This is a follow-up from #30528 to not only handle props (the critical
change), but also the owner ~and stack~ of a referenced element.
~Handling stacks here is rather academic because the Flight Server
currently does not deduplicate owner stacks. And if they are really
identical, we should probably just dedupe the whole element.~ EDIT:
Removed from the PR.
Handling owner objects on the other hand is an actual requirement as
reported in https://github.com/vercel/next.js/issues/69545. This problem
only affects the stable release channel, as the absence of owner stacks
allows for the specific kind of shared owner deduping as demonstrated in
the unit test.
This means that the owner of a Component rendered on the remote server
becomes the Component on this server.
Ideally we'd support this for the Client side too. In particular Fiber
but currently ReactComponentInfo's owner is typed as only supporting
other ReactComponentInfo and it's a bigger lift to support that.
This is important if the lazy is at the root of the chunk. I don't have
a unit test for it but @gnoff has a repro.
It also shouldn't unwrap the last value since that's the one we're
referencing.
This was already done correctly by @unstubbable in waitForReference so
this just aligns with that.
Any time we're creating a stack trace we should have a
react-stack-bottom-frame so we know what to filter out.
This is the same thing we already do for createFakeJSXCallStackInDEV but
we should do that when replaying logs too.
Stacked on #30906.
Injects the Flight Client into the DevTools hook if it `supportsFlight`.
This only injects in DEV. We could inject it in prod too but so far the
only feature this exposes is only available in DEV anyway. I also only
call `injectIntoDevTools` in the browser builds since we don't really
support DevTools on the server anyway.
The main purpose of this for now is so that DevTools can track the
Server Component owner of replayed logs. This lets us add owner stacks
where `console.createTask` is not natively supported (like Firefox). It
also lets us associate the log with the Server Component in the
Component tree #30905.
stacked on: #30731
We've refined the model of halting a prerender. Now when you abort
during a prerender we simply omit the rows that would complete the
flight render. This is analagous to prerendering in Fizz where you must
resume the prerender to actually result in errors propagating in the
postponed holes. We don't have a resume yet for flight and it's not
entirely clear how that will work however the key insight here is that
deciding whether the never resolving rows are an error or not should
really be done on the consuming side rather than in the producer.
This PR also reintroduces the logs for the abort error/postpone when
prerendering which will give you some indication that something wasn't
finished when the prerender was aborted.
Stacked on #30731.
When logging a Promise we emit it as an infinite promise instead of
blocking the replay on it.
This models that as a halted row instead. No need for this special case.
I unflag the receiving side since now it's used to replace a feature
that's already unflagged so it's used.
When printing these in DevTools they show up as the name of the
constructor so then you pass a Promise to the client it logs as "Chunk"
which is confusing.
Ideally we'd probably just name this Promise but 1) there's a slight
difference in the .then method atm 2) it's a bit tricky to name a
variable and get it from the global in the same scope. Closure compiler
doesn't let us just name a function because it removes it and just uses
the variable name.
using infinitely suspending promises isn't right because this will parse
as a promise which is only appropriate if the value we're halting at is
a promise. Instead we need to have a special marker type that says this
reference will never resolve. Additionally flight client needs to not
error any halted references when the stream closes because they will
otherwise appear as an error
addresses:
https://github.com/facebook/react/pull/30705#discussion_r1720479974
This uses a similar technique to what we use to generate fake stack
frames for server components. This generates an eval:ed wrapper function
around the Server Reference proxy we create on the client. This wrapper
function gets the original `name` of the action on the server and I also
add a source map if `findSourceMapURL` is defined that points back to
the source of the server function.
For `"use server"` on the server, there's no new API. It just uses the
callsite of `registerServerReference()` on the Server. We can infer the
function name from the actual function on the server and we already have
the `findSourceMapURL` on the client receiving it.
For `"use server"` imported from the client, there's two new options
added to `createServerReference()` (in addition to the optional
[`encodeFormAction`](#27563)). These are only used in DEV mode. The
[`findSourceMapURL`](#29708) option is the same one added in #29708. We
need to pass this these references aren't created in the context of any
specific request but globally. The other weird thing about this case is
that this is actually a case where the compiled environment is the
client so any source maps are the same as for the client layer, so the
environment name here is just `"Client"`.
```diff
createServerReference(
id: string,
callServer: CallServerCallback,
encodeFormAction?: EncodeFormActionCallback,
+ findSourceMapURL?: FindSourceMapURLCallback, // DEV-only
+ functionName?: string, // DEV-only
)
```
The key is that we use the location of the
`registerServerReference()`/`createServerReference()` call as the
location of the function. A compiler can either emit those at the same
locations as the original functions or use source maps to have those
segments refer to the original location of the function (or in the case
of a re-export the original location of the re-export is also a fine
approximate). The compiled output must call these directly without a
wrapper function because the wrapper adds a stack frame. I decided
against complicated and fragile dev-only options to skip n number of
frames that would just end up in prod code. The implementation just
skips one frame - our own. Otherwise it'll just point all source mapping
to the wrapper.
