v4 is a major version, so there are some breaking changes to be aware of:
You will need to un-/install dependencies and change the imports:
npm uninstall react-query
npm install @tanstack/react-query
npm install @tanstack/react-query-devtools
npm uninstall react-query
npm install @tanstack/react-query
npm install @tanstack/react-query-devtools
- import { useQuery } from 'react-query'
- import { ReactQueryDevtools } from 'react-query/devtools'
+ import { useQuery } from '@tanstack/react-query'
+ import { ReactQueryDevtools } from '@tanstack/react-query-devtools'
- import { useQuery } from 'react-query'
- import { ReactQueryDevtools } from 'react-query/devtools'
+ import { useQuery } from '@tanstack/react-query'
+ import { ReactQueryDevtools } from '@tanstack/react-query-devtools'
To make the import migration easier, v4 comes with a codemod.
The codemod is a best efforts attempt to help you migrate the breaking change. Please review the generated code thoroughly! Also, there are edge cases that cannot be found by the code mod, so please keep an eye on the log output.
You can easily apply it by using one (or both) of the following commands:
If you want to run it against .js or .jsx files, please use the command below:
npx jscodeshift ./path/to/src/ \
--extensions=js,jsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/replace-import-specifier.js
npx jscodeshift ./path/to/src/ \
--extensions=js,jsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/replace-import-specifier.js
If you want to run it against .ts or .tsx files, please use the command below:
npx jscodeshift ./path/to/src/ \
--extensions=ts,tsx \
--parser=tsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/replace-import-specifier.js
npx jscodeshift ./path/to/src/ \
--extensions=ts,tsx \
--parser=tsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/replace-import-specifier.js
Please note in the case of TypeScript you need to use tsx as the parser; otherwise, the codemod won't be applied properly!
Note: Applying the codemod might break your code formatting, so please don't forget to run prettier and/or eslint after you've applied the codemod!
Note: The codemod will only change the imports - you still have to install the separate devtools package manually.
In v3, Query and Mutation Keys could be a String or an Array. Internally, React Query has always worked with Array Keys only, and we've sometimes exposed this to consumers. For example, in the queryFn, you would always get the key as an Array to make working with Default Query Functions easier.
However, we have not followed this concept through to all apis. For example, when using the predicate function on Query Filters you would get the raw Query Key. This makes it difficult to work with such functions if you use Query Keys that are mixed Arrays and Strings. The same was true when using global callbacks.
To streamline all apis, we've decided to make all keys Arrays only:
;-useQuery('todos', fetchTodos) +
useQuery(['todos'], fetchTodos)
;-useQuery('todos', fetchTodos) +
useQuery(['todos'], fetchTodos)
To make this migration easier, we decided to deliver a codemod.
The codemod is a best efforts attempt to help you migrate the breaking change. Please review the generated code thoroughly! Also, there are edge cases that cannot be found by the code mod, so please keep an eye on the log output.
You can easily apply it by using one (or both) of the following commands:
If you want to run it against .js or .jsx files, please use the command below:
npx jscodeshift ./path/to/src/ \
--extensions=js,jsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/key-transformation.js
npx jscodeshift ./path/to/src/ \
--extensions=js,jsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/key-transformation.js
If you want to run it against .ts or .tsx files, please use the command below:
npx jscodeshift ./path/to/src/ \
--extensions=ts,tsx \
--parser=tsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/key-transformation.js
npx jscodeshift ./path/to/src/ \
--extensions=ts,tsx \
--parser=tsx \
--transform=./node_modules/@tanstack/react-query/codemods/v4/key-transformation.js
Please note in the case of TypeScript you need to use tsx as the parser; otherwise, the codemod won't be applied properly!
Note: Applying the codemod might break your code formatting, so please don't forget to run prettier and/or eslint after you've applied the codemod!
With the introduction of the new fetchStatus for better offline support, the idle state became irrelevant, because fetchStatus: 'idle' captures the same state better. For more information, please read Why two different states.
This will mostly affect disabled queries that don't have any data yet, as those were in idle state before:
- status: 'idle'
+ status: 'loading'
+ fetchStatus: 'idle'
- status: 'idle'
+ status: 'loading'
+ fetchStatus: 'idle'
Also, have a look at the guide on dependent queries
Due to this change, disabled queries (even temporarily disabled ones) will start in loading state. To make migration easier, especially for having a good flag to know when to display a loading spinner, you can check for isInitialLoading instead of isLoading:
;-isLoading +
isInitialLoading
;-isLoading +
isInitialLoading
See also the guide on disabling queries
The useQueries hook now accepts an object with a queries prop as its input. The value of the queries prop is an array of queries (this array is identical to what was passed into useQueries in v3).
