Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Decouple update queue from Fiber type #12600

Merged
merged 10 commits into from Apr 23, 2018

Conversation

acdlite
Copy link
Collaborator

@acdlite acdlite commented Apr 10, 2018

The update queue is in need of a refactor. Recent bugfixes (#12528) have exposed some flaws in how it's modeled. Upcoming features like Suspense and [redacted] also rely on the update queue in ways that weren't anticipated in the original design.

Major changes:

  • Instead of boolean flags for isReplace and isForceUpdate, updates have a tag field (like Fiber). This lowers the cost for adding new types of updates.
  • Render phase updates are special cased. Updates scheduled during the render phase are dropped if the work-in-progress does not commit. This is used for getDerivedStateFrom{Props,Catch}.
  • callbackList has been replaced with a generic effect list. Aside from callbacks, this is also used for componentDidCatch.

TODO:

  • Class components
  • Host roots
  • Fix Flow
  • Update React DevTools dependency on hasForceUpdate (in a separate repo) I reverted this change

@@ -248,7 +248,7 @@ exports[`ReactDebugFiberPerf recovers from caught errors 1`] = `
⛔ (Committing Changes) Warning: Lifecycle hook scheduled a cascading update
⚛ (Committing Snapshot Effects: 0 Total)
⚛ (Committing Host Effects: 2 Total)
⚛ (Calling Lifecycle Methods: 0 Total)
⚛ (Calling Lifecycle Methods: 1 Total)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is because we weren't counting componentDidCatch as a lifecycle previously

@pull-bot
Copy link

pull-bot commented Apr 10, 2018

ReactDOM: size: 🔺+0.4%, gzip: -0.2%

Details of bundled changes.

Comparing: 999b656...e4ce363

react-dom

File Filesize Diff Gzip Diff Prev Size Current Size Prev Gzip Current Gzip ENV
react-dom.development.js +0.3% +0.6% 609.63 KB 611.61 KB 140.53 KB 141.35 KB UMD_DEV
react-dom.production.min.js 🔺+0.4% -0.2% 100.2 KB 100.56 KB 31.97 KB 31.9 KB UMD_PROD
react-dom.development.js +0.3% +0.6% 594 KB 595.99 KB 136.37 KB 137.22 KB NODE_DEV
react-dom.production.min.js 🔺+0.3% 0.0% 98.65 KB 98.99 KB 31.1 KB 31.11 KB NODE_PROD
react-dom-test-utils.development.js -0.0% 0.0% 40.75 KB 40.75 KB 11.69 KB 11.69 KB UMD_DEV
react-dom-test-utils.development.js -0.0% -0.0% 35.61 KB 35.61 KB 10.27 KB 10.27 KB NODE_DEV
ReactDOM-dev.js +0.3% +0.6% 618.41 KB 620.4 KB 139.14 KB 139.94 KB FB_WWW_DEV
ReactDOM-prod.js -0.1% -0.4% 284.94 KB 284.78 KB 52.3 KB 52.09 KB FB_WWW_PROD

react-art

File Filesize Diff Gzip Diff Prev Size Current Size Prev Gzip Current Gzip ENV
react-art.development.js +0.5% +0.9% 412.73 KB 414.71 KB 89.59 KB 90.41 KB UMD_DEV
react-art.production.min.js 🔺+0.4% -0.3% 90.48 KB 90.85 KB 27.62 KB 27.54 KB UMD_PROD
react-art.development.js +0.6% +1.1% 338.56 KB 340.55 KB 70.81 KB 71.62 KB NODE_DEV
react-art.production.min.js 🔺+0.6% -0.5% 55.04 KB 55.37 KB 16.87 KB 16.79 KB NODE_PROD
ReactART-dev.js +0.6% +1.1% 346.81 KB 348.8 KB 70.42 KB 71.18 KB FB_WWW_DEV
ReactART-prod.js -0.1% -0.7% 167.61 KB 167.48 KB 27.75 KB 27.57 KB FB_WWW_PROD

react-test-renderer

File Filesize Diff Gzip Diff Prev Size Current Size Prev Gzip Current Gzip ENV
react-test-renderer.development.js +0.6% +1.1% 345.69 KB 347.67 KB 72.47 KB 73.28 KB UMD_DEV
react-test-renderer.production.min.js 🔺+0.6% -0.2% 54.77 KB 55.13 KB 16.71 KB 16.68 KB UMD_PROD
react-test-renderer.development.js +0.6% +1.2% 336.51 KB 338.5 KB 69.66 KB 70.46 KB NODE_DEV
react-test-renderer.production.min.js 🔺+0.6% -0.1% 54.02 KB 54.36 KB 16.35 KB 16.33 KB NODE_PROD
ReactTestRenderer-dev.js +0.6% +1.1% 344.98 KB 346.97 KB 69.31 KB 70.09 KB FB_WWW_DEV

