normalized-react-query
v1.0.6
Published
Wrapper around React Query to enforce type-safe, consistent key-query mappings.
Downloads
308
Maintainers
Readme
>normalized-react-query
Wrapper around React Query to enforce type-safe, consistent key-query mappings.
Contents
Introduction
React Query provides powerful API state management, with caching, pre-fetching, SSR, and hook support.
The main idea is pairing "keys" (unique to a specific API call + params) with a query function. React Query handles actual pairing of async logic with synchronous hooks/renders behind the scenes.
The "problem" with manual pairing of keys to functions is that there is no way to ensure the exact same function is paired to the same key, or that the same key is re-used for similar queries. As a result, either caching may prevent the desired function from being called, or may accidentally call the same function multiple times.
Take these example hooks, which represent inconsistent usage of keys + query handlers:
import { useQuery } from '@tanstack/react-query';
import { getUsers } from './api/users.js';
const useExample = () => {
const firstQuery = useQuery(
['users', 'get'],
async () => {
return getPosts();
};
);
const secondQuery = useQuery(
['users', 'get'],
async () => {
// NEVER RUNS
// Uses cached value of `firstQuery`.
// Result is typed to include `{ decorate: boolean }` but that will never exist.
const users = await getUsers();
return users.map(user => {
...user,
decorate: true,
};
};
);
const thirdQuery = useQuery(
// Different key, same API call.
// Triggers another fetch for data that already exists.
['users', 'fetch'],
async () => {
return getUsers();
};
);
};
This package attempts to solve this discrepancy by forcing pairing of keys and functions, with strongly typed data.
Ideally any implementations of the wrapper classes can offload any API and business logic so that actual React functions can simply access the data as it becomes available.
Install
npm i normalized-react-query
Example
import type { QueryKey } from '@tanstack/react-query';
import { useState } from 'react';
import {
Paginated,
type QueryData,
type QueryParams,
Resource,
} from 'normalized-react-query';
import { getUsers, listUsers } from './api/users.js';
class FetchUser extends Resource<User, UserId> {
protected getKey(id: QueryParams<this>): QueryKey {
return ['users', 'get', id];
}
protected async queryFn(id: QueryParams<this>): Promise<QueryData<this>> {
return getUser(id)
}
}
const fetchUser = new FetchUser();
class ListUsers extends Paginated<User> {
protected getKey(): QueryKey {
return ['posts', 'get'];
}
protected async queryFn(): Promise<QueryData<this>> {
return listUsers();
}
protected async onSuccess(
client: QueryClient,
params: QueryParams<this>,
data: QueryData<this>
): void {
for (const user of data) {
// Pre-populate lookup data of users by-id.
fetchUser.setData(client, user.id, user);
}
}
}
const listUsers = new ListUsers();
const useExample = () => {
const firstQuery = fetchUser.useQuery(123);
// Successfully cached
const secondQuery = fetchUser.useQuery(123);
// Separate query, consistent behavior
const thirdQuery = fetchUser.useQuery(456);
// Pre-populates every user
const listQuery = listUsers.useQuery();
const [showFourth, setShowFourth] = useState(false);
useEffect(() => {
setTimeout(() => setShowFourth(true), 5000);
});
// Will be cached immediately on load, due to "listUsers" pre-population.
const fourthQuery = fetchUser.useQuery(789, {
// All normal React-Query options are available
enabled: showFourth,
});
};
Usage
normalized-react-query
is an ESM module. That means it must be import
ed. To load from a CJS module, use dynamic import const { Resource } = await import('normalized-react-query');
.
Both the react and react query modules are required for this package to work. Peer dependencies are declared on both. This package merely enforces typing and structure, any caching/revalidation/subscription is still implemented by native React Query.
Class interfaces are used to best expose inheritance and override functionality for typescript. The actual query instances used should be singletons (e.g. create once and re-use).
A lowercase version of each class is available in favor of functional programming practices.
e.g.
import { resource, Resource } from 'normalized-react-query';
import { getUser, type User } from './api.js';
class GetUserClass extends Resource<User, string> {
getKey(id) {
return ['users', id];
}
queryFn(id) {
return getUser(id);
}
onError(client, id, error) {
console.error(`Failed to lookup user ${id}`, error);
}
}
const getUserClass = new GetUserClass();
// Logically same as `getUserClass`.
const getUser = resource<User, string>(
{
getKey(id) {
return ['users', id];
}
queryFn(id) {
return getUser(id);
}
},
{
onError(client, id, error) {
console.error(`Failed to lookup user ${id}`, error);
}
}
);
API
Resource
The Resource
class is the most basic wrapper around useQuery
. When in doubt, a method that asynchronously loads data (a "resource") should extend the Resource
class.
Each child class defines the parameters to the function, and generates a unique queryKey
for those params. It also provides some basic abstractions to support other React Query functionality such as pre-fetching data and invalidation.
The query function can be as simple as a direct API call, but supports side effects as necessary. A common side effect may be to take a sub-resource of the response, and pre-load another query.
These side-effects should most likely be placed in lifecycle hooks, like onSuccess
.
Paginated
The Paginated
class is a typed extension Resource
. It provides some default typing to support pagination, which React Query supports natively with keepPreviousData.
Infinite
The Infinite
class wraps the Paginated
class further, providing "infinite" queries where all pages are loaded and available in parallel. It is built off multiple individual queries, so any refetching, caching, invalidation, and de-duplication is handled smoothy.
Why not native useInfinite
?
React Query exports a useInfinite
natively, which provides very similar behavior out of the box. The decision to not use it is based on that hook conflicting with useQuery
's cache. useInfinite
queries cannot share a queryKey
with useQuery
, and therefore cannot benefit from the powerful functionality React Query defines.
Therefore, a custom useInfinite
hook was implemented using React Query's useQueries
hook. That ensures any using of Infinite
s useQuery
hook can benefit from useInfinite
and vice versa. Remember Infinite
is a child class of Paginated
and therefore supports normal query behavior as well.
Mutation
The Mutation
class is a wrapper around React Query useMutation
. "Mutations" aren't fetching data, but rather changing the server state, and handling side effects accordingly.
Most common mutations are POST and PATCH endpoints, that most likely impact a related GET endpoint. Therefore the "side effects" should either set the new state data (if returned from the mutation) or invalidate the existing query and force it to refetch.
Mutations are not "cached" in the same sense useQuery
is, but do generally benefit from the pattern of consistent mutation handling and typings.
Types
A few types are exported for convenience of accessing the generic parameters of a resource.
All accept the this
instance of a child class.
QueryData
Access the returned data type of useQuery
.
QueryParams
Access the parameter data type to useQuery
.
QueryVariables
Unique to Mutation
, access the variables interface provided to mutation execution.
EmptyObject
Represents {}
type, with no keys. Can be passed as parameter to Paginated
/Infinite
generics when there are no other parameters to provide.