How to handle RESTful web Services using Retrofit, OkHttp, Gson, Glide and Coroutines
Kriptofolio app series — Part 5
These days almost every Android app connects to internet to get/send data. You should definitely learn how to handle RESTful Web Services, as their correct implementation is the core knowledge while creating modern apps.
This part is going to be complicated. We are going to combine multiple libraries at once to get a working result. I am not going to talk about the native Android way to handle internet requests, because in the real world nobody uses it. Every good app does not try to reinvent the wheel but instead uses the most popular third party libraries to solve common problems. It would be too complicated to recreate the functionality that these well-made libraries have to offer.
- Introduction: A roadmap to build a modern Android app in 2018–2019
- Part 1: An introduction to the SOLID principles
- Part 2: How to start building your Android app: creating Mockups, UI, XML layouts
- Part 3: All about that Architecture: exploring different architecture patterns and how to use them in your app
- Part 4: How to implement Dependency Injection in your app with Dagger 2
- Part 5: Handle RESTful Web Services using Retrofit, OkHttp, Gson, Glide and Coroutines (you’re here)
What is Retrofit, OkHttp and Gson?
Retrofit is a REST Client for Java and Android. This library, in my opinion, is the most important one to learn, as it will do the main job. It makes it relatively easy to retrieve and upload JSON (or other structured data) via a REST based webservice.
In Retrofit you configure which converter is used for the data serialization. Typically to serialize and deserialize objects to and from JSON you use an open-source Java library — Gson. Also if you need, you can add custom converters to Retrofit to process XML or other protocols.
For making HTTP requests Retrofit uses the OkHttp library. OkHttp is a pure HTTP/SPDY client responsible for any low-level network operations, caching, requests and responses manipulation. In contrast, Retrofit is a high-level REST abstraction build on top of OkHttp. Retrofit is strongly coupled with OkHttp and makes intensive use of it.
Now that you know that everything is closely related, we are going to use all these 3 libraries at once. Our first goal is to get all the cryptocurrencies list using Retrofit from the Internet. We will use a special OkHttp interceptor class for CoinMarketCap API authentication when making a call to the server. We will get back a JSON data result and then convert it using the Gson library.
Quick setup for Retrofit 2 just to try it first
When learning something new, I like to try it out in practice as soon as I can. We will apply a similar approach with Retrofit 2 for you to understand it better more quickly. Don’t worry right now about code quality or any programming principles or optimizations — we’ll just write some code to make Retrofit 2 work in our project and discuss what it does.
Follow these steps to set up Retrofit 2 on My Crypto Coins app project:
First, give INTERNET permission for the app
We are going to execute HTTP requests on a server accessible via the Internet. Give this permission by adding these lines to your Manifest file:
Then you should add library dependencies
Find the latest Retrofit version. Also you should know that Retrofit doesn’t ship with an integrated JSON converter. Since we will get responses in JSON format, we need to include the converter manually in the dependencies too. We are going to use latest Google’s JSON converter Gson version. Let’s add these lines to your gradle file:
As you noticed from my comment, the OkHttp dependency is already shipped with the Retrofit 2 dependency. Versions is just a separate gradle file for convenience:
Next set up the Retrofit interface
It’s an interface that declares our requests and their types. Here we define the API on the client side.
And set up the data class
Data classes are POJOs (Plain Old Java Objects) that represent the responses of the API calls we’re going to make.
Create a special interceptor class for authentication when making a call to the server
This is the case particular for any API that requires authentication to get a successful response. Interceptors are a powerful way to customize your requests. We are going to intercept the actual request and to add individual request headers, which will validate the call with an API Key provided by CoinMarketCap Professional API Developer Portal. To get yours, you need to register there.
Finally, add this code to our activity to see Retrofit working
I wanted to get your hands dirty as soon as possible, so I put everything in one place. This is not the correct way, but it’s the fastest instead just to see a visual result quickly.
You can explore the code here. Remember this is only an initial simplified implementation version for you to get the idea better.
Final correct setup for Retrofit 2 with OkHttp 3 and Gson
Ok after a quick experiment, it is time to bring this Retrofit implementation to the next level. We already got the data successfully but not correctly. We are missing the states like loading, error and success. Our code is mixed without separation of concerns. It’s a common mistake to write all your code in an activity or a fragment. Our activity class is UI based and should only contain logic that handles UI and operating system interactions.