;-useQueries([
{ queryKey1, queryFn1, options1 },
{ queryKey2, queryFn2, options2 },
]) +
useQueries({
queries: [
{ queryKey1, queryFn1, options1 },
{ queryKey2, queryFn2, options2 },
],
})
;-useQueries([
{ queryKey1, queryFn1, options1 },
{ queryKey2, queryFn2, options2 },
]) +
useQueries({
queries: [
{ queryKey1, queryFn1, options1 },
{ queryKey2, queryFn2, options2 },
],
})
In order to make bailing out of updates possible by returning undefined, we had to make undefined an illegal cache value. This is in-line with other concepts of react-query, for example, returning undefined from the initialData function will also not set data.
Further, it is an easy bug to produce Promise<void> by adding logging in the queryFn:
useQuery(['key'], () =>
axios.get(url).then((result) => console.log(result.data)),
)
useQuery(['key'], () =>
axios.get(url).then((result) => console.log(result.data)),
)
This is now disallowed on type level; at runtime, undefined will be transformed to a failed Promise, which means you will get an error, which will also be logged to the console in development mode.
Please read the New Features announcement about online / offline support, and also the dedicated page about Network mode
Even though React Query is an Async State Manager that can be used for anything that produces a Promise, it is most often used for data fetching in combination with data fetching libraries. That is why, per default, queries and mutations will be paused if there is no network connection. If you want to opt-in to the previous behavior, you can globally set networkMode: offlineFirst for both queries and mutations:
new QueryClient({
defaultOptions: {
queries: {
networkMode: 'offlineFirst',
},
mutations: {
networkMode: 'offlineFirst',
},
},
})
new QueryClient({
defaultOptions: {
queries: {
networkMode: 'offlineFirst',
},
mutations: {
networkMode: 'offlineFirst',
},
},
})
The notifyOnChangeProps option no longer accepts a "tracked" value. Instead, useQuery defaults to tracking properties. All queries using notifyOnChangeProps: "tracked" should be updated by removing this option.
If you would like to bypass this in any queries to emulate the v3 default behavior of re-rendering whenever a query changes, notifyOnChangeProps now accepts an "all" value to opt-out of the default smart tracking optimization.
In v4, notifyOnChangeProps defaults to the "tracked" behavior of v3 instead of undefined. Now that "tracked" is the default behavior for v4, it no longer makes sense to include this config option.
The cancelRefetch option can be passed to all functions that imperatively fetch a query, namely:
Except for fetchNextPage and fetchPreviousPage, this flag was defaulting to false, which was inconsistent and potentially troublesome: Calling refetchQueries or invalidateQueries after a mutation might not yield the latest result if a previous slow fetch was already ongoing, because this refetch would have been skipped.
We believe that if a query is actively refetched by some code you write, it should, per default, re-start the fetch.
That is why this flag now defaults to true for all methods mentioned above. It also means that if you call refetchQueries twice in a row, without awaiting it, it will now cancel the first fetch and re-start it with the second one:
queryClient.refetchQueries({ queryKey: ['todos'] })
// this will abort the previous refetch and start a new fetch
queryClient.refetchQueries({ queryKey: ['todos'] })
queryClient.refetchQueries({ queryKey: ['todos'] })
// this will abort the previous refetch and start a new fetch
queryClient.refetchQueries({ queryKey: ['todos'] })
You can opt-out of this behaviour by explicitly passing cancelRefetch:false:
queryClient.refetchQueries({ queryKey: ['todos'] })
// this will not abort the previous refetch - it will just be ignored
queryClient.refetchQueries({ queryKey: ['todos'] }, { cancelRefetch: false })
queryClient.refetchQueries({ queryKey: ['todos'] })
// this will not abort the previous refetch - it will just be ignored
queryClient.refetchQueries({ queryKey: ['todos'] }, { cancelRefetch: false })
Note: There is no change in behaviour for automatically triggered fetches, e.g. because a query mounts or because of a window focus refetch.
A query filter is an object with certain conditions to match a query. Historically, the filter options have mostly been a combination of boolean flags. However, combining those flags can lead to impossible states. Specifically:
active?: boolean
- When set to true it will match active queries.
- When set to false it will match inactive queries.
inactive?: boolean
- When set to true it will match inactive queries.
- When set to false it will match active queries.
active?: boolean
- When set to true it will match active queries.
