Project Metamorphosis: Unveiling the next-gen event streaming platform. Learn More

Building the Confluent UI with React Hooks – Benefits and Lessons Learned

Updating a fundamental paradigm in your React app can be as easy as search and replace, or at other times, as difficult as convincing your entire frontend engineering to buy into it. Migrating to React Hooks may fall into the latter, but with numerous benefits, an improved user interface, and easier maintainability at scale, it is well worth the effort.

The Confluent Cloud UI is built using React Redux. As the scale of our web application and the number of developers grow, it becomes increasingly imperative to focus on three key issues:

  • Code readability
  • Code reusability
  • Test coverage for both unit tests and integration tests

Topics messages

To address these issues, we migrated our React components to utilize React Hooks. This blog post provides insight into the benefits of React Hooks and presents examples that explain how we refactored traditional React containers and components to React Hooks in the Confluent UI. Along the way, our team learned some lessons, such as how to properly apply memoization when using Hooks. This blog post dives deep into the benefits and lessons learned from using Hooks in our frontend codebase.

Benefits

Using React Hooks provides three benefits: reusability, readability, and testability.

Reusability

One benefit of React Hooks is that you can decouple unrelated logic code from lifecycle methods. For example, consider a component called “Updating Clock and Resize” that does just two things:

  1. Updates a clock every second
  2. Adds a listener to the resize event and displays the updated content

The following is a screenshot of this fictional component:

Fictional component

Below is an implementation of the class component:

 
import React from "react";
 
class Component extends React.Component {
 constructor(props) {
   super(props);
   this.tickId = null;
   this.resizeListener = null;
   this.state = {
     time: new Date().toLocaleTimeString(),
     windowWidth: window.innerWidth
   };
 }
 
 componentDidMount() {
   this.tickId = setInterval(() => {
     this.setState({ time: new Date().toLocaleTimeString() });
   }, 1000);
   this.resizeListener = () => {
     this.setState({ windowWidth: window.innerWidth });
   };
   window.addEventListener("resize", this.resizeListener);
 }
 
 componentWillUnmount() {
   clearInterval(this.tickId);
   window.removeEventListener("resize", this.resizeListener);
 }
 
 render() {
   return (
     <div>
       <div>Time {this.state.time}</div>
       <div>The window width is {this.state.windowWidth}</div>
     </div>
   );
 }
}
export default Component;

One downside to this approach is that you mix both setInterval and addEventListener in your componentDidMount. You generally want to avoid building out apps in lifecycle methods (i.e., componentDidMount, componentDidUpdate, and componentWillUnmount) because your code becomes more tightly coupled and less scalable as your logic gets more complex—and not to mention more difficult to understand.

Fortunately, you can separate the setInterval and addEventListener into two useEffect Hooks after refactoring the component. Moreover, if you would like to reuse the setInterval and addEventListener in different components, you can define two custom Hooks, such as useInterval and useEventListener.

Before you put together these custom Hooks, take a look at a Hook called useEventCallback from Formik. useEventCallback utilizes useRef to store the updated callback function into a mutable ref and returns a memoized callback. In other words, you want to ensure that the reference to your savedCallback never changes, as it will be used as a dependency for your useEffect. You can read more about useEffect and dependency array in the article Using the Effect Hook.

// from formik’s useEventCallback
// that stores the callback in a mutable Hook when fn is updating
// and the callback's reference address will not be changed
function useEventCallback(fn) {
 const ref = React.useRef(fn);
 useEffect(() => {
   ref.current = fn;
 }, [fn]);
 return React.useCallback((...args) => ref.current.apply(void 0, args), []);
}

Then, extract the logic for useInterval and useEventListener:

function useInterval (fn, delay){
 const savedCallback = useEventCallback(fn);
 useEffect(() => {
   if (delay != null) {
     const tickId = setInterval(savedCallback, delay);
     return () => {
       clearInterval(tickId);
     };
   }
 }, [delay, savedCallback]);
};
 
function useEventListener(event, fn) {
 const savedCallback = useEventCallback(fn);
 useEffect(() => {
   window.addEventListener(event, savedCallback);
   return () => {
     window.removeEventListener(event, savedCallback);
   };
 }, [event, savedCallback]);
}

