Experiment: Unsuspend all lanes on update (#20660)

acdlite · web-flow · commit d13f5b9538e4 · 2021-01-26T12:23:34.000-08:00
Adds a feature flag to tweak the internal heuristic used to "unsuspend"
lanes when a new update comes in.

A lane is "suspended" if we couldn't finish rendering it because it was
missing data, and we chose not to commit the fallback. (In this context,
"suspended" does not include updates that finished with a fallback.)

When we receive new data in the form of an update, we need to retry
rendering the suspended lanes, since the new data may have unblocked the
previously suspended work. For example, the new update could navigate
back to an already loaded route.

It's impractical to retry every combination of suspended lanes, so we
need some heuristic that decides which lanes to retry and in
which order.

The existing heuristic roughly approximates the old Expiration Times
model. It unsuspends all lower priority lanes, but leaves higher
priority lanes suspended.

Then when we start rendering, we choose the lanes that have the highest
LanePriority and render those -- and then we add to that all the lanes
that are highher priority.

If this sounds terribly confusing, it's because it barely makes sense.
(It made more sense in the Expiration Times world, I promise, but it
was still confusing.) I don't think it's worth me trying to explain the
old behavior too much because the point here is that we can replace it
with something simpler.

The new heurstic is to unsuspend all suspended lanes whenever there's
an update.

This is effectively what we already do except in a few very specific
edge cases, ever since we removed the delayed suspense feature from
everything that's not a refresh transition.

We can optimize this in the future to only unsuspend lanes that are
either 1) in the `lanes` or `subtreeLanes` of the node that was updated,
or 2) in the `lanes` of the return path of the node that was updated.
This would exclude lanes that are only located in unrelated sibling
trees. But, this optimization wouldn't be useful currently because we
assign the same transition lane to all transitions. It will become
relevant again once we start assigning arbitrary lanes to transitions
-- but that in turn requires us to implement entanglement of overlapping
transitions, one of our planned projects.

So to sum up: the goal here is to remove the weird edge cases and switch
to a simpler model, on top of which we can make more substantial
improvements.

I put it behind a flag so I can run an A/B test and confirm it doesn't
cause a regression.
diff --git a/packages/react-reconciler/src/ReactFiberLane.new.js b/packages/react-reconciler/src/ReactFiberLane.new.js
@@ -36,7 +36,10 @@ export type Lane = number;
 export type LaneMap<T> = Array<T>;
 
 import invariant from 'shared/invariant';
-import {enableCache} from 'shared/ReactFeatureFlags';
+import {
+  enableCache,
+  enableTransitionEntanglement,
+} from 'shared/ReactFeatureFlags';
 
 import {
   ImmediatePriority as ImmediateSchedulerPriority,
@@ -306,9 +309,15 @@ export function getNextLanes(root: FiberRoot, wipLanes: Lanes): Lanes {
     return NoLanes;
   }
 
-  // If there are higher priority lanes, we'll include them even if they
-  // are suspended.
-  nextLanes = pendingLanes & getEqualOrHigherPriorityLanes(nextLanes);
+  if (enableTransitionEntanglement) {
+    // We don't need to include higher priority lanes, because in this
+    // experiment we always unsuspend all transitions whenever we receive
+    // an update.
+  } else {
+    // If there are higher priority lanes, we'll include them even if they
+    // are suspended.
+    nextLanes = pendingLanes & getEqualOrHigherPriorityLanes(nextLanes);
+  }
 
   // If we're already in the middle of a render, switching lanes will interrupt
   // it and we'll lose our progress. We should only do this if the new lanes are
@@ -673,8 +682,27 @@ export function markRootUpdated(
   // Unsuspend any update at equal or lower priority.
   const higherPriorityLanes = updateLane - 1; // Turns 0b1000 into 0b0111
 
-  root.suspendedLanes &= higherPriorityLanes;
-  root.pingedLanes &= higherPriorityLanes;
+  if (enableTransitionEntanglement) {
+    // If there are any suspended transitions, it's possible this new update
+    // could unblock them. Clear the suspended lanes so that we can try rendering
+    // them again.
+    //
+    // TODO: We really only need to unsuspend only lanes that are in the
+    // `subtreeLanes` of the updated fiber, or the update lanes of the return
+    // path. This would exclude suspended updates in an unrelated sibling tree,
+    // since there's no way for this update to unblock it.
+    //
+    // We don't do this if the incoming update is idle, because we never process
+    // idle updates until after all the regular updates have finished; there's no
+    // way it could unblock a transition.
+    if ((updateLane & IdleLanes) === NoLanes) {
+      root.suspendedLanes = NoLanes;
+      root.pingedLanes = NoLanes;
+    }
+  } else {
+    root.suspendedLanes &= higherPriorityLanes;
+    root.pingedLanes &= higherPriorityLanes;
+  }
 