Actually, after this quick setup, I worked a lot and made many changes. There is no point to put all the code that was changed in the article. Better instead you should browse the final Part 5 code repo here. I have commented everything very well and my code should be clear for you to understand. But I am going to talk about most important things I have done and why I did them.
The first step to improve was to start using Dependency Injection. Remember from the previous part we already have Dagger 2 implemented inside the project correctly. So I used it for the Retrofit setup.
Now as you see, Retrofit is separated from the activity class as it should be. It will be initialized only once and used app-wide.
As you may have noticed while creating the Retrofit builder instance, we added a special Retrofit calls adapter using
addCallAdapterFactory. By default, Retrofit returns a
Call<T>, but for our project we require it to return a
LiveData<T> type. In order to do that we need to add
LiveDataCallAdapter by using
Now we will get
LiveData<T> instead of
Call<T> as the return type from Retrofit service methods defined in the
Another important step to make is to start using the Repository pattern. I have talked about it in Part 3. Check out our MVVM architecture schema from that post to remember where it goes.
As you see in the picture, Repository is a separate layer for the data. It’s our single source of contact for getting or sending data. When we use Repository, we are following the separation of concerns principle. We can have different data sources (like in our case persistent data from an SQLite database and data from web services), but Repository is always going to be single source of truth for all app data.
Instead of communicating with our Retrofit implementation directly, we are going to use Repository for that. For each kind of entity, we are going to have a separate Repository.
As you notice in the
CryptocurrencyRepository class code, I am using the
NetworkBoundResource abstract class. What is it and why do we need it?
NetworkBoundResource is a small but very important helper class that will allow us to maintain a synchronization between the local database and the web service. Our goal is to build a modern application that will work smoothly even when our device is offline. Also with the help of this class we will be able to present different network states like errors or loading for the user visually.
NetworkBoundResource starts by observing the database for the resource. When the entry is loaded from the database for the first time, it checks whether the result is good enough to be dispatched or if it should be re-fetched from the network. Note that both of these situations can happen at the same time, given that you probably want to show cached data while updating it from the network.
If the network call completes successfully, it saves the response into the database and re-initializes the stream. If the network request fails, the
NetworkBoundResource dispatches a failure directly.
Under the hood, the
NetworkBoundResource class is made by using MediatorLiveData and its ability to observe multiple LiveData sources at once. Here we have two LiveData sources: the database and the network call response. Both of those LiveData are wrapped into one MediatorLiveData which is exposed by
Let’s take a closer look how the
NetworkBoundResource will work in our app. Imagine the user will launch the app and click on a floating action button on the bottom right corner. The app will launch the add crypto coins screen. Now we can analyze
NetworkBoundResource's usage inside it.
If the app is freshly installed and it is its first launch, then there will not be any data stored inside the local database. Because there is no data to show, a loading progress bar UI will be shown. Meanwhile the app is going to make a request call to the server via a web service to get all the cryptocurrencies list.
If the response is unsuccessful then the error message UI will be shown with the ability to retry a call by pressing a button. When a request call is successful at last, then the response data will be saved to a local SQLite database.
If we come back to the same screen the next time, the app will load data from the database instead of making a call to the internet again. But the user can ask for a new data update by implementing pull-to-refresh functionality. Old data information will be shown whilst the network call is happening. All this is done with the help of
Another class used in our Repository and
LiveDataCallAdapter where all the "magic" happens is
ApiResponse is just a simple common wrapper around the
Retrofit2.Response class that converts each response to an instance of LiveData.
Inside this wrapper class, if our response has an error, we use the Gson library to convert the error to a JSON object. However, if the response was successful, then the Gson converter for JSON to POJO object mapping is used. We already added it when creating the retrofit builder instance with
GsonConverterFactory inside the Dagger
Glide for image loading
What is Glide? From the docs:
Glide is a fast and efficient open source media management and image loading framework for Android that wraps media decoding, memory and disk caching, and resource pooling into a simple and easy to use interface.
Glide’s primary focus is on making scrolling any kind of a list of images as smooth and fast as possible, but it is also effective for almost any case where you need to fetch, resize, and display a remote image.
Sounds like a complicated library which offers many useful features that you would not want to develop all by yourself. In My Crypto Coins app, we have several list screens where we need to show multiple cryptocurrency logos — pictures taken from the internet all at once — and still ensure a smooth scrolling experience for the user. So this library fits our needs perfectly. Also this library is very popular among Android developers.
Steps to setup Glide on My Crypto Coins app project:
Get the latest Glide version. Again versions is a separate file
versions.gradle for the project.