- When set to false it will match inactive queries.
inactive?: boolean
- When set to true it will match inactive queries.
- When set to false it will match active queries.
Those flags don't work well when used together, because they are mutually exclusive. Setting false for both flags could match all queries, judging from the description, or no queries, which doesn't make much sense.
With v4, those filters have been combined into a single filter to better show the intent:
- active?: boolean
- inactive?: boolean
+ type?: 'active' | 'inactive' | 'all'
- active?: boolean
- inactive?: boolean
+ type?: 'active' | 'inactive' | 'all'
The filter defaults to all, and you can choose to only match active or inactive queries.
queryClient.invalidateQueries had two additional, similar flags:
refetchActive: Boolean
- Defaults to true
- When set to false, queries that match the refetch predicate and are actively being rendered
via useQuery and friends will NOT be refetched in the background, and only marked as invalid.
refetchInactive: Boolean
- Defaults to false
- When set to true, queries that match the refetch predicate and are not being rendered
via useQuery and friends will be both marked as invalid and also refetched in the background
refetchActive: Boolean
- Defaults to true
- When set to false, queries that match the refetch predicate and are actively being rendered
via useQuery and friends will NOT be refetched in the background, and only marked as invalid.
refetchInactive: Boolean
- Defaults to false
- When set to true, queries that match the refetch predicate and are not being rendered
via useQuery and friends will be both marked as invalid and also refetched in the background
For the same reason, those have also been combined:
- refetchActive?: boolean
- refetchInactive?: boolean
+ refetchType?: 'active' | 'inactive' | 'all' | 'none'
- refetchActive?: boolean
- refetchInactive?: boolean
+ refetchType?: 'active' | 'inactive' | 'all' | 'none'
This flag defaults to active because refetchActive defaulted to true. This means we also need a way to tell invalidateQueries to not refetch at all, which is why a fourth option (none) is also allowed here.
This was confusing to many and also created infinite loops if setQueryData was called from within onSuccess. It was also a frequent source of error when combined with staleTime, because if data was read from the cache only, onSuccess was not called.
Similar to onError and onSettled, the onSuccess callback is now tied to a request being made. No request -> no callback.
If you want to listen to changes of the data field, you can best do this with a useEffect, where data is part of the dependency Array. Since React Query ensures stable data through structural sharing, the effect will not execute with every background refetch, but only if something within data has changed:
const { data } = useQuery({ queryKey, queryFn })
React.useEffect(() => mySideEffectHere(data), [data])
const { data } = useQuery({ queryKey, queryFn })
React.useEffect(() => mySideEffectHere(data), [data])
The plugins createWebStoragePersistor and createAsyncStoragePersistor have been renamed to createSyncStoragePersister and createAsyncStoragePersister respectively. The interface Persistor in persistQueryClient has also been renamed to Persister. Checkout this stackexchange for the motivation of this change.
Since these plugins are no longer experimental, their import paths have also been updated:
- import { persistQueryClient } from 'react-query/persistQueryClient-experimental'
- import { createWebStoragePersistor } from 'react-query/createWebStoragePersistor-experimental'
- import { createAsyncStoragePersistor } from 'react-query/createAsyncStoragePersistor-experimental'
+ import { persistQueryClient } from '@tanstack/react-query-persist-client'
+ import { createSyncStoragePersister } from '@tanstack/query-sync-storage-persister'
+ import { createAsyncStoragePersister } from '@tanstack/query-async-storage-persister'
- import { persistQueryClient } from 'react-query/persistQueryClient-experimental'
- import { createWebStoragePersistor } from 'react-query/createWebStoragePersistor-experimental'
- import { createAsyncStoragePersistor } from 'react-query/createAsyncStoragePersistor-experimental'
+ import { persistQueryClient } from '@tanstack/react-query-persist-client'
+ import { createSyncStoragePersister } from '@tanstack/query-sync-storage-persister'
+ import { createAsyncStoragePersister } from '@tanstack/query-async-storage-persister'
The old cancel method that allowed you to define a cancel function on promises, which was then used by the library to support query cancellation, has been removed. We recommend to use the newer API (introduced with v3.30.0) for query cancellation that uses the AbortController API internally and provides you with an AbortSignal instance for your query function to support query cancellation.
Types now require using TypeScript v4.1 or greater
As of v4, React Query is optimized for modern browsers. We have updated our browserslist to produce a more modern, performant and smaller bundle. You can read about the requirements here.
It was possible to change the logger globally by calling setLogger. In v4, that function is replaced with an optional field when creating a QueryClient.
