How I architected a single-page React application
With Data Structures, Components and integration with Redux
I recently built a single-page application that interacts with a backend JSON API server. I chose to use React to deepen my understanding of React fundamentals and how each tool can help in building a scalable frontend.
The stack of this application consists of:
- Frontend with React/Redux
- A backend JSON API server with Sinatra, integrated with Postgres for database persistence
- An API client that fetches data from OMDb API, written in Ruby
For this post, we’ll assume that we have the backend completed. So let’s focus on how design decisions are made on the frontend.
Side note: The decisions presented here are for reference only and may vary depending on the needs of your application. An example OMDb Movie Tracker app is used here for demonstration.
The App
The application consists of a search input form. A user can input a movie title to return a movie result from OMDb. The user can also save a movie with a rating and short comment into a favorites list.
To view the final app, click here. To view the source code, click here.
When a user searches a movie on the homepage, it looks like this:
For the sake of simplicity, we’ll only focus on designing the core features of the application in this article. You can also skip to Part II: Redux of the series.
Data Structure
Defining appropriate data structures should be one of the most important aspects of designing an app. This should come as the first step, as it determines not only how the frontend should render the elements, but also how the API server should return the JSON responses.
For this app, we’ll need two main pieces of information to properly render our UI: a single movie result and a list of favorited movies.
Movie result object
A single movie result will contain information such as the title, year, description, and poster image. With this, we need to define an object that can store these attributes:
{
"title": "Star Wars: Episode IV - A New Hope",
"year": "1977",
"plot": "Luke Skywalker joins forces with a Jedi Knight...",
"poster": "https://m.media-amazon.com/path/to/poster.jpg",
"imdbID": "tt0076759"
}The poster property is simply a URL to the poster image that will be displayed in the results. If there’s no poster available for that movie, it will be “N/A”, which we will display a placeholder. We will also need an imdbID attribute to uniquely identify each movie. This is useful for determining whether or not a movie result already exists in the favorites list. We’ll explore later on how it works.
Favorites list
The favorites list will contain all of the movies saved as favorites. The list will look something like this:
[
{ title: "Star Wars", year: "1977", ..., rating: 4 },
{ title: "Avatar", year: "2009", ..., rating: 5 }
]Keep in mind that we’ll need to look up a specific movie from the list, and the time complexity for this approach is O(N). While it works fine for smaller datasets, imagine having to search for a movie in a favorites list that grows indefinitely.
With this in mind, I chose to go with a hash table with keys as imdbID and values as favorited movie objects:
{
tt0076759: {
title: "Star Wars: Episode IV - A New Hope",
year: "1977",
plot: "...",
poster: "...",
rating: "4",
comment: "May the force be with you!",
},
tt0499549: {
title: "Avatar",
year: "2009",
plot: "...",
poster: "...",
rating: "5",
comment: "Favorite movie!",
}
}With this, we can look up a movie in the favorites list in O(1) time by its imdbID.
Note: the runtime complexity is probably not going to matter in most cases since the datasets are usually small on the client-side. We are also going to perform slicing and copying (also O(N) operations) in Redux anyway. But as an engineer, it’s good to be aware of potential optimizations that we can perform.
Components
Components are at the heart of React. We’ll need to determine which ones that will interact with the Redux store, and which ones that are only for presentation. We can also reuse some of the presentational components too. Our component hierarchy will look something like this:
Main page
We designate our App component at the top level. When the root path is visited, it needs to render the SearchContainer. It also needs to display flash messages to the user and handle the client-side routing.
The SearchContainer will retrieve the movie result from our Redux store, providing information as props to MovieItem for rendering. It will also dispatch a search action when a user submits a search in SearchInputForm. More on Redux later.
Add To Favorites Form
When the user clicks on the “Add To Favorites” button, we will display the AddFavoriteForm, a controlled component.
We are constantly updating its state whenever a user changes the rating or input text in the comment text area. This is useful for validation upon form submission.
The RatingForm is responsible to render the yellow stars when the user clicks on them. It also informs the current rating value to AddFavoriteForm.
Favorites Tab
When a user clicks on the “Favorites” tab, the App renders FavoritesContainer.
The FavoritesContainer is responsible for retrieving the favorites list from the Redux store. It also dispatches actions when a user changes a rating or clicks on the “Remove” button.
Our MovieItem and FavoritesInfo are simply presentational components that receive props from FavoritesContainer.
We’ll reuse the RatingForm component here. When a user clicks on a star in the RatingForm, the FavoritesContainer receives the rating value and dispatches an update rating action to the Redux store.
Redux Store
Our Redux store will include reducers that handle the search and favorites actions. Additionally, we’ll need to include a status reducer to track state changes when a user initiates an action. We’ll explore more on the status reducer later.