Because we want to use the networking library OkHttp in our project for all network operations, we need to include the specific Glide integration for it instead of the default one. Also since Glide is going to perform a network request to load images via the internet, we need to include the permission
INTERNET in our
AndroidManifest.xml file — but we already did that with the Retrofit setup.
Glide v4, which we will be using, offers a generated API for Applications. It will use an annotation processor to generate an API that allows applications to extend Glide’s API and include components provided by integration libraries. For any app to access the generated Glide API we need to include an appropriately annotated
AppGlideModule implementation. There can be only a single implementation of the generated API and only one
AppGlideModule per application.
Let’s create a class extending
AppGlideModule somewhere in your app project:
Even if our application is not changing any additional settings or implementing any methods in
AppGlideModule, we still need to have its implementation to use Glide. You're not required to implement any of the methods in
AppGlideModule for the API to be generated. You can leave the class blank as long as it extends
AppGlideModule and is annotated with
Use Glide-generated API
AppGlideModule, applications can use the API by starting all loads with
GlideApp.with(). This is the code that shows how I have used Glide to load and show cryptocurrency logos in the add crypto coins screen all cryptocurrencies list.
As you see, you can start using Glide with just few lines of code and let it do all the hard work for you. It is pretty straightforward.
While building this app, we are going to face situations when we will run time consuming tasks such as writing data to a database or reading from it, fetching data from the network and other. All these common tasks take longer to complete than allowed by the Android framework’s main thread.
The main thread is a single thread that handles all updates to the UI. Developers are required not to block it to avoid the app freezing or even crashing with an Application Not Responding dialog. Kotlin coroutines is going to solve this problem for us by introducing main thread safety. It is the last missing piece that we want to add for My Crypto Coins app.
Coroutines are a Kotlin feature that convert async callbacks for long-running tasks, such as database or network access, into sequential code. With coroutines, you can write asynchronous code, which was traditionally written using the Callback pattern, using a synchronous style. The return value of a function will provide the result of the asynchronous call. Code written sequentially is typically easier to read, and can even use language features such as exceptions.
So we are going to use coroutines everywhere in this app where we need to wait until a result is available from a long-running task and than continue execution. Let’s see one exact implementation for our ViewModel where we will retry getting the latest data from the server for our cryptocurrencies presented on the main screen.
First add coroutines to the project:
Then we will create abstract class which will become the base class to be used for any ViewModel that needs to have common functionality like coroutines in our case:
Here we create specific coroutine scope, which will control the lifetime of coroutines through its job. As you see, scope allows you to specify a default dispatcher that controls which thread runs a coroutine. When the ViewModel is no longer used, we cancel
viewModelJob and with that every coroutine started by
uiScope will be cancelled as well.
Finally, implement the retry functionality:
Here we call a function marked with a special Kotlin keyword
suspend for coroutines. This means that the function suspends execution until the result is ready, then it resumes where it left off with the result. While it is suspended waiting for a result, it unblocks the thread that it is running on.
Also, in one suspend function we can call another suspend function. As you see we do that by calling new suspend function marked
withContext that is executed on different thread.
The idea of all this code is that we can combine multiple calls to form nice-looking sequential code. First we request to get the ids of the cryptocurrencies we own from the local database and wait for the response. Only after we get it do we use the response ids to make a new call with Retrofit to get those updated cryptocurrency values. That is our retry functionality.
We made it! Final thoughts, repository, app & presentation
Congratulations, I am happy if you managed to reach to the end. All the most significant points for creating this app have been covered. There was plenty of new stuff done in this part and a lot of that is not covered by this article, but I commented my code everywhere very well so you should not get lost in it. Check out final code for this part 5 here on GitHub:
The biggest challenge for me personally was not to learn new technologies, not to develop the app, but to write all these articles. Actually I am very happy with myself that I completed this challenge. Learning and developing is easy compared to teaching others, but that is where you can understand the topic even better. My advice if you are looking for the best way to learn new things is to start creating something yourself immediately. I promise you will learn a lot and quickly.
All these articles are based on version 1.0.0 of “Kriptofolio” (previously “My Crypto Coins”) app which you can download as a separate APK file here. But I will be very happy if you install and rate the latest app version from the store directly:
Also please feel free to visit this simple presentation website that I made for this project:
Ačiū! Thanks for reading! I originally published this post for my personal blog www.baruckis.com on May 11, 2019.