We don't have a `"use server"` imported from the client implementation
in the reference implementation/fixture so it's a bit tricky to test
that. In the case of CJS on the server, we just use a runtime instead of
compiler so it's tricky to source map those appropriately. We can
implement it for ESM on the server which is the main thing we're testing
in the fixture. It's easier in a real implementation where all the
compilation is just one pass. It's a little tricky since we have to
parse and append to other source maps but I'd like to do that as a
follow up. Or maybe that's just an exercise for the reader.
You can right click an action and click "Go to Definition".
<img width="1323" alt="Screenshot 2024-08-17 at 6 04 27 PM"
src="https://github.com/user-attachments/assets/94d379b3-8871-4671-a20d-cbf9cfbc2c6e">
For now they simply don't point to the right place but you can still
jump to the right file in the fixture:
<img width="1512" alt="Screenshot 2024-08-17 at 5 58 40 PM"
src="https://github.com/user-attachments/assets/1ea5d665-e25a-44ca-9515-481dd3c5c2fe">
In Firefox/Safari given that the location doesn't exist in the source
map yet, the browser refuses to open the file. Where as Chrome does
nearest (last) line.
There's a special case that happens when we replay logs on the client
because this doesn't happen within the context of any particular
rendered component. So we need to reimplement things that would normally
be handled by a full client like Fiber.
The implementation of `getOwnerStackByComponentInfoInDev` is the
simplest version since it doesn't have any client components in it so I
move it to `shared/`. It's only used by Flight but both `react-server/`
and `react-client/` packages. The ReactComponentInfo type is also more
generic than just Flight anyway.
In a follow up I still need to implement this in React DevTools when
native tasks are not available so that it appends it to the console.
This is a major nit but this avoids an extra stack frame when we're
replaying logs.
Normally the `printToConsole` frame doesn't show up because it'd be
ignore listed.
<img width="421" alt="Screenshot 2024-07-25 at 11 49 39 AM"
src="https://github.com/user-attachments/assets/81334c2f-e19e-476a-871e-c4db9dee294e">
When you expand to show ignore listed frames a ton of other frames show
up.
<img width="516" alt="Screenshot 2024-07-25 at 11 49 47 AM"
src="https://github.com/user-attachments/assets/2ab8bdfb-464c-408d-9176-ee2fabc114b6">
The annoying thing about this frame is that it's at the top of the stack
where as typically framework stuff ends up at the bottom and something
you can ignore. The user space stack comes first.
With this fix there's no longer any `printToConsole` frame.
<img width="590" alt="Screenshot 2024-07-25 at 12 09 09 PM"
src="https://github.com/user-attachments/assets/b8365d53-31f3-43df-abce-172d608d3c9c">
Am I wiling to eat the added complexity and slightly slower performance
for this nit? Definitely.
This enables configuring the name of the requested environment.
When we currently use createTask, we start with a `"use server"`
annotation. This option basically configures that string.
I now also deal with the case when switching environments along the
owner path. If you go from `"Third Party"` to `"Server"` to `"Client"`,
it'll have a task named `"use third party"` at the root, then `"use
server"` and then finally `"use client"`.
We don't really have the concept of a Server Component making a request
during render to then create another Server Component. Really the inner
one should conceptually have the first one as its owner in that case. So
currently the inner one will always have a null owner. We could somehow
connect them in this server-to-server case.
We don't currently have a way to configure the `"use client"` option but
I figured maybe that could be inferred by the server environment that
the Flight Client is executed within.
Note: We did talk before about annotating each stack frame with the
environment. You can effectively do that manually when parsing
`rsc://React/{environment}/` from `captureOwnerStack`. However, we can't
do that natively. At least not without deeper integration. Because it's
the source map that's responsible for the actual function name of each
stack frame - not what we give it at runtime. So for the native stacks,
the task showing the change in environment is more practical.
This way you can use the environment to know where to look for the
source map in case you have multiple server environments.
This becomes part of the public protocol since it's part of what you'll
parse out of the `rsc://React/` prefixed URLs inside of
`captureOwnerStack`.
When a model references a deduped object of a blocked element that has
subsequently been turned into a lazy element, we need to wait for the
lazy element's chunk to resolve before resolving the reference.
Without the fix, the new test failed with the following runtime error:
```
TypeError: Cannot read properties of undefined (reading 'children')
1003 | let value = chunk.value;
1004 | for (let i = 1; i < path.length; i++) {
> 1005 | value = value[path[i]];
| ^
1006 | }
1007 | const chunkValue = map(response, value);
1008 | if (__DEV__ && chunk._debugInfo) {
at getOutlinedModel (packages/react-client/src/ReactFlightClient.js:1005:26)
```
The bug was uncovered after updating React in Next.js in
https://github.com/vercel/next.js/pull/66711.
Stacked on #30401.
Previously we were transferring the original V8 stack trace string to
the client and then parsing it there. However, really the server is the
one that knows what format it is and it should be able to vary by server
environment.
We also don't use the raw string anymore (at least not in
enableOwnerStacks). We always create the native Error stacks.
The string also made it unclear which environment it is and it was
tempting to just use it as is.