//store.jsimport { createStore, combineReducers, applyMiddleware } from 'redux';
import thunk from "redux-thunk";import search from './reducers/searchReducer';
import favorites from './reducers/favoritesReducer';
import status from './reducers/statusReducer';export default createStore(
combineReducers({
search,
favorites,
status
}),
{},
applyMiddleware(thunk)
)
We’ll also apply the Redux Thunk middleware right away. We’ll go more into detail on that later. Now, let’s figure out how we manage the state changes when a user submits a search.
Search Reducer
When a user performs a search action, we want to update the store with a new search result via searchReducer. We can then render our components accordingly. The general flow of events looks like this:
We’ll treat “Get search result” as a black box for now. We’ll explore how that works later with Redux Thunk. Now, let’s implement the reducer function.
//searchReducer.jsconst initialState = {
"title": "",
"year": "",
"plot": "",
"poster": "",
"imdbID": "",
}export default (state = initialState, action) => {
if (action.type === 'SEARCH_SUCCESS') {
state = action.result;
}
return state;
}
The initialState will represent the data structure defined earlier as a single movie result object. In the reducer function, we handle the action where a search is successful. If the action is triggered, we simply reassign the state to the new movie result object.
//searchActions.js
export const searchSuccess = (result) => ({
type: 'SEARCH_SUCCESS', result
});We define an action called searchSuccess that takes in a single argument, the movie result object, and returns an action object of type “SEARCH_SUCCESS”. We will dispatch this action upon a successful search API call.
Redux Thunk: Search
Let’s explore how the “Get search result” from earlier works. First, we need to make a remote API call to our backend API server. When the request receives a successful JSON response, we’ll dispatch the searchSuccess action along with the payload to searchReducer.
Knowing that we’ll need to dispatch after an asynchronous call completes, we’ll make use of Redux Thunk. Thunk comes into play for making multiple dispatches or delaying a dispatch. With Thunk, our updated flow of events looks like this:
For this, we define a function that takes in a single argument title and serves as the initial search action. This function is responsible for fetching the search result and dispatching a searchSuccess action:
//searchActions.js
import apiClient from '../apiClient';...export function search(title) {
return (dispatch) => {
apiClient.query(title)
.then(response => {
dispatch(searchSuccess(response.data))
});
}
}
We’ve set up our API client beforehand, and you can read more about how I set up the API client here. The apiClient.query method simply performs an AJAX GET request to our backend server and returns a Promise with the response data.
We can then connect this function as an action dispatch to our SearchContainer component:
//SearchContainer.jsimport React from 'react';
import { connect } from 'react-redux';
import { search } from '../actions/searchActions';...const mapStateToProps = (state) => (
{
result: state.search,
}
);const mapDispatchToProps = (dispatch) => (
{
search(title) {
dispatch(search(title))
},
}
);export default connect(mapStateToProps, mapDispatchToProps)(SearchContainer);
When a search request succeeds, our SearchContainer component will render the movie result:
Handling Other Search Statuses
Now we have our search action working properly and connected to our SearchContainer component, we’d like to handle other cases other than a successful search.
Search request pending
When a user submits a search, we’ll display a loading animation to indicate that the search request is pending:
Search request succeeds
If the search fails, we’ll display an appropriate error message to the user. This is useful to provide some context. A search failure could happen in cases where a movie title is not available, or our server is experiencing issues communicating with the OMDb API.
To handle different search statuses, we’ll need a way to store and update the current status along with any error messages.
Status Reducer
The statusReducer is responsible for tracking state changes whenever a user performs an action. The current state of an action can be represented by one of the three “statuses”:
- Pending (when a user first initiates the action)
- Success (when a request returns a successful response)
- Error (when a request returns an error response)
With these statuses in place, we can render different UIs based on the current status of a given action type. In this case, we’ll focus on tracking the status of the search action.
We’ll start by implementing the statusReducer. For the initial state, we need to track the current search status and any errors:
// statusReducer.js
const initialState = {
search: '', // status of the current search
searchError: '', // error message when a search fails
}Next, we need to define the reducer function. Whenever our SearchContainer dispatches a “SEARCH_[STATUS]” action, we will update the store by replacing the search and searchError properties.
// statusReducer.js...export default (state = initialState, action) => {
const actionHandlers = {
'SEARCH_REQUEST': {
search: 'PENDING',
searchError: '',
},
'SEARCH_SUCCESS': {
search: 'SUCCESS',
searchError: '',
},
'SEARCH_FAILURE': {
search: 'ERROR',
searchError: action.error,
},
}
const propsToUpdate = actionHandlers[action.type];
state = Object.assign({}, state, propsToUpdate);
return state;
}
We use an actionHandlers hash table here since we are only replacing the state’s properties. Furthermore, it improves readability more than using if/else or case statements.
