ARC — Simplifying async requests in Redux apps

Vinicius Dacal
Frontend Weekly
Published in
11 min readSep 12, 2017

--

TL-DR: This article’s goal is to explain in details how Arc works. If you are only looking for instructions on how to get started with Arc, you can go directly to the repository’s README or its docs. However, it’s always good to know the tools you are working with.

A year ago, I wrote an article about how to handle async request in Redux apps. Since then, I passed through other projects and I end up using other technologies.

I have been working with redux apps for almost two years now, and I have tried different approaches to handle async requests. From the pain and frustrations I had on this process, an idea came up, to create a lib that abstracts the annoying parts. Something that cut off the code repetition, but at the same time were flexible and scalable, allowing the user to invest his time where it really matters. It that’s how Arc was born.

Arc is a 2kb lib, totally free of dependencies and 100% covered by tests.

We can split Arc in two independently parts:

  • Types and creators factory.
  • AsyncMiddleware.

Core concepts

First of all, let explore the Arc’s core concepts, to understand better how it works and what it really does.

Action creator

An action creator is a function that is used to create actions. An action creator allows us to centralize the creation of a certain kind of action, what helps us to create a consistency in the application. You can see an action creator example bellow:

Accordingly to the above action creator, to get an action ADD_TODO , we just need to execute the function addTodo, providing description and date, as the follow example:

Async Actions

An async action, is an action that describes a request. Bellow, you can see an example of an async action from Arc:

The first thing different we can notice is the type. The type of a regular action is usually a string, and in the case above, we have an array with two types, one to be used when the request starts and another to be used when it ends.

Another thing we can notice, is that we have an object meta, which has an attribute url and another attribute method. The url would be the request’s endpoint and the method would be the one used to perform the request(get, post, put, delete).

Last but not least, we have the payload, that will be the content sent in the request. In the last example, we are creating a new to-do item, and the service in question is waiting the parameters description and date.

In the case you need any additional data to be sent in the action, you can include them in the meta object.

{
...
meta:{
url,
method,
anyOtherParam,
},

}

Arc follows the flux-standard-action pattern to define its action. Their README is very simple and can clear possible doubts regarding to the decisions about the Arc actions’s format.

The async action is only a design pattern, the action its own doesn’t do the request. Who is responsible for that is the async middleware, that understands the async actions and performs the necessary actions.

Middleware

A middleware is a function that is part of a function chain, that carries data from one point to other, modifying it or not.

In Redux, the data that passes through the middlewares, are the actions. That said, inside a middleware, you can change an action’s content or even intercept, removing it from the application flow.

A middleware in Redux, is a function that must respect the following signature:

In Redux, the middlewares are executed every time an action is dispatched in the application. Inside it, we have access to the store object, which contains the methods getState and dispatch, that can be used to interact with the redux state.

We also receive the method next, that should be executed passing the action as a param, to let the action follow its flow through the middlewares.

Lastly, we receive the action that was dispatched.

Notice that to modify an action’s content inside the middleware, you only need to clone the action, change the clone’s values and execute the method next with it:

const middleware = store => next => action => {
const newAction = {
...action,
payload: anyNewPayload,
};
next(newAction);
};

To intercept an action, you don’t call the method next:

const TYPE_TO_INTERCEPT = 'X';const middleware = store => next => action => {
if (action.type === TYPE_TO_INTERCEPT) {
return;
}
next(newAction);
};

In the above example, we intercept all the actions whose type is X.

Can you see the possibilities a middleware gives us? No? Don’t worry, everything will become clear a little further!

Async Middleware

Once we have the definition for the async action, we need to have a middleware that understands it and performs the request accordingly to its data.

Bellow, you can see an example of an async middleware.

Basic async middleware example

The above example is using axios, but in the Arc, you can choose whatever lib you like, or even implement the request in vanilla JS.

Right in the beginning of the middleware, we already check if the action type is an array. If it is not, we only call the function next, with the action we just received, to let it follow its natural flow.

If the action type is an array, we assume that the action in question is an async action. Given that, we extract both strings, requestType and responseType and the objects payload and meta:

const [requestType, responseType] = action.type;const { meta, payload } = action;

Before the request starts, we dispatch an action, whose type is the requestType and we also inject in it, the payload and meta. This way, we can have access to all the relevant request data inside the reducers:

store.dispatch({ type: requestType, payload, meta });

After dispatch the first action, we start the request, using the url and method from meta object, and also the payload:

axios[meta.method](meta.url, payload)

Once the request is done, we dispatch a second action, having responseType as its type and also having the meta .

In the case of success, the action.payload will be the result of the request:

store.dispatch({
type: responseType,
payload: response,
meta,
});

In the case of error, the action.payload becomes the request error and we set action.error to true:

store.dispatch({
type: responseType,
payload: error,
error: true,
meta,
});

With async middleware, we already simplified our work in terms of requests. However, a part of our work, remains mechanical and repetitive: The creation of types and action creators for each new kind of request.

Considering a RESTFul API, the types and action creators for a CRUD, would look like this:

If we pay attention, we can find numerals patterns that keeps repeating. Lets observe the code to add a to-do:

  • The types are always in uppercase, composed by: namespace, action name and request period.
  • The action creator name is the same as the action.
  • The payload will be always the request’s content, and it will be the dynamic content for the action

The url need to be dynamic in a certain manner, to allow us passing params to it when needed, as an id, for example.

Considering the points above and inspired by redux-sauce, I got following idea: What if we could define the name of our actions, inform their urls and methods, and a function gives us the types and the action creators?