Now, it’s time to directly reuse useInterval and useEventListener in the original implementation (see the source code for reference):

const getCurrentLocalTimeString = () => new Date().toLocaleTimeString();
const getWindowWidth = () => window.innerWidth;

export default function Component(props) {
  const [time, setTime] = useState(getCurrentLocalTimeString);
  const [windowWidth, setWindowSize] = useState(getWindowWidth);

  useInterval(() => {
    setTime(getCurrentLocalTimeString());
  }, 1000);

  useEventListener("resize", () => {
    setWindowSize(getWindowWidth());
  });

  return (
    <div>
      <div>Time (time)</dev>
      <div>The window width is (windowWidth)</div>
    </div>
  );
}

With React Hooks reusing component logic, you can now see the logic behind setInterval and setEventListener from the component, which immediately makes the code cleaner. In addition, two new useful custom Hooks are created that enhance the reusability of the codebase.

Readability

Like most React apps at the time (prior to React 16.7), the Confluent UI was divided into two categories: containers and presentational components, as suggested by Dan Abramov. A container was typically composed with a variety of higher-order components (HOCs) to reuse the logic and inject data. The presentational components received the data and callbacks via props passed from the containers.

Two common issues with using HOCs include the use of render props and the potential to have components surrounded by layers of providers, consumers, higher-order components, render props, and other abstractions—otherwise known as wrapper hell. The following uses the topic edit form in the Confluent UI as an example of wrapper hell.

Topic edit form

The form container has the following logic:

  1. Inject the API endpoint and token from the Redux store
  2. Send a request using the API endpoint and token from step 1
  3. Inject the API response from the Redux store
  4. Inject and initialize a Formik form using the response from API
  5. Conditionally render the component if the API is fetching or failed
const FormContainer = compose(
  connectTokenAndApiEndpoint,           // get token/endpoint from redux store
  connectRequest(mapPropsToRequestFn1), // send query to api
  connect(mapStateToPropsFn1),          // inject redux’s data to component
  withFormik(formikProps),              // inject formik props 
  conditionalComponentLoader,           // render loading component
)(FormComponent);

Nested wrappers

The code produces several nested wrappers in React, as shown in the figure above. Developers new to React may find it difficult to understand the purpose of each nested wrapper. Moreover, it is also difficult to figure out the data relationship between containers and presentational components.

Below is an alternative look at the same component using React Hooks:

import React from "react";
import { useSelector } from "react-redux";
import { useQuery } from "react-query";
import { useFormik } from "formik";
  
function Form(props) {
 const apiEndpoint= useSelector((state) => state.apiEndpoint);
 const token = useSelector((state) => state.token);
 const { isLoading, error, data } = useQuery(
   "getFormData",
   getFormData({ apiEndpoint, token, ... })  //inject data from props+redux
 );
 const formikProps = useFormik({
   initialValues: { ...data, ... },          //inject data from props
   onSubmit: (values) => {}
 });
 if (isLoading) {
   return "loading";
 }
 if (error != null) {
   return "error";
 }
 return <form onSubmit={formikProps.handleSubmit}>...</form>;
}

export default Form;

The following figure shows the mapping between HOCs and Hooks. Compared to the HOCs solution, the Hooks approach provides cleaner logic and a better way to understand the relationship between the data and component.

Mapping between HOCs and Hooks

Now the data flow and logic is cleaner:

  1. Get apiEndpoint and token from Redux’s store by useSelector
  2. Send an API request with apiEndpoint and token from step 1
  3. Get the response data from the API request and pass the response data as initial values for the form
  4. Conditionally render the component if the API is fetching or returned a failed response

As a result, you may consider using Hooks to replace old HOCs and render props to improve the readability of your React application.

Testability

A common testing pattern in React is to test container and presentational components separately. This works for unit tests. However, if one container is composed of several HOCs, you will be required to test the container and components together in order to do the integration tests, which is generally more cumbersome.

