In #29010, we started throwing an error if the res was mutated after
getServerSideProps() returned. This was to support classic streaming,
where it would be possible to accidentally mutate the response headers
after they were already sent. However, this change also caught [a few
non-streaming cases](https://github.com/vercel/next.js/pull/29010#issuecomment-943482743) that we don't want to break.
As such, with this change, we only throw the error if res is mutated
after gSSP returns *and* you've opted into using classic streaming.
Otherwise, we will only add a warning to the console.
This PR adds support for [Middleware as per RFC ](https://github.com/vercel/next.js/discussions/29750).
## 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
This refactor is the first of a few changes to support "classic" (two-part)
streaming. This one should be a noop that doesn't actually change the behavior.
It re-organizes the way that functions are wrapped in Document Head/NextScript
so anything that will be part of the second flush can be separated out from the
first flush. It also adds the structure for a useMaybeDeferContent hook, but
currently always assumes that nothing should be deferred.
The next PRs will actually implement streaming.
* Throw error if res is accessed after gSSP returns
Currently it's possible to access the `ServerResponse` through `context.res`
in `getServerSideProps()`. If one was to store that response and mutate
its headers or status code after gSSP returns (e.g. during rendering), it
would currently happen to work because of when headers are sent. However,
this is an anti-pattern that relies an implementation detail of the framework
and shouldn't be allowed. This will be particularly important once Next.js
starts to support basic streaming (two-part flush: routing then data) because
then the headers will be sent as soon as gSSP returns, which explicitly breaks
this pattern.
With this commit, the framework now throws an error in development mode if
the ServerResponse is accessed after gSSP returns.
* fixup! Throw error if res is accessed after gSSP returns
Use `Writable` instead of `Observable` and remove the `zen-observable` dependencies. I initially opted to use `Observable` for simplicity and fast iteration, but we should really just use `Writable` directly (or some other stream in the future).
React's streaming SSR has some [specific requirements](https://github.com/reactwg/react-18/discussions/66#discussioncomment-944266) on the stream API. Rather than trying to also squeeze a `Readable` in here, which might be more standard for node apps, I've just followed React's lead. By limiting ourselves to just `Writable`, it ought to be easier to adopt a different stream type in the future if desired.
The React `pipeToNodeWritable` API requires us to pass a stream immediately, but we don't actually have a `ServerResponse` to give it until `RenderResult.pipe(...)` is called later. For that reason, we pass React a `Writable` that we will simply forward to `res` later. This mechanism of deferring is `NodeWritablePiper`, which is just a function that can be called with `ServerResponse` (or another `Writable`, as we now do to render to string for static results) to have content written to it. `NodeWritablePiper` takes a `next` argument so that we can chain both synchronous and asynchronous pipers together.
Also does some clean up and adds another streaming test for backpressure.
We're no longer currently planning on supporting caching for dynamic responses, so we can do some cleaning & simplification:
* Multiplexing can be removed since we only ever subscribe once (via `RenderResult.pipe`, described below)
* `RenderResult.toUnchunkedString` can become synchronous since static responses are never chunked
* `RenderResult.forEach` can become `RenderResult.pipe` which helps encapsulate some of the details of `RenderResult`
Previous to this change, getServerSideProps could only return plain objects
for props, e.g.:
```javascript
export async function getServerSideProps() {
return {
props: {
text: 'some value',
}
}
}
```
With this commit, the props object can also be a Promise, e.g.
```javascript
export async function getServerSideProps() {
return {
props: (async function () {
return {
text: 'promise value',
}
})(),
}
}
```
For now, the framework simply waits for the results of the props Promise to resolve,
but this small change sets the groundwork for later allowing props to be streamed (cc @devknoll).
## Feature
- [x] Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR. -- *This is part of @devknoll's ongoing work to support streaming.*
- [ ] 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. *not sure if this applies here*
- [ ] Errors have helpful link attached, see `contributing.md`
Our `Observable` use has gotten sufficiently complex that it makes sense to just use a 3rd party implementation and not worry about maintaining it ourselves. As a bonus, it doesn't rely on Node APIs.
This fixes revalidation not occurring correctly when `notFound: true` is returned during build, additional tests have been added to ensure this is working correctly for dynamic and non-dynamic pages returning `notFound: true` during build and then revalidating afterwards.