Something that would be declarative and would allow our configs to be validated. That would reduce the possibility of mistakes from devs and would increase a lot their productivity. Based on that, I wrote an interface to define the requests:

In the above example, we have a config for a request that creates a to-do item. We define the action name and the respective url and method for the request. On the image bellow, we can see an example of what createApiAction returns to us:

The action name and the namespace must be always provided in lower camelcase.

The function createApiActions, returns to us, the creators and the types for the configured requests. A creator accepts as parameter, an object that may have:

  • a payload, which will become the payload for the request and the action.
  • Any other extra param, which will become part of the action’s meta and will also be used to parse the dynamic urls

Given that, to start a request, we only have to use a creator with dispatch, as the following example:

And then, the following action would be dispatched:

Dynamic URLs

In the above example, we are using static URLs, which will not attend all our needs. For that reason, Arc allows us to provide dynamic urls, which contains params as part of itself.

To use them, we only need to add : and the param name in the url definition, as the example bellow:

{ url: path/to/to-do/:id, method: 'anymethod' }

To parse the params that were defined in the url, will be used the values passed to the action creator, as the following example:

Above, we configured the requests: read, update e otherDynamicUrl. After define, we use them passing to them an object which contains the necessary values to parse the url. Bellow we have a detailed explanation about the action read:

As we can notice in the above image, Arc uses the parameter id, that was provided to the creator, to parse the url. It also takes any additional param, except the payload, and injects them in the action’s meta.

For that reason, it’s possible to see that beyond the url and method, we also have the id inside the meta.

On the image bellow, we can se an explanation about the action update:

Above, we can see that is possible to pass either params to the url, even the payload for the action.

Based on Arc’s config, our CRUD would look like this now:

If we compare this code with the previous CRUD example, we can notice that we save 55 lines, and we have a code considerably more simple.

Types and Errors Treatment

The types returned from createApiAction can be used inside the reducers, as the following example:

Above, we are covering all the requests phases, and we also handle the errors.

Accordingly to this example, we can use the action regarding to the beginning of the request to change the state, indicating that there is a request in progress.

When the action regarding to the end of the request be dispatched, the request result will come under action.payload, in the case of success.

When an error occurs, action.error will be true and the action.payload will be the request error itself.

Installation

To integrate Arc in a Redux project, it’s fairly simple:

First, you need to add it to your project’s dependencies, running:

yarn add redux-arc

or

npm install --save redux-arc

Then, you need to configure the async middleware:

Accordingly to the above code, first we need to import the function createAsyncMiddleware. Then we execute that function passing the function that will perform the request. createAsyncMiddleware will return the middleware, which should be applied to the store.

asyncTask

For more flexibility, Arc expects you to provide the function that will be responsible to perform the request. In the above example, we are using axios, but you can use any other lib. The options object will contain the url, the method and also the payload for the request. Once the request is finished, you only need to execute the function done, passing the respective error and response

Until here, we approached the most use cases regarding to api calls. However, there are some scenarios we didn’t cover, such as when we need to perform some changes to the payload before the request, or when we need to modify the response, or take any action once the request ends. For all those cases, Arc has the policies.

Policies

Policies are middlewares that you register on Arc. Once registered, the policy can be passed in the request’s config, as the example below:

First, we declared a policy named anyPolicy. Every policy must respect the following signature:

function PolicyName(store) {
done => (action, error, response) => {
done(action, error, response)
};
}
PolicyName.applyPoint = 'beforeRequest' or 'onResponse';

In the line 8, we are defining the policy’s applyPoint, which is required. They indicate which point of the request, the policy will be executed. The available applyPoints are: beforeRequest and onResponse.

In the line 9, we registered the policy on Arc, using the method register. With that done, the policy is already available to be used in any request.

To use the policies, as we do in the line 15, you only need to declare a parameter policies in the request’s config and pass an array with the policies you want to apply to it.

The above example gives us an idea of how to declare and use a policy. Now, imagine you want to create and update to-do items, but you don’t want to create two configs for that. You want to define a request named save, that creates or updates the item accordingly to the payload. That could be easily achieved with a policy, as below:

In the above policy, we receive an action and we clone it, to avoid data mutability.

We verify if the action’s payload contain an id and if so, we consider it as an update action. In that case, we add the id to the url, we change the method to put and we remove the id from the payload. Lastly, we pass the new action ahead.

Notice that the applyPoint we used was the beforeRequest, because we need to change the action’s content before the request starts.

The createOrUpdate policy, would be used as the following example:

Firt we define our request named save and define the url without the id. We define the method as post, which will be used in the creation, and we include the policy createOrUpdate to the request.

Beyond line 12, we can se the action creator being used in both cases, to create and update.

There are cases when you need to change the response instead the request. For those cases, you only need to change the applyPoint to 'onResponse'. Inside the policy, clone the response, change its data and pass it ahead, using the function done.

Conclusion

As we saw on this post, Arc reduces a lot the boilerplate when we are talking about api calls, in the same time, it gives flexibility to us developers.

The policies by themselves, give you the ability to structure and organize the edge cases in a clear and easy way.

You can go ahead and take a look at Arc’s source code. Contributions are welcome, including the ones to the docs.

Did you enjoy the post and think it’s useful? Give a clap 👏 below and help us spread the word :)

--

--

Vinicius Dacal
Frontend Weekly

Brazilian Software Engineer, working remotely for BEN Group UK. Loves learning, creating and sharing.