After applying Hooks in a React app, containers and a component can be combined into one component. Thanks to the react-testing-library, it is easier to test the component with multiple custom Hooks. Furthermore, it makes it easier to write separate unit tests for customized Hooks.

For example, let’s test the following Hook useToggleSelectionSet (see source code for reference) with a component TodoList. The TodoList contains two props:

  • items: the whole items of TodoList
  • preSelections: the items will be marked as done when added to the list

Items in TodoList

TodoList:

import { useToggleSelectionSet } from "./useToggleSelectionSet";

export const TodoList = ({ items, preSelections }) => {
  const [selectionSet, toggleSelectionItem] = useToggleSelectionSet(
    preSelections
  );
  return (
    <ul>
      {items.map((item) => {
        return (
          <li key={item}>
            <input
              data-testid={`${item}-checkbox`}
              checked={selectionSet.has(item)}
              onChange={(event) =>
                toggleSelectionItem(item, event.target.checked)
              }
              type="checkbox"
            ></input>
            <span>{item}</span>
          </li>
        );
      })}
    </ul>
  );
};

useToggleSelectionSet:

import { useCallback, useEffect, useState } from "react";

export function useToggleSelectionSet(preSelections) {
  const [selectionSet, setSelectionSet] = useState(
    () => new Set(preSelections)
  );
  // toggle an item between selected and unselected
  const toggleSelectionItem = useCallback((item, isChecked) => {
    setSelectionSet((prevSet) => {
      const newSet = new Set(prevSet);
      if (isChecked) {
        newSet.add(item);
      } else {
        newSet.delete(item);
      }
      return newSet;
    });
  }, []);

  // keep the old selections and appending new preSelections
  useEffect(() => {
    setSelectionSet((prevSelectionSet) => {
      const newSelections = preSelections.filter(
        (item) => !prevSelectionSet.has(item)
      );
      if (newSelections.length > 0) {
        return new Set([...prevSelectionSet, ...newSelections]);
      }
      return prevSelectionSet;
    });
  }, [preSelections]);

  return [selectionSet, toggleSelectionItem];
}

With the react-testing-library’s help, it is easy to test a component during the rerendering process. For example, the TodoList component test first renders with the items “Apple” and “Banana,” with “Apple” selected.

Items in TodoList

Here is the corresponding code:

    const { getByTestId, rerender } = render(
      <TodoList items={["Apple", "Banana"]} preSelections={["Apple"]} />
    );

You can also check whether “Apple” is selected and whether “Banana” is not selected:

    // Apple is pre selected
    expect(getByTestId("Apple-checkbox")).toHaveProperty("checked", true);
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", false);

Now imagine you’ve added “Cat” to the list, so the items are now “Apple,” “Banana,” and “Cat.” And you would like to select the newly added item “Cat.”

Added "Cat" to the TodoList

You can simulate this using rerender with your updated props and check whether “Apple” is still selected.

    rerender(
      <TodoList items={["Apple", "Banana", "Cat"]} preSelections={["Cat"]} />
    );

    // Apple's selection is still there
    expect(getByTestId("Apple-checkbox")).toHaveProperty("checked", true);

In addition, you can use fireEvent to test the checkbox’s functionality:

    // check toggle click
    fireEvent.click(getByTestId("Banana-checkbox"));
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", true);
    fireEvent.click(getByTestId("Banana-checkbox"));
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", false);

The resulting unit test would look like this:

import { render, fireEvent } from "@testing-library/react";
import { TodoList } from "../TodoList";

describe("TodoList", () => {
  it("TodoList keep the old list selections when there are new items", () => {
    const { getByTestId, rerender } = render(
      <TodoList items={["Apple", "Banana"]} preSelections={["Apple"]} />
    );
    // Apple is pre selected
    expect(getByTestId("Apple-checkbox")).toHaveProperty("checked", true);
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", false);

    rerender(
      <TodoList items={["Apple", "Banana", "Cat"]} preSelections={["Cat"]} />
    );

    // Apple's selection is still there
    expect(getByTestId("Apple-checkbox")).toHaveProperty("checked", true);
    expect(getByTestId("Cat-checkbox")).toHaveProperty("checked", true);