- import { QueryClient, setLogger } from 'react-query';
+ import { QueryClient } from '@tanstack/react-query';
- setLogger(customLogger)
- const queryClient = new QueryClient();
+ const queryClient = new QueryClient({ logger: customLogger })
- import { QueryClient, setLogger } from 'react-query';
+ import { QueryClient } from '@tanstack/react-query';
- setLogger(customLogger)
- const queryClient = new QueryClient();
+ const queryClient = new QueryClient({ logger: customLogger })
In v3, React Query would cache query results for a default of 5 minutes, then manually garbage collect that data. This default was applied to server-side React Query as well.
This lead to high memory consumption and hanging processes waiting for this manual garbage collection to complete. In v4, by default the server-side cacheTime is now set to Infinity effectively disabling manual garbage collection (the NodeJS process will clear everything once a request is complete).
This change only impacts users of server-side React Query, such as with Next.js. If you are setting a cacheTime manually this will not impact you (although you may want to mirror behavior).
Starting with v4, react-query will no longer log errors (e.g. failed fetches) to the console in production mode, as this was confusing to many. Errors will still show up in development mode.
React Query now supports package.json "exports" and is fully compatible with Node's native resolution for both CommonJS and ESM. We don't expect this to be a breaking change for most users, but this restricts the files you can import into your project to only the entry points we officially support.
Subscribing manually to the QueryCache has always given you a QueryCacheNotifyEvent, but this was not true for the MutationCache. We have streamlined the behavior and also adapted event names accordingly.
- type: 'queryAdded'
+ type: 'added'
- type: 'queryRemoved'
+ type: 'removed'
- type: 'queryUpdated'
+ type: 'updated'
- type: 'queryAdded'
+ type: 'added'
- type: 'queryRemoved'
+ type: 'removed'
- type: 'queryUpdated'
+ type: 'updated'
The MutationCacheNotifyEvent uses the same types as the QueryCacheNotifyEvent.
Note: This is only relevant if you manually subscribe to the caches via queryCache.subscribe or mutationCache.subscribe
With version 3.22.0, hydration utilities moved into the React Query core. With v3, you could still use the old exports from react-query/hydration, but these exports have been removed with v4.
- import { dehydrate, hydrate, useHydrate, Hydrate } from 'react-query/hydration'
+ import { dehydrate, hydrate, useHydrate, Hydrate } from '@tanstack/react-query'
- import { dehydrate, hydrate, useHydrate, Hydrate } from 'react-query/hydration'
+ import { dehydrate, hydrate, useHydrate, Hydrate } from '@tanstack/react-query'
The methods cancelMutatations and executeMutation on the QueryClient were undocumented and unused internally, so we removed them. Since it was just a wrapper around a method available on the mutationCache, you can still use the functionality of executeMutation
- executeMutation<
- TData = unknown,
- TError = unknown,
- TVariables = void,
- TContext = unknown
- >(
- options: MutationOptions<TData, TError, TVariables, TContext>
- ): Promise<TData> {
- return this.mutationCache.build(this, options).execute()
- }
- executeMutation<
- TData = unknown,
- TError = unknown,
- TVariables = void,
- TContext = unknown
- >(
- options: MutationOptions<TData, TError, TVariables, TContext>
- ): Promise<TData> {
- return this.mutationCache.build(this, options).execute()
- }
Additionally, query.setDefaultOptions was removed because it was also unused. mutation.cancel was removed because it didn't actually cancel the outgoing request.
Previously, React Query had a directory named react which imported from the react module. This could cause problems with some Jest configurations, resulting in errors when running tests like:
TypeError: Cannot read property 'createContext' of undefined
TypeError: Cannot read property 'createContext' of undefined
With the renamed directory this no longer is an issue.
If you were importing anything from 'react-query/react' directly in your project (as opposed to just 'react-query'), then you need to update your imports:
- import { QueryClientProvider } from 'react-query/react';
+ import { QueryClientProvider } from '@tanstack/react-query/reactjs';
- import { QueryClientProvider } from 'react-query/react';
+ import { QueryClientProvider } from '@tanstack/react-query/reactjs';
v4 comes with an awesome set of new features:
React 18 was released earlier this year, and v4 now has first class support for it and the new concurrent features it brings.
In v3, React Query has always fired off queries and mutations, but then taken the assumption that if you want to retry it, you need to be connected to the internet. This has led to several confusing situations:
With v4, React Query introduces a new networkMode to tackle all these issues. Please read the dedicated page about the new Network mode for more information.