   const eventTimes = root.eventTimes;
   const index = laneToIndex(updateLane);
diff --git a/packages/react-reconciler/src/ReactFiberLane.old.js b/packages/react-reconciler/src/ReactFiberLane.old.js
@@ -36,7 +36,10 @@ export type Lane = number;
 export type LaneMap<T> = Array<T>;
 
 import invariant from 'shared/invariant';
-import {enableCache} from 'shared/ReactFeatureFlags';
+import {
+  enableCache,
+  enableTransitionEntanglement,
+} from 'shared/ReactFeatureFlags';
 
 import {
   ImmediatePriority as ImmediateSchedulerPriority,
@@ -306,9 +309,15 @@ export function getNextLanes(root: FiberRoot, wipLanes: Lanes): Lanes {
     return NoLanes;
   }
 
-  // If there are higher priority lanes, we'll include them even if they
-  // are suspended.
-  nextLanes = pendingLanes & getEqualOrHigherPriorityLanes(nextLanes);
+  if (enableTransitionEntanglement) {
+    // We don't need to include higher priority lanes, because in this
+    // experiment we always unsuspend all transitions whenever we receive
+    // an update.
+  } else {
+    // If there are higher priority lanes, we'll include them even if they
+    // are suspended.
+    nextLanes = pendingLanes & getEqualOrHigherPriorityLanes(nextLanes);
+  }
 
   // If we're already in the middle of a render, switching lanes will interrupt
   // it and we'll lose our progress. We should only do this if the new lanes are
@@ -673,8 +682,27 @@ export function markRootUpdated(
   // Unsuspend any update at equal or lower priority.
   const higherPriorityLanes = updateLane - 1; // Turns 0b1000 into 0b0111
 
-  root.suspendedLanes &= higherPriorityLanes;
-  root.pingedLanes &= higherPriorityLanes;
+  if (enableTransitionEntanglement) {
+    // If there are any suspended transitions, it's possible this new update
+    // could unblock them. Clear the suspended lanes so that we can try rendering
+    // them again.
+    //
+    // TODO: We really only need to unsuspend only lanes that are in the
+    // `subtreeLanes` of the updated fiber, or the update lanes of the return
+    // path. This would exclude suspended updates in an unrelated sibling tree,
+    // since there's no way for this update to unblock it.
+    //
+    // We don't do this if the incoming update is idle, because we never process
+    // idle updates until after all the regular updates have finished; there's no
+    // way it could unblock a transition.
+    if ((updateLane & IdleLanes) === NoLanes) {
+      root.suspendedLanes = NoLanes;
+      root.pingedLanes = NoLanes;
+    }
+  } else {
+    root.suspendedLanes &= higherPriorityLanes;
+    root.pingedLanes &= higherPriorityLanes;
+  }
 
   const eventTimes = root.eventTimes;
   const index = laneToIndex(updateLane);
diff --git a/packages/react-reconciler/src/__tests__/ReactExpiration-test.js b/packages/react-reconciler/src/__tests__/ReactExpiration-test.js
@@ -733,13 +733,19 @@ describe('ReactExpiration', () => {
       // Both normal pri updates should have expired.
       expect(Scheduler).toFlushExpired([
         'Sibling',
-        // Note: we also flushed the high pri update here, because in the
-        // current implementation, once we pick the next lanes to work on, we
-        // entangle it with all pending at equal or higher priority. We could
-        // feasibly change this heuristic so that the high pri update doesn't
-        // render until after the expired updates have finished. But the
-        // important thing in this test is that the normal updates expired.
-        'High pri: 1',
+        gate(flags => flags.enableTransitionEntanglement)
+          ? // Notice that the high pri update didn't flush yet. Expiring one lane
+            // doesn't affect other lanes. (Unless they are intentionally
+            // entangled, like we do for overlapping transitions that affect the
+            // same state.)
+            'High pri: 0'
+          : // In the current implementation, once we pick the next lanes to work
+            // on, we entangle it with all pending at equal or higher priority.
+            // We could feasibly change this heuristic so that the high pri
+            // update doesn't render until after the expired updates have
+            // finished. But the important thing in this test is that the normal
+            // updates expired.
+            'High pri: 1',
         'Normal pri: 2',
         'Sibling',
       ]);
diff --git a/packages/react-reconciler/src/__tests__/ReactSuspenseList-test.js b/packages/react-reconciler/src/__tests__/ReactSuspenseList-test.js
@@ -2443,41 +2443,30 @@ describe('ReactSuspenseList', () => {
 
     expect(ReactNoop).toMatchRenderedOutput(null);
 
