Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[Experiment] Lazily propagate context changes (#20890)
* Move context comparison to consumer In the lazy context implementation, not all context changes are propagated from the provider, so we can't rely on the propagation alone to mark the consumer as dirty. The consumer needs to compare to the previous value, like we do for state and context. I added a `memoizedValue` field to the context dependency type. Then in the consumer, we iterate over the current dependencies to see if something changed. We only do this iteration after props and state has already bailed out, so it's a relatively uncommon path, except at the root of a changed subtree. Alternatively, we could move these comparisons into `readContext`, but that's a much hotter path, so I think this is an appropriate trade off. * [Experiment] Lazily propagate context changes When a context provider changes, we scan the tree for matching consumers and mark them as dirty so that we know they have pending work. This prevents us from bailing out if, say, an intermediate wrapper is memoized. Currently, we propagate these changes eagerly, at the provider. However, in many cases, we would have ended up visiting the consumer nodes anyway, as part of the normal render traversal, because there's no memoized node in between that bails out. We can save CPU cycles by propagating changes only when we hit a memoized component — so, instead of propagating eagerly at the provider, we propagate lazily if or when something bails out. Most of our bailout logic is centralized in `bailoutOnAlreadyFinishedWork`, so this ended up being not that difficult to implement correctly. There are some exceptions: Suspense and Offscreen. Those are special because they sometimes defer the rendering of their children to a completely separate render cycle. In those cases, we must take extra care to propagate *all* the context changes, not just the first one. I'm pleasantly surprised at how little I needed to change in this initial implementation. I was worried I'd have to use the reconciler fork, but I ended up being able to wrap all my changes in a regular feature flag. So, we could run an experiment in parallel to our other ones. I do consider this a risky rollout overall because of the potential for subtle semantic deviations. However, the model is simple enough that I don't expect us to have trouble fixing regressions if or when they arise during internal dogfooding. --- This is largely based on [RFC#118](reactjs/rfcs#118), by @gnoff. I did deviate in some of the implementation details, though. The main one is how I chose to track context changes. Instead of storing a dirty flag on the stack, I added a `memoizedValue` field to the context dependency object. Then, to check if something has changed, the consumer compares the new context value to the old (memoized) one. This is necessary because of Suspense and Offscreen — those components defer work from one render into a later one. When the subtree continues rendering, the stack from the previous render is no longer available. But the memoized values on the dependencies list are. This requires a bit more work when a consumer bails out, but nothing considerable, and there are ways we could optimize it even further. Conceptually, this model is really appealing, since it matches how our other features "reactively" detect changes — `useMemo`, `useEffect`, `getDerivedStateFromProps`, the built-in cache, and so on. I also intentionally dropped support for `unstable_calculateChangedBits`. We're planning to remove this API anyway before the next major release, in favor of context selectors. It's an unstable feature that we never advertised; I don't think it's seen much adoption. Co-Authored-By: Josh Story <[email protected]> * Propagate all contexts in single pass Instead of propagating the tree once per changed context, we can check all the contexts in a single propagation. This inverts the two loops so that the faster loop (O(numberOfContexts)) is inside the more expensive loop (O(numberOfFibers * avgContextDepsPerFiber)). This adds a bit of overhead to the case where only a single context changes because you have to unwrap the context from the array. I'm also unsure if this will hurt cache locality. Co-Authored-By: Josh Story <[email protected]> * Stop propagating at nearest dependency match Because we now propagate all context providers in a single traversal, we can defer context propagation to a subtree without losing information about which context providers we're deferring — it's all of them. Theoretically, this is a big optimization because it means we'll never propagate to any tree that has work scheduled on it, nor will we ever propagate the same tree twice. There's an awkward case related to bailing out of the siblings of a context consumer. Because those siblings don't bail out until after they've already entered the begin phase, we have to do extra work to make sure they don't unecessarily propagate context again. We could avoid this by adding an earlier bailout for sibling nodes, something we've discussed in the past. We should consider this during the next refactor of the fiber tree structure. Co-Authored-By: Josh Story <[email protected]> * Mark trees that need propagation in readContext Instead of storing matched context consumers in a Set, we can mark when a consumer receives an update inside `readContext`. I hesistated to put anything in this function because it's such a hot path, but so are bail outs. Fortunately, we only need to set this flag once, the first time a context is read. So I think it's a reasonable trade off. In exchange, propagation is faster because we no longer need to accumulate a Set of matched consumers, and fiber bailouts are faster because we don't need to consult that Set. And the code is simpler. Co-authored-by: Josh Story <[email protected]>
- Loading branch information
