser and node
and we wanted to use the browser variant b/c we wanted the same code to
work in edge/node runtimes. React now publishes a `.edge` variant which
is like `.browser` but expects to be server only. This is necessary to
get the opt-in for `AsyncLocalStorage`.
2. Even with the above change, AsyncLocalStorage was not patched on the
global scope until after React was loaded. I moved this into a module
which is loaded first
3. The component passed to RSC's `renderToReadableStream` is not
actually part of the RSC module graph. If I tried to call
`ReactDOM.preload()` inside that function or any other function defined
inside `app-render.tsx` file it would actually see the wrong instance of
`react-dom`. I added a new export on the RSC top level export which
exposes a `preloadStyle(...)` function which just delegates to
`ReactDOM.preload(...)`. This makes the preload run in the right module
graph
~There are a couple of bugs in React that this work uncovered that I
will upstream. We may want to delay merging until they are addressed.
I'll update this comment when that is complete.~
1. ~React, during SSR, can emit a preload for a style twice in some
circumstances because late discovered stylesheets don't consider whether
a preload has already been flushed when preparing to reveal a boundary
they are within~
2. ~React, during RSC updates on the client, can preload a style that is
already in the document because it currently only looks for existing
preload links and does not consider if there is a stylesheet link with
the same href.~
~both of these issues will not break functionality, they just make the
network tab look at bit more noisy. We would expect network deduping to
prevent multiple actual loads~
The above React bugs were fixed and included now in the React update in
this PR
---------
Co-authored-by: Shu Ding <g@shud.in>
Draft Mode is very similar to Preview Mode but doesn't include any
additional data.
This PR implements support for Draft Mode in `pages` and a future PR
will implement support in `app`.
fix NEXT-992
Reverts vercel/next.js#48038
fix NEXT-926
---
The root cause was that when copying the package.json, I removed all
fields except for a few (such as `exports`) but missed the `browser`
field. That caused the client bundle to resolve to the Node.js version
of React DOM, and then we had the `async_hooks` error. Added it back in
99c9b9e51f8b0d4e4503ece9d07bce09161f3341.
I reproduced the error with next-site earlier and confirmed that this
fix is good.
Next.js includes various feature sets that depend on specific release
channels of React. However, our current setup only includes the `next`
channel of React, which restricts our ability to integrate with features
available on the `experimental` channel.
To address this limitation, this pull request introduces the following
changes:
- Vendors the `react@experimental` version, along with the corresponding
`react-dom` and `scheduler` packages.
- Modifies the `sync-react` script to also update the `experimental`
channel and removes `--version` as they're always synced to the latest
now.
- Retains the default behavior of using the `next` channel in the
`appDir` directory.
- Adds an option to switch to the `experimental` channel by setting
`experimental.experimentalReact: true` in the configuration.
fix NEXT-926 ([link](https://linear.app/vercel/issue/NEXT-926))