-    ReactNoop.act(() => {
+    await ReactNoop.act(async () => {
       // Add a few items at the end.
       updateLowPri(true);
 
       // Flush partially through.
       expect(Scheduler).toFlushAndYieldThrough(['B', 'C']);
 
       // Schedule another update at higher priority.
-      Scheduler.unstable_runWithPriority(
-        Scheduler.unstable_UserBlockingPriority,
-        () => updateHighPri(true),
-      );
+      ReactNoop.flushSync(() => updateHighPri(true));
 
       // That will intercept the previous render.
-    });
+      expect(Scheduler).toHaveYielded([
+        'Suspend! [A]',
+        'Loading A',
+        // Re-render at forced.
+        'Suspend! [A]',
+        'Loading A',
+      ]);
+      expect(ReactNoop).toMatchRenderedOutput(<span>Loading A</span>);
 
-    jest.runAllTimers();
-
-    expect(Scheduler).toHaveYielded([
-      // First attempt at high pri.
-      'Suspend! [A]',
-      'Loading A',
-      // Re-render at forced.
-      'Suspend! [A]',
-      'Loading A',
-      // We auto-commit this on DEV.
       // Try again on low-pri.
-      'Suspend! [A]',
-      'Loading A',
-      // Re-render at forced.
-      'Suspend! [A]',
-      'Loading A',
-    ]);
-
-    expect(ReactNoop).toMatchRenderedOutput(<span>Loading A</span>);
+      expect(Scheduler).toFlushAndYield(['Suspend! [A]', 'Loading A']);
+      expect(ReactNoop).toMatchRenderedOutput(<span>Loading A</span>);
+    });
 
     await AsyncA.resolve();
 
diff --git a/packages/shared/ReactFeatureFlags.js b/packages/shared/ReactFeatureFlags.js
@@ -149,3 +149,6 @@ export const enableUseRefAccessWarning = false;
 export const enableRecursiveCommitTraversal = false;
 
 export const disableSchedulerTimeoutInWorkLoop = false;
+
+// Experiment to simplify/improve how transitions are scheduled
+export const enableTransitionEntanglement = false;
diff --git a/packages/shared/forks/ReactFeatureFlags.native-fb.js b/packages/shared/forks/ReactFeatureFlags.native-fb.js
@@ -58,6 +58,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.native-oss.js b/packages/shared/forks/ReactFeatureFlags.native-oss.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.test-renderer.js b/packages/shared/forks/ReactFeatureFlags.test-renderer.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.test-renderer.native.js b/packages/shared/forks/ReactFeatureFlags.test-renderer.native.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.test-renderer.www.js b/packages/shared/forks/ReactFeatureFlags.test-renderer.www.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.testing.js b/packages/shared/forks/ReactFeatureFlags.testing.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.testing.www.js b/packages/shared/forks/ReactFeatureFlags.testing.www.js
@@ -57,6 +57,7 @@ export const enableUseRefAccessWarning = false;
 
 export const enableRecursiveCommitTraversal = false;
 export const disableSchedulerTimeoutInWorkLoop = false;
+export const enableTransitionEntanglement = false;
 
 // Flow magic to verify the exports of this file match the original version.
 // eslint-disable-next-line no-unused-vars
diff --git a/packages/shared/forks/ReactFeatureFlags.www-dynamic.js b/packages/shared/forks/ReactFeatureFlags.www-dynamic.js
@@ -55,3 +55,4 @@ export const enableUseRefAccessWarning = __VARIANT__;
 
 export const enableProfilerNestedUpdateScheduledHook = __VARIANT__;
 export const disableSchedulerTimeoutInWorkLoop = __VARIANT__;
+export const enableTransitionEntanglement = __VARIANT__;
diff --git a/packages/shared/forks/ReactFeatureFlags.www.js b/packages/shared/forks/ReactFeatureFlags.www.js
@@ -31,6 +31,7 @@ export const {
   enableUseRefAccessWarning,
   disableNativeComponentFrames,
   disableSchedulerTimeoutInWorkLoop,
+  enableTransitionEntanglement,
 } = dynamicFeatureFlags;
 
 // On WWW, __EXPERIMENTAL__ is used for a new modern build.