React Query defaults to "tracking" query properties, which should give you a nice boost in render optimization. The feature has existed since v3.6.0 and has now become the default behavior with v4.
When using the functional updater form of setQueryData, you can now bail out of the update by returning undefined. This is helpful if undefined is given to you as previousValue, which means that currently, no cached entry exists and you don't want to / cannot create one, like in the example of toggling a todo:
queryClient.setQueryData(['todo', id], (previousTodo) =>
previousTodo ? { ...previousTodo, done: true } : undefined,
)
queryClient.setQueryData(['todo', id], (previousTodo) =>
previousTodo ? { ...previousTodo, done: true } : undefined,
)
Mutations can now also be garbage collected automatically, just like queries. The default cacheTime for mutations is also set to 5 minutes.
Custom contexts can now be specified to pair hooks with their matching Provider. This is critical when there may be multiple React Query Provider instances in the component tree, and you need to ensure your hook uses the correct Provider instance.
An example:
// Our first data package: @my-scope/container-data
const context = React.createContext<QueryClient | undefined>(undefined)
const queryClient = new QueryClient()
export const useUser = () => {
return useQuery(USER_KEY, USER_FETCHER, {
context,
})
}
export const ContainerDataProvider = ({
children,
}: {
children: React.ReactNode
}) => {
return (
<QueryClientProvider client={queryClient} context={context}>
{children}
</QueryClientProvider>
)
}
// Our first data package: @my-scope/container-data
const context = React.createContext<QueryClient | undefined>(undefined)
const queryClient = new QueryClient()
export const useUser = () => {
return useQuery(USER_KEY, USER_FETCHER, {
context,
})
}
export const ContainerDataProvider = ({
children,
}: {
children: React.ReactNode
}) => {
return (
<QueryClientProvider client={queryClient} context={context}>
{children}
</QueryClientProvider>
)
}
// Our second data package: @my-scope/my-component-data
const context = React.createContext<QueryClient | undefined>(undefined)
const queryClient = new QueryClient()
export const useItems = () => {
return useQuery(ITEMS_KEY, ITEMS_FETCHER, {
context,
})
}
export const MyComponentDataProvider = ({
children,
}: {
children: React.ReactNode
}) => {
return (
<QueryClientProvider client={queryClient} context={context}>
{children}
</QueryClientProvider>
)
}
// Our second data package: @my-scope/my-component-data
const context = React.createContext<QueryClient | undefined>(undefined)
const queryClient = new QueryClient()
export const useItems = () => {
return useQuery(ITEMS_KEY, ITEMS_FETCHER, {
context,
})
}
export const MyComponentDataProvider = ({
children,
}: {
children: React.ReactNode
}) => {
return (
<QueryClientProvider client={queryClient} context={context}>
{children}
</QueryClientProvider>
)
}
// Our application
import { ContainerDataProvider, useUser } from "@my-scope/container-data";
import { AppDataProvider } from "@my-scope/app-data";
import { MyComponentDataProvider, useItems } from "@my-scope/my-component-data";
<ContainerDataProvider> // <-- Provides container data (like "user") using its own React Query provider
...
<AppDataProvider> // <-- Provides app data using its own React Query provider (unused in this example)
...
<MyComponentDataProvider> // <-- Provides component data (like "items") using its own React Query provider
<MyComponent />
</MyComponentDataProvider>
...
</AppDataProvider>
...
</ContainerDataProvider>
// Example of hooks provided by the "DataProvider" components above:
const MyComponent = () => {
const user = useUser() // <-- Uses the context specified in ContainerDataProvider.
const items = useItems() // <-- Uses the context specified in MyComponentDataProvider
...
}
// Our application
import { ContainerDataProvider, useUser } from "@my-scope/container-data";
import { AppDataProvider } from "@my-scope/app-data";
import { MyComponentDataProvider, useItems } from "@my-scope/my-component-data";
<ContainerDataProvider> // <-- Provides container data (like "user") using its own React Query provider
...
<AppDataProvider> // <-- Provides app data using its own React Query provider (unused in this example)
...
<MyComponentDataProvider> // <-- Provides component data (like "items") using its own React Query provider
<MyComponent />
</MyComponentDataProvider>
...
</AppDataProvider>
...
</ContainerDataProvider>
// Example of hooks provided by the "DataProvider" components above:
const MyComponent = () => {
const user = useUser() // <-- Uses the context specified in ContainerDataProvider.
const items = useItems() // <-- Uses the context specified in MyComponentDataProvider
...
}