## Bug
- [x] Related issues linked using `fixes #number`
- [x] Integration tests added
- [x] Errors have helpful link attached, see `contributing.md`
Fixes: https://github.com/vercel/next.js/issues/21453
Implements `renderToString` in terms of a new `renderToStream`. The former is used for legacy documents that generate the body HTML as part of `getInitialProps`. The latter will be used directly in #27794 when streaming dynamic HTML.
Since we're exposing an actual streaming response for dynamic HTML (instead of buffering with `resultFromChunks`), we use `multiplexResult` to buffer and multiplex the underlying result to multiple subscribers.
We generate the HTML for a document in two steps: First, we generate the body (i.e. everything under `<div id="__next">`). Then we generate the rest of the document and embed the body in it.
This doesn't work when the body is a stream, because React can't render the body for us unless we buffer it, and buffering it means not streaming. This PR takes the existing approach for AMP and uses it for all scenarios: instead of rendering HTML, we just render a placeholder that we can replace with HTML later. This will be used in a follow-up PR to let us know where to concatenate the body stream.
I also used the opportunity to split out `HtmlContext` from `DocumentProps`, as these will not be the same thing with functional document components.
Allows opting in to support for new concurrent features, like server-side Suspense.
**!!! DO NOT USE !!!**
This is highly experimental. We **will** be gating additional breaking changes behind this same flag.
**!!! DO NOT USE !!!**
Also resolves suspense for static pages (i.e. `getStaticProps` or `next build`/`next export`) since we can't currently support streaming for those cases anyway.
Adds `RenderResult`, replacing the `string` that `renderToHTML` used to return, with an `Observable`-like API that callers can use to subscribe and get a callback when chunks are available to flush, etc.
This is the last architectural change needed for streaming. There are, however, other things currently standing in the way of streaming. For example, it is common to mutate `res` in `getServerSideProps` to do routing work, or write headers, before fetching page data. This pattern effectively nullifies any advantages of streaming. I may do a follow-up PR that adds an experimental alternative for applications not using React 18, but the main purpose for this support is for Suspense and Server Components.
For that reason, there's no actual streaming here yet: instead we just flush a single chunk. A follow-up PR will add support for streaming suspense boundaries in React 18.
This PR fixes#25927 bug which prevents to use `assetPrefix` config with the experimental critical CSS feature.
Fixes: #25927
## Bug
- [x] Related issues linked using `fixes #number`
- [ ] Integration tests added
- [ ] Errors have helpful link attached, see `contributing.md`
From https://github.com/vercel/next.js/pull/20628, when the page is rendered server-side, `Router`'s `isReady` field needs to be initially set to `true`. However, when `_app` has custom `getInitialProps`, it seems that it is not the case, even though the page is rendered on the server.
This leads to a bug that `Router.isReady` is never set to `true`.
This pull request fixes the problem by fixing the initial calculation logic of `isReady` of `Router`.
## 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
- [x] Make sure the linting passes
Currently there is a lot of mutation in the Next.js Server and the checks for Locale are directly coded in the general request handler. Ideally, we should have a function where we just pass the request input (url + headers + config) and generate a bunch of metadata that analyzes it generating all metadata we might require for both the URL and i18n + basePath information.
This PR brings:
- A new parsing function `parseUrl` that joins parsing an absolute/relative URL into a data structure compatible with the Node parsing output but missing redundant properties.
- A wrapper `parseNextURL` that extends `parseUrl` analyzing `i18n` and `basePath` based on the provided configuration, url and headers. This function is pure and stateless so it can be used outside of the Next.js context.
- Types improvements and reuse.
- Refactors `next-server.ts` request handling using the above mentioned functions so that the code there just apply effects to the `req` object and the `parsedUrl.query` leaving the code much more straightforward.
- Refactors `getRouteRegex` decomposing in two different functions where `getParametrizedRoute` can be used to retrieve the serializable data that is used to generate the Regex.
How a page is rendered depends on whether or not we're streaming. For example, if we're just rendering to a `string` or we're generating a response for a crawler or other robot, we don't want React 18 to dynamically flush `<script>` tags to update Suspense boundaries as they resolve. Instead, we just want to wait for the full HTML to resolve and return a result similar to `renderToString`.
This is what `RequestContext` and the new/refactored `pipe` and `getStaticHTML` methods allow. They add a `requireStaticHTML` option that gets passed down. A follow-up PR will make sure this is `true` when serving a robot, and also ensure React is invoked appropriately.