NoSleep Javascript

Stop using Redux for Data Fetching, use this instead

November 10, 2020 • 10 min read
#TypeScript#Javascript#React#react-query#redux#redux-toolkit#data fetching#reactive data flow

For the longest time it has been an idiom in React applications to perform data fetching through Redux, this approach has always have problems in its efficiency and effectiveness but modern tools and techniques have showed us better ways. Let’s together explore what better alternatives the ecosystem has to offer, but first let’s quickly cover what’s Redux and how it is typically used to fetch data.

Redux is a global state management system mostly associated with React applications that is great for managing global app state and the states transitions in a predictable way but when dealing with application data fetching requirements it can become a problem:

The typical data fetching in Redux

This is a very simplistic way to data fetch to Redux, it is probably not production ready and has unresolved problems.

post.action.ts
import {
  LOADING_START,
  LOADING_END,
  FETCH_ERROR,
  UPDATE,
} from "./post.reducer";

export function loadingStart() {
  return { type: LOADING_START };
}

export function loadingEnd() {
  return { type: LOADING_END };
}

export function fetchError(error: string) {
  return { type: FETCH_ERROR, payload: string };
}

export function update(post: IPost) {
  return { type: FETCH_ERROR, payload: post };
}

// this requires redux-thunk
export function getPost(postId: number) {
  return async (dispatch) => {
    dispatch(loadingStart());
    try {
      const post = await api.getPost(postId);
      dispatch(update(post));
    } catch (err) {
      dispatch(fetchError(err.toString()));
    } finally {
      dispatch(loadingEnd());
    }
  };
}
post.reducer.ts
export const LOADING_START = "posts/LOADING_START";
export const LOADING_END = "posts/LOADING_END";
export const FETCH_ERROR = "posts/FETCH_ERROR";
export const UPDATE = "posts/UPDATE";

export const initialState = {
  loading: false,
  error: "",
  data: undefined,
};

export interface IState {
  loading: boolean;
  error: string;
  data: IPost | undefined;
}

export default function postReducer(
  prevState: IPostState = initialState,
  action: IAction<any>
): IPostState {
  switch (action.type) {
    case LOADING_START: {
      return { ...prevState, loading: true };
    }
    case LOADING_END: {
      return { ...prevState, loading: false };
    }

    case FETCH_ERROR: {
      const error = action.payload;
      return { ...prevState, error };
    }

    case UPDATE: {
      const post = action.payload as IPost;
      const postId = post.id;

      return { ...prevState, data: post };
    }

    default: {
      return prevState;
    }
  }
}
root.reducer.ts
import post from "./post.reducer";

export default combineReducers({ post });

And finally you can use it

export class MyComponent() {
  // fetch data when the component mounts
  onComponentDidMount() {
    this.props.getPost(this.props.postId)
  }
}

// or
export function MyComponent({postId, getPost, post, loading, error}) {
  useEffect(() => {
    getPost()
  }, [])
}

function mapStateToProps(appState: IAppState) {
  // reading the result of the data being fetched,
  // but who asked for it? do I need to do it?
  // did someone already did it?
  const { data: post, loading, error } = appState.post;
  return { post, loading, error };
}

function mapDispatchToProps() {
  // this will be called and will trigger the http call,
  // let's hope that `MyComponent` does it and does it correctly,
  // i.e. in onComponentDidMount AND onComponentDidUpdate
  return { getPost: actions.getPost };
}

export default connect(
  mapStateToProps,
  mapDispatchToProps
)(MyComponent);

Data fetching with Redux is too complicated

The current implementation has the following problems

  • it is way too verbose (hooks and libraries might make it better)
  • is that the right semantics for my actions?
  • is that the right structure and naming for my actions?
  • is that the right file structure? Official Redux docs once upon a time opposed to the “ducks” approach but now they seem to be strongly recommending it
  • should I use switch or some other way of casing?
  • should I normalize?
  • is the state shape correct?
  • are my meta flags correct?
  • how would I manage multiple posts being fetched at the same time? (hint: you need to change your state shape and do more stuff)
  • should I use immer (which I love BTW) or similar?
  • should I use redux-toolkit?
  • should I use sagas? (which is even more verbose and complex BTW).
  • where are the selectors?
  • how do I make my team code consistently across the multiple action/reducers that we will need to write?
  • what happens if more than one component triggers the same data fetching action? Do I need to dedup http calls manually?
  • do I need to make all my data fetching actions at the root of my app? (in big enough apps this causes a bad decoupling between the data requirements of your components and your components).
  • what about if my app is too big and I want to more granularly define what data fetching requirements a medium level component needs to have?
  • when does this fetched data expires? how do I clean it up?

All these are somewhat open problems that you and your team need to fix in the life cycle of your app development. It is really hard to structure this 100% correctly.

And finally, these things illustrate the crazy amount of decisions, considerations and code a developer needs to make in order to fetch a simple post.

