Sourced from @​reduxjs/toolkit's releases.

v2.9.0

This feature release rewrites RTK Query's internal subscription and polling systems and the useStableQueryArgs hook for better perf, adds automatic AbortSignal handling to requests still in progress when a cache entry is removed, fixes a bug with the transformResponse option for queries, adds a new builder.addAsyncThunk method, and fixes assorted other issues.

Changelog

RTK Query Performance Improvements

We had reports that RTK Query could get very slow when there were thousands of subscriptions to the same cache entry. After investigation, we found that the internal polling logic was attempting to recalculate the minimum polling time after every new subscription was added. This was highly inefficient, as most subscriptions don't change polling settings, and it required repeated O(n) iteration over the growing list of subscriptions. We've rewritten that logic to debounce the update check and ensure a max of one polling value update per tick for the entire API instance.

Related, while working on the request abort changes, testing showed that use of plain Records to hold subscription data was inefficient because we have to iterate keys to check size. We've rewritten the subscription handling internals to use Maps instead, as well as restructuring some additional checks around in-flight requests.

These two improvements drastically improved runtime perf for the thousands-of-subscriptions-one-cache-entry repro, eliminating RTK methods as visible hotspots in the perf profiles. It likely also improves perf for general usage as well.

We've also changed the implementation of our internal useStableQueryArgs hook to avoid calling serializeQueryArgs on its value, which can avoid potential perf issues when a query takes a very large object as its cache key.

[!NOTE] The internal logic switched from serializing the query arg to doing reference checks on nested values. This means that if you are passing a non-POJO value in a query arg, such as useSomeQuery({a: new Set()}), and you have refetchOnMountOrArgChange enabled, this will now trigger refeteches each time as the Set references are now considered different based on equality instead of serialization.

Abort Signal Handling on Cleanup

We've had numerous requests over time for various forms of "abort in-progress requests when the data is no longer needed / params change / component unmounts / some expensive request is taking too long". This is a complex topic with multiple potential use cases, and our standard answer has been that we don't want to abort those requests - after all, cache entries default to staying in memory for 1 minute after the last subscription is removed, so RTKQ's cache can still be updated when the request completes. That also means that it doesn't make sense to abort a request "on unmount".

However, it does then make sense to abort an in-progress request if the cache entry itself is removed. Given that, we've updated our cache handling to automatically call the existing resPromise.abort() method in that case, triggering the AbortSignal attached to the baseQuery. The handling at that point depends on your app - fetchBaseQuery should handle that, a custom baseQuery or queryFn would need to listen to the AbortSignal.

We do have an open issue asking for further discussions of potential abort / cancelation use cases and would appreciate further feedback.

New Options

The builder callback used in createReducer and createSlice.extraReducers now has builder.addAsyncThunk available, which allows handling specific actions from a thunk in the same way that you could define a thunk inside createSlice.reducers:

        const slice = createSlice({
          name: 'counter',
          initialState: {
            loading: false,
            errored: false,
            value: 0,
          },
          reducers: {},
          extraReducers: (builder) =>
            builder.addAsyncThunk(asyncThunk, {
              pending(state) {
                state.loading = true
              },
              fulfilled(state, action) {
                state.value = action.payload
              },
              rejected(state) {
                state.errored = true
              },
</tr></table> 

... (truncated)

• 98c54c6 Release 2.9.0
• 0a86abd export some useful async thunk types
• d02c27a add docs for addAsyncThunk (#5066)
• ab346b9 fixdoc: incorrect position of keepUnusedDataFor for two scenarios (#5062)
• d2bbb8d Rewrite subscription handling and polling calculations for better perf (#5064)
• 3c6de47 Merge pull request #5065 from reduxjs/feature/TS5.9-matrix
• c37b977 Add missing error value
• 38537c3 Don't run Node10 checks for TS 6+
• 821be20 Add TS 5.9 to the matrix
• 407688d Abort pending requests if the cache entry is removed (#5061)
• Additional commits viewable in compare view

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot merge will merge this PR after your CI passes on it
• @dependabot squash and merge will squash and merge this PR after your CI passes on it
• @dependabot cancel merge will cancel a previously requested merge and block automerging
• @dependabot reopen will reopen this PR if it is closed
• @dependabot close will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)