    // check toggle click
    fireEvent.click(getByTestId("Banana-checkbox"));
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", true);
    fireEvent.click(getByTestId("Banana-checkbox"));
    expect(getByTestId("Banana-checkbox")).toHaveProperty("checked", false);
  });
});

Lastly, it is cleaner and more intuitive to test the useToggleSelectionSet Hook by itself with the React Hooks Testing Library than to create a full integration test with TodoList.

import { renderHook, act } from "@testing-library/react-hooks/dom/pure";
import { useToggleSelectionSet } from "../useToggleSelectionSet";

describe("useToggleSelectionSet", () => {
  it("useToggleSelectionSet keep the old selections when there are new preSelections", () => {
    const hook = renderHook(
      (preSelections) => useToggleSelectionSet(preSelections),
      { initialProps: ["Apple"] }
    );
    let [selectionSet, handleToggleItem] = hook.result.current;
    // Apple is pre selected
    expect(selectionSet).toMatchObject(new Set(["Apple"]));

    // check toggle
    act(() => {
      handleToggleItem("Banana", true);
    });
    [selectionSet] = hook.result.current;
    expect(selectionSet).toMatchObject(new Set(["Apple", "Banana"]));

    // rerender with new props
    hook.rerender(["Cat"]);
    [selectionSet] = hook.result.current;
    // Apple's selection is still there
    expect(selectionSet).toMatchObject(new Set(["Apple", "Banana", "Cat"]));
  });
});

Everything looks good! Both the component and Hook tests have passed.

Component and Hook tests have passed

Another vital part of React Hook testing is the capability of mocking out modules. Mocking Hooks is easier than mocking HOCs because the former is similar to mocking a function.

More lessons learned

While React Hooks solve wrapper hell well and improve the readability and reusability significantly in the codebase, it’s not a silver bullet. This section shares a couple of pitfalls that we encountered while migrating to React Hooks due to unmemoized code.

Pitfall 1: Unnecessary rerendering

It’s a challenge to figure out the unnecessary rerendering when migrating to Hooks. For example, you may migrate React Redux’s connect HOC to the useSelector Hook. With the connect HOC, all fields are combined into an object by the mapStateToProps function. The connected component will not rerender if all fields in the object are the same as the previous one. For example, the ConnectedComponent will not be rendered if the age and name are not changed.

function mapStateToProps(state, props) {
 return {
   age: state.age,
   name: state.name
 };
}
const ConnectedComponent = connect(mapStateToProps)(MyComponent);

However, if you replace connect with useSelector, there will be unnecessary rerendering.

function MyComponent(state, props) {
 const mappedProps = useSelector((state) => ({
   age: state.age,
   name: state.name
 }));
 ...
}

In this example, the mappedProps object is recreated whenever an action is dispatched to the Redux store, so the mappedProps is not the same as the previous one, which leads to unnecessary rerendering. Passing shallowEqual to useSelector, or calling useSelector multiple times separately, with each returning an individual value, may solve this problem.

Additionally, useSelector is a Hook, not an HOC, so it cannot wrap a component with React.PureComponent as connect() did. Therefore, even if a component’s props and all returned data from useSelector are not changed, the component may still be rerendered unnecessarily.

To solve this problem, you may consider using React.memo to optimize the performance. Moreover, useMemo and useCallback are also handful Hooks for memoization.

Pitfall 2: Infinite rerendering

When migrating from class components to Hooks, it is common practice to replace the componentDidMount method with a useEffect. However, if a developer includes a callback as a dependency of useEffect, and the callback contains setState from its parent’s component, you may encounter an infinite rerendering issue.

Consider the following class component with a prop onMount. The onMount method will only be invoked once when the component mounts.

class MyComponent extends Component {
  componentDidMount() {
    const onMount = this.props.onMount;
    onMount && onMount();
  }
  ...
}

In its parent component, the onMount method is created with an inline function that includes a setState method.