Instead I parse it on the server and make it a structured stack in the
transfer format. It also makes it clear that it needs to be formatted in
the current environment before presented.
Stacked on https://github.com/facebook/react/pull/30400 and
https://github.com/facebook/react/pull/30369
Previously we were using fake evals to recreate a stack for console
replaying and thrown errors. However, for owner stacks we just used the
raw string that came from the server.
This means that the format of the owner stack could include different
formats. Like Spidermonkey format for the client components and V8 for
the server components. This means that this stack can't be parsed
natively by the browser like when printing them as error like in
https://github.com/facebook/react/pull/30289. Additionally, since
there's no source file registered with that name and no source mapping
url, it can't be source mapped.
Before:
<img width="1329" alt="before-firefox"
src="https://github.com/user-attachments/assets/cbe03f9c-96ac-48fb-b58f-f3a224a774f4">
Instead, we need to create a fake stack like we do for the other things.
That way when it's printed as an Error it gets source mapped. It also
means that the format is consistently in the native format of the
current browser.
After:
<img width="753" alt="after-firefox"
src="https://github.com/user-attachments/assets/b436f1f5-ca37-4203-b29f-df9828c9fad3">
So this is nice because you can just take the result from
`captureOwnerStack()` and append it to an `Error` stack and print it
natively. E.g. this is what React DevTools will do.
If you want to parse and present it yourself though it's a bit awkward
though. The `captureOwnerStack()` API now includes a bunch of
`rsc://React/` URLs. These don't really have any direct connection to
the source map. Only the browser knows this connection from the eval.
You basically have to strip the prefix and then manually pass the
remainder to your own `findSourceMapURL`.
Another awkward part is that since Safari doesn't support eval sourceURL
exposed into `error.stack` - it means that `captureOwnerStack()` get an
empty location for server components since the fake eval doesn't work
there. That's not a big deal since these stacks are already broken even
for client modules for many because the `eval-source-map` strategy in
Webpack doesn't work in Safari for this same reason.
A lot of this refactoring is just clarifying that there's three kind of
ReactComponentInfo fields:
- `stack` - The raw stack as described on the original server.
- `debugStack` - The Error object containing the stack as represented in
the current client as fake evals.
- `debugTask` - The same thing as `debugStack` but described in terms of
a native `console.createTask`.
Unfortunately, Firefox doesn't include the name of a function in stack
traces if you set it as either `.name` or `.displayName` at runtime.
Only if you include it declarative.
We also can't include it into a named function expression because not
all possible names are expressible declaratively. E.g. spaces or
punctuations.
However, we can express any name if it's an object property and since
object properties now give their name declarative to the function
defined inside of them, we can declaratively express any name this way.
Stacked on #30308.
This is now a noop module so we can stop applying the transform of
console.error using the Babel plugin in the mainline builds. I'm keeping
the transform for RN/WWW for now although it might be nice if the
transform moved into those systems as it gets synced instead of keeping
it upstream.
In jest tests we're already not running the forks for RN/WWW so we don't
need it at all there.
This marker can then be emitted as a getter. When this object gets
accessed we use a special error to let the user know what is going on.
<img width="1350" alt="Screenshot 2024-07-08 at 10 13 46 PM"
src="https://github.com/facebook/react/assets/63648/e3eb698f-e02d-4394-a051-ba9ac3236480">
When you click the `...`:
<img width="1357" alt="Screenshot 2024-07-08 at 10 13 56 PM"
src="https://github.com/facebook/react/assets/63648/4b8ce1cf-d762-49a4-97b9-aeeb1aa8722c">
I also increased the object limit in console logs. It was arbitrarily
set very low before.
These limits are per message. So if you have a loop of many logs it can
quickly add up a lot of strain on the server memory and the client. This
is trying to find some tradeoff. Unfortunately we don't really do much
deduping in these logs so with cyclic objects it ends up maximizing the
limit and then siblings aren't logged.
Ideally we should be able to lazy load them but that requires a lot of
plumbing to wire up so if we can avoid it we should try to. If we want
to that though one idea is to use the getter to do a sync XHR to load
more data but the server needs to retain the objects in memory for an
unknown amount of time. The client could maybe send a signal to retain
them until a weakref clean up but even then it kind of needs a heartbeat
to let the server know the client is still alive. That's a lot of
complexity. There's probably more we can do to optimize deduping and
other parts of the protocol to make it possible to have even higher
limits.
Defaults to true in browser builds, otherwise defaults to false. The
assumption is that the server logs will already contain a log from the
original Flight server.
We currently always replay console logs but this leads to duplicates on
the server by default when you use SSR, because the Flight Client on the
server replays the logs. This can be nice since those logs gets badged.
It can also be nice if they're running in separate servers but when
they're logging to the same stream it's annoying. Which is really the
typical set up so we should just make that the default but leave it
configurable.
It can be efficient to accept raw string chunks to pass through a stream
instead of encoding them into a binary copy first.
Previously our Flight parsers didn't accept receiving string chunks.
That's partly because we sometimes need to encode binary chunks anyway
so string only transport isn't enough but some chunks can be strings.
This adds a partial ability for chunks to be received as strings.