react-reconciler

File Filesize Diff Gzip Diff Prev Size Current Size Prev Gzip Current Gzip ENV
react-reconciler.development.js +0.6% +1.2% 316.9 KB 318.89 KB 65.12 KB 65.92 KB NODE_DEV
react-reconciler.production.min.js 🔺+0.7% -0.5% 46.91 KB 47.24 KB 14.3 KB 14.23 KB NODE_PROD
react-reconciler-persistent.development.js +0.6% +1.2% 316.23 KB 318.22 KB 64.88 KB 65.69 KB NODE_DEV
react-reconciler-persistent.production.min.js 🔺+0.7% 🔺+0.2% 45.83 KB 46.16 KB 14.08 KB 14.11 KB NODE_PROD
react-reconciler-reflection.development.js -0.0% 0.0% 11.06 KB 11.06 KB 3.42 KB 3.42 KB NODE_DEV

react-native-renderer

File Filesize Diff Gzip Diff Prev Size Current Size Prev Gzip Current Gzip ENV
ReactNativeRenderer-dev.js +0.4% +0.8% 456.63 KB 458.64 KB 97.38 KB 98.15 KB RN_FB_DEV
ReactNativeRenderer-prod.js -0.3% -0.7% 218.26 KB 217.69 KB 36.66 KB 36.39 KB RN_FB_PROD
ReactNativeRenderer-dev.js +0.4% +0.8% 456.38 KB 458.39 KB 97.32 KB 98.08 KB RN_OSS_DEV
ReactNativeRenderer-prod.js 🔺+0.9% -0.1% 215.01 KB 216.91 KB 36.29 KB 36.26 KB RN_OSS_PROD
ReactFabric-dev.js +0.5% +0.9% 439.04 KB 441.05 KB 92.94 KB 93.75 KB RN_FB_DEV
ReactFabric-prod.js 🔺+1.0% 🔺+0.3% 200.59 KB 202.5 KB 33.55 KB 33.65 KB RN_FB_PROD
ReactFabric-dev.js +0.5% +0.9% 439.08 KB 441.09 KB 92.96 KB 93.76 KB RN_OSS_DEV
ReactFabric-prod.js 🔺+1.0% 🔺+0.3% 200.63 KB 202.54 KB 33.57 KB 33.67 KB RN_OSS_PROD

Generated by 🚫 dangerJS

@gaearon
Copy link
Collaborator

gaearon commented Apr 10, 2018

Need to add “update DevTools” to todos. I think it might rely on isForceUpdate somewhere.

@bvaughn
Copy link
Contributor

bvaughn commented Apr 10, 2018

Need to add “update DevTools” to todos. I think it might rely on isForceUpdate somewhere.

I don't think so? I don't see any references to isForced.

@gaearon
Copy link
Collaborator

gaearon commented Apr 10, 2018

@acdlite
Copy link
Collaborator Author

acdlite commented Apr 10, 2018

Note: the bundle size changes are mostly because I haven't deleted the old update queue yet :D

@acdlite acdlite force-pushed the refactor-updatequeue branch 3 times, most recently from e325697 to a99b4ca Compare April 11, 2018 02:23
@acdlite acdlite changed the title [WIP] Decouple update queue from Fiber type Decouple update queue from Fiber type Apr 11, 2018
@acdlite
Copy link
Collaborator Author

acdlite commented Apr 11, 2018

@sebmarkbage Ready for review

⚛ (Committing Host Effects: 1 Total)
⚛ (Calling Lifecycle Methods: 1 Total)
⚛ (Committing Host Effects: 0 Total)
⚛ (Calling Lifecycle Methods: 0 Total)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't know why this wasn't already 0. There are no host effects or lifecycles in the second flush. New snapshot seems correct.

@@ -274,7 +274,7 @@ exports[`ReactDebugFiberPerf recovers from fatal errors 1`] = `
⚛ (Committing Changes)
⚛ (Committing Snapshot Effects: 0 Total)
⚛ (Committing Host Effects: 1 Total)
⚛ (Calling Lifecycle Methods: 0 Total)
⚛ (Calling Lifecycle Methods: 1 Total)
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same

@acdlite
Copy link
Collaborator Author

acdlite commented Apr 11, 2018

Needless to say, this is a risky change. I'll do the next www sync and fix any bugs that may come up.