With our statusReducer in place, we can render the UI based on different search statuses. We will update our flow of events to this:
We now have additional searchRequest and searchFailure actions available to dispatch to the store:
//searchActions.jsexport const searchRequest = () => ({
type: 'SEARCH_REQUEST'
});export const searchFailure = (error) => ({
type: 'SEARCH_FAILURE', error
});
To update our search action, we will dispatch searchRequest immediately and will dispatch searchSuccess or searchFailure based on the eventual success or failure of the Promise returned by Axios:
//searchActions.js...export function search(title) {
return (dispatch) => {
dispatch(searchRequest());apiClient.query(title)
.then(response => {
dispatch(searchSuccess(response.data))
})
.catch(error => {
dispatch(searchFailure(error.response.data))
});
}
}
We can now connect the search status state to our SearchContainer, passing it as a prop. Whenever our store receives the state changes, our SearchContainer renders a loading animation, an error message, or the search result:
//SearchContainer.js...(imports omitted)const SearchContainer = (props) => (
<main id='search-container'>
<SearchInputForm
placeholder='Search movie title...'
onSubmit={ (title) => props.search(title) }
/>
{
(props.searchStatus === 'SUCCESS')
? <MovieItem
movie={ props.result }
...(other props)
/>
: null
}
{
(props.searchStatus === 'PENDING')
? <section className='loading'>
<img src='../../images/loading.gif' />
</section>
: null
}
{
(props.searchStatus === 'ERROR')
? <section className='error'>
<p className='error'>
<i className="red exclamation triangle icon"></i>
{ props.searchError }
</p>
</section>
: null
}
</main>
);const mapStateToProps = (state) => (
{
searchStatus: state.status.search,
searchError: state.status.searchError,
result: state.search,
}
);...
Favorites Reducer
We’ll need to handle CRUD actions performed by a user on the favorites list. Recalling from our API endpoints earlier, we’d like to allow users to perform the following actions and update our store accordingly:
- Save a movie into the favorites list
- Retrieve all favorited movies
- Update a favorite’s rating
- Delete a movie from the favorites list
To ensure that the reducer function is pure, we simply copy the old state into a new object together with any new properties usingObject.assign. Note that we only handle actions with types of _SUCCESS:
//favoritesReducer.jsexport default (state = {}, action) => {
switch (action.type) {
case 'SAVE_FAVORITE_SUCCESS':
state = Object.assign({}, state, action.favorite);
break;case 'GET_FAVORITES_SUCCESS':
state = action.favorites;
break;case 'UPDATE_RATING_SUCCESS':
state = Object.assign({}, state, action.favorite);
break;case 'DELETE_FAVORITE_SUCCESS':
state = Object.assign({}, state);
delete state[action.imdbID];
break;default: return state;
}
return state;
}
We’ll leave the initialState as an empty object. The reason is that if our initialState contains placeholder movie items, our app will render them immediately before waiting for the actual favorites list response from our backend API server.
From now on, each of the favorites action will follow a general flow of events illustrated below. The pattern is similar to the search action in the previous section, except right now we’ll skip handling any “PENDING” status.
Save Favorites Action
Take the save favorites action for example. The function makes an API call to with our apiClient and dispatches either a saveFavoriteSuccess or a saveFavoriteFailure action, depending on whether or not we receive a successful response:
//favoritesActions.js
import apiClient from '../apiClient';export const saveFavoriteSuccess = (favorite) => ({
type: 'SAVE_FAVORITE_SUCCESS', favorite
});export const saveFavoriteFailure = (error) => ({
type: 'SAVE_FAVORITE_FAILURE', error
});export function save(movie) {
return (dispatch) => {
apiClient.saveFavorite(movie)
.then(res => {
dispatch(saveFavoriteSuccess(res.data))
})
.catch(err => {
dispatch(saveFavoriteFailure(err.response.data))
});
}
}
We can now connect the save favorite action to AddFavoriteForm through React Redux.
To read more about how I handled the flow to display flash messages, click here.
Conclusion
Designing the frontend of an application requires some forethought, even when using a popular JavaScript library such as React. By thinking about how the data structures, components, APIs, and state management work as a whole, we can better anticipate edge cases and effectively fix errors when they arise. By using certain design patterns such as controlled components, Redux, and handling AJAX workflow using Thunk, we can streamline managing the flow of providing UI feedback to user actions. Ultimately, how we approach the design will have an impact on usability, clarity, and future scalability.
References
Fullstack React: The Complete Guide to ReactJS and Friends
About me
I am a software engineer located in NYC and co-creator of SpaceCraft. I have experience in designing single-page applications, synchronizing state between multiple clients, and deploying scalable applications with Docker.
I am currently looking for my next full-time opportunity! Please get in touch if you think that I will be a good fit for your team.