However, accepting strings comes with some downsides. E.g. if the
strings are split up we need to buffer it which compromises the perf for
the common case. If the chunk represents binary data, then we'd need to
encode it back into a typed array which would require a TextEncoder
dependency in the parser. If the string chunk represents a byte length
encoded string we don't know how many unicode characters to read without
measuring them in terms of binary - also requiring a TextEncoder.
This PR is mainly intended for use for pass-through within the same
memory. We can simplify the implementation by assuming that any string
chunk is passed as the original chunk. This requires that the server
stream config doesn't arbitrarily concatenate strings (e.g. large
strings should not be concatenated which is probably a good heuristic
anyway). It also means that this is not suitable to be used with for
example receiving string chunks on the client by passing them through
SSR hydration data - except if the encoding that way was only used with
chunks that were already encoded as strings by Flight.
Web streams mostly just work on binary data anyway so they can't use
this.
In Node.js streams we concatenate precomputed and small strings into
larger buffers. It might make sense to do that using string ropes
instead. However, in the meantime we can at least pass large strings
that are outside our buffer view size as raw strings. There's no benefit
to us eagerly encoding those.
Also, let Node accept string chunks as long as they're following our
expected constraints. This lets us test the mixed protocol using
pass-throughs. This can also be useful when the RSC server is in the
same environment as the SSR server as they don't have to go from strings
to typed arrays back to strings.
Now we can also use this in the pass-through used in renderToMarkup.
This lets us avoid the dependency on TextDecoder/TextEncoder in that
package.
When we replay logs we badge them with e.g. `[Server]`. That way it's
easy to identify that the source of the log actually happened on the
Server (RSC). However, when we threw an error we didn't have any such
thing. The error was rethrown on the client and then handled just like
any other client error.
This transfers the `environmentName` in DEV to our restored Error
"sub-class" (conceptually) along with `digest`. That way you can read
`error.environmentName` to print this in your own UI.
I also updated our default for `onCaughtError` (and `onError` in Fizz)
to use the `printToConsole` helper that the Flight Client uses to log it
with the badge format. So by default you get the same experience as
console.error for caught errors:
<img width="810" alt="Screenshot 2024-06-10 at 9 25 12 PM"
src="https://github.com/facebook/react/assets/63648/8490fedc-09f6-4286-9332-fbe6b0faa2d3">
<img width="815" alt="Screenshot 2024-06-10 at 9 39 30 PM"
src="https://github.com/facebook/react/assets/63648/bdcfc554-504a-4b1d-82bf-b717e74975ac">
Unfortunately I can't do the same thing for `onUncaughtError` nor
`onRecoverableError` because they use `reportError` which doesn't have
custom formatting (unless we also prevented default on window.onerror).
However maybe that's ok because 1) you should always have an error
boundary 2) it's not likely that an RSC error can actually recover
because it's not going to be rendered again so shouldn't really happen
outside some parent conditionally rendering maybe.
The other problem with this approach is that the default is no longer
trivial - so reimplementing the default in user space is trickier and
ideally we shouldn't expose our default to be called.
Stacked on #29807.
This lets the nearest Suspense/Error Boundary handle it even if that
boundary is defined by the model itself.
It also ensures that when we have an error during serialization of
properties, those can be associated with the nearest JSX element and
since we have a stack/owner for that element we can use it to point to
the source code of that line. We can't track the source of any nested
arbitrary objects deeper inside since objects don’t track their stacks
but close enough. Ideally we have the property path but we don’t have
that right now. We have a partial in the message itself.
<img width="813" alt="Screenshot 2024-06-09 at 10 08 27 PM"
src="https://github.com/facebook/react/assets/63648/917fbe0c-053c-4204-93db-d68a66e3e874">
Note: The component name (Counter) is lost in the first message because
we don't print it in the Task. We use `"use client"` instead because we
expect the next stack frame to have the name. We also don't include it
in the actual error message because the Server doesn't know the
component name yet. Ideally Client References should be able to have a
name. If the nearest is a Host Component then we do use the name though.
However, it's not actually inside that Component that the error happens
it's in App and that points to the right line number.
An interesting case is that if something that's actually going to be
consumed by the props to a Suspense/Error Boundary or the Client
Component that wraps them fails, then it can't be handled by the
boundary. However, a counter intuitive case might be when that's on the
`children` props. E.g.
`<ErrorBoundary>{clientReferenceOrInvalidSerialization}</ErrorBoundary>`.
This value can be inspected by the boundary so it's not safe to pass it
so if it's errored it is not caught.
## Implementation
The first insight is that this is best solved on the Client rather than
in the Server because that way it also covers Client References that end
up erroring.
The key insight is that while we don't have a true stack when using
`JSON.parse` and therefore no begin/complete we can still infer these
phases for Elements because the first child of an Element is always
`'$'` which is also a leaf. In depth first that's our begin phase. When
the Element itself completes, we have the complete phase. Anything in
between is within the Element.
Using this idea I was able to refactor the blocking tracking mechanism
to stash the blocked information on `initializingHandler` and then on
the way up do we let whatever is nearest handle it - whether that's an
Element or the root Chunk. It's kind of like an Algebraic Effect.
cc @unstubbable This is something you might want to deep dive into to
find more edge cases. I'm sure I've missed something.