@gaearon gaearon mentioned this pull request Apr 11, 2018
@NE-SmallTown
Copy link
Contributor

Upcoming features like Suspense and [redacted] also rely on the update queue in ways that weren't anticipated in the original design.

What does [redacted] mean?

@acdlite acdlite mentioned this pull request Apr 14, 2018
5 tasks
@@ -749,10 +684,11 @@ export default function(
): boolean {
const ctor = workInProgress.type;
const instance = workInProgress.stateNode;
resetInputPointers(workInProgress, instance);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What is this change? You didn't replicate the other assignment so something is new.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Object.assign(resultState, partialState);
}

function enqueueDerivedStateFromProps(
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

DRY for DRYness sake? Makes it harder to read this inline IMO. It also encourages reading the getDerivedStateFromProps property in multiple places if they're also needed at the callsite. Only abstract if it is chunk of work. (On the flip side, this PR successfully unabstract memoizeProps and memoizeState in places.)

if (updateQueue !== null && updateQueue.capturedValues !== null) {
workInProgress.effectTag &= ~DidCapture;
if (typeof instance.componentDidCatch === 'function') {
workInProgress.effectTag |= ErrLog;
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There was a reason we had to apply these effects in the complete phase. Why are we able to move them to the begin phase now?

Copy link
Collaborator Author

@acdlite acdlite Apr 20, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think there was a reason. It happened to be a convenient place to do it. Now we do it while processing the update queue.

@@ -19,14 +19,14 @@ export const Update = /* */ 0b000000000100;
export const PlacementAndUpdate = /* */ 0b000000000110;
export const Deletion = /* */ 0b000000001000;
export const ContentReset = /* */ 0b000000010000;
export const Callback = /* */ 0b000000100000;
export const UpdateQueue = /* */ 0b000000100000;
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What are theses? They're not described in terms of side-effects. Either this needs to change, or the whole thing renamed.

const ClassUpdateQueue = 0;
const RootUpdateQueue = 1;

type UpdateShared<Payload, U> = {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I really had the term "Shared". This has been sneaking into RN too.

It doesn't describe anything about the concept other than DRYness which indicates that maybe you should repeat yourself. If it is real concept, name it.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Stop denying me muh mixins!

type RootUpdateQueueType = UpdateQueueShared<RootUpdate, ReactNodeList>;

type UpdateQueueOwner<Queue, State> = {
alternate: UpdateQueueOwner<Queue> | null,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is clearly not going to be how we describe "redacted" so seems like a too early abstraction. Even just assuming that it is a single argument is a big assumption. Just use Fiber until you have something else.

}
// Clear the list of render phase updates.
finishedQueue.firstRenderPhaseUpdate = finishedQueue.lastRenderPhaseUpdate = null;
}
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You are forking into a first class concept with typeOfUpdateQueue then you're picking separate code paths again in commitEffect. The only thing gain by doing that instead of specializing each function is that you can reuse the code above here. Just break this out into a reusable function and instead specialize each code path without making typeOfUpdateQueue a first class thing.

@acdlite
Copy link
Collaborator Author

acdlite commented Apr 18, 2018

TODO after chatting to @sebmarkbage:

  • Use first-class functions instead of typeOfUpdateQueue
  • Separate effect list into "own" effects and children effects
  • Rename UpdateQueue side effect type to something that makes more sense
  • Remove ForceUpdate effect and go back to queue.hasForceUpdate instead

@acdlite acdlite force-pushed the refactor-updatequeue branch 4 times, most recently from 995a437 to 73d4d48 Compare April 20, 2018 22:57
The update queue is in need of a refactor. Recent bugfixes (facebook#12528) have
exposed some flaws in how it's modeled. Upcoming features like Suspense
and [redacted] also rely on the update queue in ways that weren't
anticipated in the original design.

Major changes:

- Instead of boolean flags for `isReplace` and `isForceUpdate`, updates
have a `tag` field (like Fiber). This lowers the cost for adding new
types of updates.
- Render phase updates are special cased. Updates scheduled during
the render phase are dropped if the work-in-progress does not commit.
This is used for `getDerivedStateFrom{Props,Catch}`.
- `callbackList` has been replaced with a generic effect list. Aside
from callbacks, this is also used for `componentDidCatch`.
I tried to avoid this at first, since we avoid it everywhere else in the Fiber
codebase, but since updates are not in a hot path, the trade off with file size
seems worth it.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

8 participants