<MyComponent
onMount={()=>{
setState(state => {...})
}}
...
/>

When refactoring to React Hooks, the code looks like this:

function MyComponent (props) {
  const {onMount} = props;
  useEffect(()=> {
    onMount();
  }, []); 
 ...
}

If the onMount prop changes, you want to fire the new event, so we added onMount to the dependency array of useEffect.

useEffect(()=> {
   onMount();
// added a depen
}, [onMount]);

However, the onMount is created by an inline function, which is never the same as the previous onMount during rerendering. Unfortunately, we forget to memoize this function properly. Moreover, onMount includes setState from the parent component, which will trigger a rerender from a parent component. Thus, it leads to an infinite rerendering process displayed in the figure below. Eventually, the page will crash.

onMount infinite rendering process

There are two ways to solve this bug:

  1. Wrap onMount with useCallback from the parent component.
    const onMount = useCallback(()=>{setState(state=>{...})} ,[]);

    Now onMount is memoized since the dependency array of useCallback is empty. Therefore, the onMount will only be invoked once, which will not lead to additional rerendering. However, the useCallback approach requires all parent components to wrap the onMount with useCallback. To solve this problem, there is a second option.

  2. Wrap onMount with useEventCallback.If MyComponent is shared with multiple components and you forget to wrap one of the components’s onMount with useCallback, the bug still exists. In that case, you should solve the problem in your child component.As mentioned earlier in the “Updating Clock” example, the output of useEventCallback will be memoized. You may wrap the onMount with useEventCallback. Since memodOnMount is memoized, the memodOnMount in useEffect will not execute infinitely.
    function MyComponent (props) {
      const {onMount} = props;
      const memodOnMount = useEventCallback(onMount);
      useEffect(()=> {
        memodOnMount();
      }, [memodOnMount]); 
      ...
    }
    

    Hooks without memoization will result in these issues. Thus, when using React Hooks, attention must be paid to memoizing and especially to the variable that is part of the dependency array.

Pitfall 3: Side effects in other components during rendering

Sometimes it is possible to forget to wrap useEffect when calling setState or dispatch from Redux, for example:

function MyForm(props) {
 const { setOpenModal } = props;
 const [{ success }, submit] = useFormSubmit();
 const dispatch = useDispatch();
// We should not call dispatch or setState during render 
 if (success) {
   dispatch(push("/fakepath"));
   setOpenModal(false);
 }
 return <form onSubmit={(values) => submit(getFakeQuery(values))}>...</form>;
}

React throws warnings for some updates while rendering:

Warning: Cannot update a component from inside the function body of a different component.

Wrapping dispatch/setState with useEffect solves the problem:

useEffect(() => {
  if (success) {
    dispatch(push("/fakepath"));
    setOpenModal(false);
  }
}, [...]);

Conclusion

React Hooks significantly improves code reusability, readability, and ease of integration testing. At Confluent, we’ve learned to address rendering issues by properly applying memoization when using React Hooks. To help you apply React Hook best practices, be sure to familiarize yourself with React Hooks’ rules and FAQ.

And if you haven’t already, get started with fully managed Apache Kafka® as a service with Confluent Cloud to receive $200 off usage each month for your first three months. You can also use the promo code CL60BLOG to receive $60 of additional free usage.*

Start Free

Mi Li joined Confluent as a software engineer in 2019 and works on a variety of data monitoring projects. He earned a master’s degree in computer science from the University of Southern California and a bachelor’s degree in computer science from the University of Liverpool.

Did you like this blog post? Share it now

Subscribe to the Confluent blog

More Articles Like This

Apache Kafka Made Simple: A First Glimpse of a Kafka Without ZooKeeper

At the heart of Apache Kafka® sits the log—a simple data structure that uses sequential operations that work symbiotically with the underlying hardware. Efficient disk buffering and CPU cache usage,

Lessons Learned from Running Apache Kafka at Scale at Pinterest

Apache Kafka® is at the heart of the data transportation layer at Pinterest. The amount of data that runs through Kafka has constantly grown over the years. This growth sometimes

How Real-Time Stream Processing Safely Scales with ksqlDB, Animated

Software engineering memes are in vogue, and nothing is more fashionable than joking about how complicated distributed systems can be. Despite the ribbing, many people adopt them. Why? Distributed systems