---------
Co-authored-by: eps1lon <sebastian.silbermann@vercel.com>
Stacked on #29807.
Conceptually the error's owner/task should ideally be captured when the
Error constructor is called but neither `console.createTask` does this,
nor do we override `Error` to capture our `owner`. So instead, we use
the nearest parent as the owner/task of the error. This is usually the
same thing when it's thrown from the same async component but not if you
await a promise started from a different component/task.
Before this stack the "owner" and "task" of a Lazy that errors was the
nearest Fiber but if the thing erroring is a Server Component, we need
to get that as the owner from the inner most part of debugInfo.
To get the Task for that Server Component, we need to expose it on the
ReactComponentInfo object. Unfortunately that makes the object not
serializable so we need to special case this to exclude it from
serialization. It gets restored again on the client.
Before (Shell):
<img width="813" alt="Screenshot 2024-06-06 at 5 16 20 PM"
src="https://github.com/facebook/react/assets/63648/7da2d4c9-539b-494e-ba63-1abdc58ff13c">
After (App):
<img width="811" alt="Screenshot 2024-06-08 at 12 29 23 AM"
src="https://github.com/facebook/react/assets/63648/dbf40bd7-c24d-4200-81a6-5018bef55f6d">
We know from Fiber that inline objects with more than 16 properties in
V8 turn into dictionaries instead of optimized objects. The trick is to
use a constructor instead of an inline object literal. I don't actually
know if that's still the case or not. I haven't benchmarked/tested the
output. Better safe than sorry.
It's unfortunate that this can have a negative effect for Hermes and JSC
but it's not as bad as it is for V8 because they don't deopt into
dictionaries. The time to construct these objects isn't a concern - the
time to access them frequently is.
We have to beware the Task objects in Fizz. Those are currently on 16
fields exactly so we shouldn't add anymore ideally.
We should ideally have a lint rule against object literals with more
than 16 fields on them. It might not help since sometimes the fields are
conditional.
This lets you click a stack frame on the client and see the Server
source code inline.
<img width="871" alt="Screenshot 2024-06-01 at 11 44 24 PM"
src="https://github.com/facebook/react/assets/63648/581281ce-0dce-40c0-a084-4a6d53ba1682">
<img width="840" alt="Screenshot 2024-06-01 at 11 43 37 PM"
src="https://github.com/facebook/react/assets/63648/00dc77af-07c1-4389-9ae0-cf1f45199efb">
We could do some logic on the server that sends a source map url for
every stack frame in the RSC payload. That would make the client
potentially config free. However regardless we need the config to
describe what url scheme to use since that’s not built in to the bundler
config. In practice you likely have a common pattern for your source
maps so no need to send data over and over when we can just have a
simple function configured on the client.
The server must return a source map, even if the file is not actually
compiled since the fake file is still compiled.
The source mapping strategy can be one of two models depending on if the
server’s stack traces (`new Error().stack`) are source mapped back to
the original (`—enable-source-maps`) or represents the location in
compiled code (like in the browser).
If it represents the location in compiled code it’s actually easier. You
just serve the source map generated for that file by the tooling.
If it is already source mapped it has to generate a source map where
everything points to the same location (as if not compiled) ideally with
a segment per logical ast node.
This lets any element created from the server, to bottom out with a
client "owner" which is the creator of the Flight request. This could be
a Server Action being invoked or a router.
This is similar to how a client element bottoms out in the creator of
the root element without an owner. E.g. where the root app element was
created.
Without this, we inherit the task of whatever is currently executing
when we're parsing which can be misleading.
Before:
<img width="507" alt="Screenshot 2024-05-30 at 12 06 57 PM"
src="https://github.com/facebook/react/assets/63648/e234db7e-67f7-404c-958a-5c5500ffdf1f">
After:
<img width="555" alt="Screenshot 2024-05-30 at 4 59 04 PM"
src="https://github.com/facebook/react/assets/63648/8ba6acb4-2ffd-49d4-bd44-08228ad4200e">
The before/after doesn't show much of a difference here but that's just
because our Flight parsing loop is an async, which maybe it shouldn't be
because it can be unnecessarily deep, and it creates a hidden line for
every loop. That's what the `Promise.then` is. If the element is lazily
initialized it's worse because we can end up in an unrelated render task
as the owner - although that's its own problem.
Follow up to https://github.com/facebook/react/pull/29632.
It's possible for `eval` to throw such as if we're in a CSP environment.
This is non-essential debug information. We can still proceed to create
a fake stack entry. It'll still have the right name. It just won't have
the right line/col number nor source url/source map. It might also be
ignored listed since it's inside Flight.
We have three kinds of stacks that we send in the RSC protocol:
- The stack trace where a replayed `console.log` was called on the
server.
- The JSX callsite that created a Server Component which then later
called another component.
- The JSX callsite that created a Host or Client Component.