Simple and common tasks should have simple solutions.

The usage problem

Let’s say we already figured out how to do all the boilerplate, we still have to deal with the usage problem: the decoupling of the triggering of http calls (the action) and the reading of the result (the state) as seen in above.

What you really are trying to express is something like the following, notice how in a single place we have the triggering of the http side effect and the reading of the result and the meta data.

const { result, loading, error } = getPost();
// UI

Decoupling things that are naturally tightly coupled is as bad as tightly coupling things that shouldn’t be.

By using redux hooks it can, partially, get better.

Redux forces both data fetching and mutations to be imperative

Data fetching (when you retrieve data) is usually reactive in a React applications, meaning that a prop will change, a component will render and an associated side effect of fetching new data should be triggered. Mutations, like saving a form are typically imperative. Redux forces both data fetching and mutations to be imperative.

The library problem

Any library that relies on Redux makes its integration surface bigger. Lib consumers now need setup Redux and set it up the way the library expects it to be with the appropriate reducers and middlewares it needs to work.

The alternative that we will explore are much more straight forward to encapsulate at the library level.

What is data fetching?

Let’s take a step back and analyse what we really want of our data fetching layer:

  • dedup http calls; any “smart” component that needs data to be fetched should be able to declare it without fearing it might cause duplicated http calls.
  • meta flags: loading, error, last fetched, etc.
  • updating or invalidating the state when necessary, example:

    • when data is too old
    • when you mutate the data in the server and you need to reflect that
  • reactivity (when a prop changes new data will be fetched if necessary).
  • medium term response storing, reading and writing or updating.
  • heavy abstraction, simplicity and a really easy to use interface.

These things have been a staple of GraphQL clients such as Apollo for quite some time now.

Just a cache

When you think about it what you really want is a cache layer for your http requests that will simply maintain a key value store for your data, where the key will identify the resource and the value will be the resource plus its metadata:

export const cache = {
  'posts/124': {
    data: {/* the post */}
    meta: {loading, error, lastFetched, keepAlive, etc}
  }
  'posts/456': {
    data: {/* the post */}
    meta: {loading, error, lastFetched, keepAlive, etc}
  }
  'users/barry@white.com': {
    data: {/* the user */}
    meta: {loading, error, lastFetched, keepAlive, etc}
  }
}

And this data structure will have the appropriate functionality on top of it to:

  • fetch the data and store it the right key
  • update other cache keys after a successful data fetch (i.e. after fetching multiple posts we update the individual post cache)
  • update cache after a mutation (i.e. saving a post form)
  • cache time: how long I want to keep this data?
  • stale time: how long I will consider this data fresh?
  • automatically managing meta flags.

Forget about

  • state shape (it is just a key value store of responses)
  • no normalization (it is just a cache)
  • data fetching meta state like loading, error, etc
  • actions
  • reducers
  • thunks
  • imperative data fetching onComponentDidMount

Do this instead!

Enter react-query.

There is one alternative built on top of Redux called redux-query. I’ve used it extensively and although it is an improvement over pure Redux, some of the core problems that we covered remained.

react-query is a really nice new library that builds on the things we have previously discussed, solves all the problems, is relatively simple and easy to understand and will reduce the amount of code you need to write for data fetching by an absurd amount, all the previous code gets replaced by this:

export function useGetPost(postId): QueryResult<IPost> {
  return userQuery<IPost>({
    queryKey: `posts/${postId}`,
    queryFn: api.getPost,
  });
}

And its usage

export default function MyComponent({ postId }) {
  const { data: post, isLoading, error } = useGetPost(
    postId
  );
  //...
}

This covers what we did in Redux and more, as you can see it requires a considerable less amount of code.

this is production ready!

redux-query deals with a cache of responses instead of the Redux state, but can do most of the same things if necessary i.e. updating the state, it just have a different semantic and a much less verbose interface.

There are some new problems with react-query that mostly rise from its reactive interface like dependent queries et al but if you know how to deal with react you will know how to deal with that, and in my production experience they tend to be the exception and not the rule.

TLDR: try react-query

The programming world is an ever evolving technical landscape where the community grows and learns new tips and tricks from experience, Redux has a historical significance and it is still suitable for certain scenarios but modern tools have provided better ways of doing some of the things we typically used to do with Redux, data fetching is one of them.

This new data fetching paradigm is mostly possible because of this crazy idea the react team had: react-hooks. I agree they are weird but the new things we can build with them and the composition models they enable is really great, and I expect more awesome and unexpected tools like react-query to come in the near future as we develop our understanding of hooks.

Enjoyed the ride? Consider subscribing to our awesome newsletter and become part of the family and consider donating and supporting my work.

Cheers!

Subscribe to our mailing list!

Get exclusive content, offers and become a friend!


Written by franleplant: Tech Lead and Software developer with a degree in Engineering. Woodworker, all things Javascript, Rust and Software Architecture. Github

© 2021, nosleepjavascript.com