These stack frames disappear in native stacks on the client since
they're executed on the server. This evals a fake file which only has
one call in it on the same line/column as the server. Then we call
through these fake modules to "replay" the callstack. We then replay the
`console.log` within this stack, or call `console.createTask` in this
stack to recreate the stack.
The main concern with this approach is the performance. It adds
significant cost to create all these eval:ed functions but it should
eventually balance out.
This doesn't yet apply source maps to these. With source maps it'll be
able to show the server source code when clicking the links.
I don't love how these appear.
- Because we haven't yet initialized the client module we don't have the
name of the client component we're about to render yet which leads to
the `<...>` task name.
- The `(async)` suffix Chrome adds is still a problem.
- The VMxxxx prefix is used to disambiguate which is noisy. Might be
helped by source maps.
- The continuation of the async stacks end up rooted somewhere in the
bootstrapping of the app. This might be ok when the bootstrapping ends
up ignore listed but it's kind of a problem that you can't clear the
async stack.
<img width="927" alt="Screenshot 2024-05-28 at 11 58 56 PM"
src="https://github.com/facebook/react/assets/63648/1c9d32ce-e671-47c8-9d18-9fab3bffabd0">
<img width="431" alt="Screenshot 2024-05-28 at 11 58 07 PM"
src="https://github.com/facebook/react/assets/63648/52f57518-bbed-400e-952d-6650835ac6b6">
<img width="327" alt="Screenshot 2024-05-28 at 11 58 31 PM"
src="https://github.com/facebook/react/assets/63648/d311a639-79a1-457f-9a46-4f3298d07e65">
<img width="817" alt="Screenshot 2024-05-28 at 11 59 12 PM"
src="https://github.com/facebook/react/assets/63648/3aefd356-acf4-4daa-bdbf-b8c8345f6d4b">
This is necessary to simplify the component stack handling to make way
for owner stacks. It also solves some hacks that we used to have but
don't quite make sense. It also solves the problem where things like key
warnings get silenced in RSC because they get deduped. It also surfaces
areas where we were missing key warnings to begin with.
Almost every type of warning is issued from the renderer. React Elements
are really not anything special themselves. They're just lazily invoked
functions and its really the renderer that determines there semantics.
We have three types of warnings that previously fired in
JSX/createElement:
- Fragment props validation.
- Type validation.
- Key warning.
It's nice to be able to do some validation in the JSX/createElement
because it has a more specific stack frame at the callsite. However,
that's the case for every type of component and validation. That's the
whole point of enableOwnerStacks. It's also not sufficient to do it in
JSX/createElement so we also have validation in the renderers too. So
this validation is really just an eager validation but also happens
again later.
The problem with these is that we don't really know what types are valid
until we get to the renderer. Additionally, by placing it in the
isomorphic code it becomes harder to do deduping of warnings in a way
that makes sense for that renderer. It also means we can't reuse logic
for managing stacks etc.
Fragment props validation really should just be part of the renderer
like any other component type. This also matters once we add Fragment
refs and other fragment features. So I moved this into Fiber. However,
since some Fragments don't have Fibers, I do the validation in
ChildFiber instead of beginWork where it would normally happen.
For `type` validation we already do validation when rendering. By
leaving it to the renderer we don't have to hard code an extra list.
This list also varies by context. E.g. class components aren't allowed
in RSC but client references are but we don't have an isomorphic way to
identify client references because they're defined by the host config so
the current logic is flawed anyway. I kept the early validation for now
without the `enableOwnerStacks` since it does provide a nicer stack
frame but with that flag on it'll be handled with nice stacks anyway. I
normalized some of the errors to ensure tests pass.
For `key` validation it's the same principle. The mechanism for the
heuristic is still the same - if it passes statically through a parent
JSX/createElement call then it's considered validated. We already did
print the error later from the renderer so this also disables the early
log in the `enableOwnerStacks` flag.
I also added logging to Fizz so that key warnings can print in SSR logs.
Flight is a bit more complex. For elements that end up on the client we
just pass the `validated` flag along to the client and let the client
renderer print the error once rendered. For server components we log the
error from Flight with the server component as the owner on the stack
which will allow us to print the right stack for context. The factoring
of this is a little tricky because we only want to warn if it's in an
array parent but we want to log the error later to get the right debug
info.
Fiber/Fizz has a similar factoring problem that causes us to create a
fake Fiber for the owner which means the logs won't be associated with
the right place in DevTools.
Follow up to https://github.com/facebook/react/pull/29201. If a chunk
had listeners attached already (e.g. because `.then` was called on the
chunk returned from `createFromReadableStream`),
`wakeChunkIfInitialized` would overwrite any listeners added during
chunk initialization. This caused cyclic [path
references](https://github.com/facebook/react/pull/28996) within that
chunk to never resolve. Fixed by merging the two arrays of listeners.
Fixes#29200
The cyclic state might have added listeners that will still need to be
invoked. This happens if we have a cyclic reference AND end up blocked.
We have already cleared these before entering the parsing when we enter
the CYCLIC state so we they already have the right type. If listeners
are added during this phase they should carry over to the blocked state.
---------
Co-authored-by: Hendrik Liebau <mail@hendrik-liebau.de>
Stacked on #28997.
We can use the technique of referencing an object by its row + property
name path for temporary references - like we do for deduping. That way
we don't need to generate an ID for temporary references. Instead, they
can just be an opaque marker in the slot and it has the implicit ID of
the row + path.
Then we can stash all objects, even the ones that are actually available
to read on the server, as temporary references. Without adding anything
to the payload since the IDs are implicit. If the same object is
returned to the client, it can be referenced by reference instead of
serializing it back to the client. This also helps preserve object
identity.
We assume that the objects are immutable when they pass the boundary.
I'm not sure if this is worth it but with this mechanism, if you return
the `FormData` payload from a `useActionState` it doesn't have to be
serialized on the way back to the client. This is a common pattern for
having access to the last submission as "default value" to the form
fields. However you can still control it by replacing it with another
object if you want. In MPA mode, the temporary references are not
configured and so it needs to be serialized in that case. That's
required anyway for hydration purposes.
I'm not sure if people will actually use this in practice though or if
FormData will always be destructured into some other object like with a
library that turns it into typed data, and back. If so, the object
identity is lost.
Instead of forcing an object to be outlined to be able to refer to it
later we can refer to it by the property path inside another parent
object.
E.g. this encodes such a reference as `'$123:props:children:foo:bar'`.
That way we don't have to preemptively outline object and we can dedupe
after the first time we've found it.
There's no cost on the client if it's not used because we're not storing
any additional information preemptively.
This works mainly because we only have simple JSON objects from the root
reference. Complex objects like Map, FormData etc. are stored as their
entries array in the look up and not the complex object. Other complex
objects like TypedArrays or imports don't have deeply nested objects in
them that can be referenced.
This solves the problem that we only dedupe after the third instance.
This dedupes at the second instance. It also solves the problem where
all nested objects inside deduped instances also are outlined.
The property paths can get pretty large. This is why a test on payload
size increased. We could potentially outline the reference itself at the
first dupe. That way we get a shorter ID to refer to in the third
instance.
This is the first step to experimenting with a new type of stack traces
behind the `enableOwnerStacks` flag - in DEV only.
The idea is to generate stacks that are more like if the JSX was a
direct call even though it's actually a lazy call. Not only can you see
which exact JSX call line number generated the erroring component but if
that's inside an abstraction function, which function called that
function and if it's a component, which component generated that
component. For this to make sense it really need to be the "owner" stack
rather than the parent stack like we do for other component stacks. On
one hand it has more precise information but on the other hand it also
loses context. For most types of problems the owner stack is the most
useful though since it tells you which component rendered this
component.
The problem with the platform in its current state is that there's two
ways to deal with stacks:
1) `new Error().stack`
2) `console.createTask()`
The nice thing about `new Error().stack` is that we can extract the
frames and piece them together in whatever way we want. That is great
for constructing custom UIs like error dialogs. Unfortunately, we can't
take custom stacks and set them in the native UIs like Chrome DevTools.
The nice thing about `console.createTask()` is that the resulting stacks
are natively integrated into the Chrome DevTools in the console and the
breakpoint debugger. They also automatically follow source mapping and
ignoreLists. The downside is that there's no way to extract the async
stack outside the native UI itself so this information cannot be used
for custom UIs like errors dialogs. It also means we can't collect this
on the server and then pass it to the client for server components.
The solution here is that we use both techniques and collect both an
`Error` object and a `Task` object for every JSX call.
The main concern about this approach is the performance so that's the
main thing to test. It's certainly too slow for production but it might
also be too slow even for DEV.
This first PR doesn't actually use the stacks yet. It just collects them
as the first step. The next step is to start utilizing this information
in error printing etc.
For RSC we pass the stack along across over the wire. This can be
concatenated on the client following the owner path to create an owner
stack leading back into the server. We'll later use this information to
restore fake frames on the client for native integration. Since this
information quickly gets pretty heavy if we include all frames, we strip
out the top frame. We also strip out everything below the functions that
call into user space in the Flight runtime. To do this we need to figure
out the frames that represents calling out into user space. The
resulting stack is typically just the one frame inside the owner
component's JSX callsite. I also eagerly strip out things we expect to
be ignoreList:ed anyway - such as `node_modules` and Node.js internals.
For [`AsyncIterable`](https://github.com/facebook/react/pull/28847) we
encode `AsyncIterator` as a separate tag.
Previously we encoded `Iterator` as just an Array. This adds a special
encoding for this. Technically this is a breaking change.
This is kind of an edge case that you'd care about the difference but it
becomes more important to treat these correctly for the warnings here
#28853.
So that when we end up referring to it in more places, it's only one.
We don't do this same pattern for regular `Symbol.iterator` because we
also support the string `"@@iterator"` for backwards compatibility.
This adds support in Flight for serializing four kinds of streams:
- `ReadableStream` with objects as a model. This is a single shot
iterator so you can read it only once. It can contain any value
including Server Components. Chunks are encoded as is so if you send in
10 typed arrays, you get the same typed arrays out on the other side.
- Binary `ReadableStream` with `type: 'bytes'` option. This supports the
BYOB protocol. In this mode, the receiving side just gets `Uint8Array`s
and they can be split across any single byte boundary into arbitrary
chunks.
- `AsyncIterable` where the `AsyncIterator` function is different than
the `AsyncIterable` itself. In this case we assume that this might be a
multi-shot iterable and so we buffer its value and you can iterate it
multiple times on the other side. We support the `return` value as a
value in the single completion slot, but you can't pass values in
`next()`. If you want single-shot, return the AsyncIterator instead.
- `AsyncIterator`. These gets serialized as a single-shot as it's just
an iterator.
`AsyncIterable`/`AsyncIterator` yield Promises that are instrumented
with our `.status`/`.value` convention so that they can be synchronously
looped over if available. They are also lazily parsed upon read.
We can't do this with `ReadableStream` because we use the native
implementation of `ReadableStream` which owns the promises.
The format is a leading row that indicates which type of stream it is.
Then a new row with the same ID is emitted for every chunk. Followed by
either an error or close row.
`AsyncIterable`s can also be returned as children of Server Components
and then they're conceptually the same as fragment arrays/iterables.
They can't actually be used as children in Fizz/Fiber but there's a
separate plan for that. Only `AsyncIterable` not `AsyncIterator` will be
valid as children - just like sync `Iterable` is already supported but
single-shot `Iterator` is not. Notably, neither of these streams
represent updates over time to a value. They represent multiple values
in a list.
When the server stream is aborted we also close the underlying stream.
However, closing a stream on the client, doesn't close the underlying
stream.
A couple of possible follow ups I'm not planning on doing right now:
- [ ] Free memory by releasing the buffer if an Iterator has been
exhausted. Single shots could be optimized further to release individual
items as you go.
- [ ] We could clean up the underlying stream if the only pending data
that's still flowing is from streams and all the streams have cleaned
up. It's not very reliable though. It's better to do cancellation for
the whole stream - e.g. at the framework level.
- [ ] Implement smarter Binary Stream chunk handling. Currently we wait
until we've received a whole row for binary chunks and copy them into
consecutive memory. We need this to preserve semantics when passing
typed arrays. However, for binary streams we don't need that. We can
just send whatever pieces we have so far.
With the enableBinaryFlight flag on we should encode typed arrays and
blobs in the Reply direction too for parity.
It's already possible to pass Blobs inside FormData but you should be
able to pass them inside objects too.
We encode typed arrays as blobs and then unwrap them automatically to
the right typed array type.
Unlike the other protocol, I encode the type as a reference tag instead
of row tag. Therefore I need to rename the tags to avoid conflicts with
other tags in references. We are running out of characters though.
We used to assume that outlined models are emitted before the reference
(which was true before Blobs). However, it still wasn't safe to assume
that all the data will be available because an "import" (client
reference) can be async and therefore if it's directly a child of an
outlined model, it won't be able to update in place.
This is a similar problem as the one hit by @unstubbable in #28669 with
elements, but a little different since these don't follow the same way
of wrapping.
I don't love the structuring of this code which now needs to pass a
first class mapper instead of just being known code. It also shares the
host path which is just an identity function. It wouldn't necessarily
pass my own review but I don't have a better one for now. I'd really
prefer if this was done at a "row" level but that ends up creating even
more code.
Add test for Blob in FormData and async modules in Maps.
We currently support FormData for Replies mainly for Form Actions. This
supports it in the other direction too which lets you return it from an
action as the response. Mainly for parity.
We don't really recommend that you just pass the original form data back
because the action is supposed to be able to clear fields and such but
you could potentially at least use this as the format and could clear
some fields.
We could potentially optimize this with a temporary reference if the
same object was passed to a reply in case you use it as a round trip to
avoid serializing it back again. That way the action has the ability to
override it to clear fields but if it doesn't you get back the same as
you sent.
#28755 adds support for Blobs when the `enableBinaryFlight` is enabled
which allows them to be used inside FormData too.
We currently support Blobs when passing from Client to Server so this
adds it in the other direction for parity - when `enableFlightBinary` is
enabled.
We intentionally only support the `Blob` type to pass-through, not
subtype `File`. That's because passing additional meta data like
filename might be an accidental leak. You can still pass a `File`
through but it'll appear as a `Blob` on the other side. It's also not
possible to create a faithful File subclass in all environments without
it actually being backed by a file.
This implementation isn't great but at least it works. It creates a few
indirections. This is because we need to be able to asynchronously emit
the buffers but we have to "block" the parent object from resolving
while it's loading.
Ideally, we should be able to create the Blob on the client early and
then stream in it lazily. Because the Blob API doesn't guarantee that
the data is available synchronously. Unfortunately, the native APIs
doesn't have this. We could implement custom versions of all the data
read APIs but then the blobs still wouldn't work with native APIs. So we
just have to wait until Blob accepts a stream in the constructor.
We should be able to stream each chunk early in the protocol though even
though we can't unblock the parent until they've all loaded. I didn't do
this yet mostly because of code structure and I'm lazy.
Hendrik Liebau
Sebastian Markbåge
Josh Story
Jan Kassens
Janka Uryga