How-To: Quickly Create a Strongly-Typed REST Client for an API

When accessing resources with a REST-style API is a significant part of your application, it is common to look for an SDK that provides a strongly-typed client for the API. This allows you to avoid writing boilerplate code to make requests and deserialize responses. There is sometimes no SDK available yet for the API you want to use, or it's not compatible with your target platform. With the design goal of avoiding boilerplate code in mind, we will walk through how to use Refit to quickly create a strongly-typed client for a REST API and register it with the service collection as an endpoint.

Step-by-steps

Important

This guide assumes you used the template wizard or dotnet new unoapp to create your solution. If not, it is recommended that you follow the instructions for creating an application from the template.

1. Enable HTTP

When working with a complex application, centralized registration of your API endpoints is a good practice. This allows you to easily change the endpoint for a given service. It also reduces the complexity to adding new services which can then have their HttpClient instance reused across multiple instances of view models or other services.

The first step to centrally registering any API endpoint is to enable HTTP on the IHostBuilder:

protected override void OnLaunched(LaunchActivatedEventArgs args)
{
    var appBuilder = this.CreateBuilder(args)
        .Configure(hostBuilder =>
        {
            hostBuilder.UseHttp();
        });
...

This feature requires the Uno.Extensions.Http.WinUI package. It uses Microsoft Extensions for any HTTP-related work such as naming or configuring the HttpClient instance associated with your endpoints.

2. Define the live interface

For the purposes of this tutorial, we will create and register an interface for the Chuck Norris facts web API which is documented here
This interface will be used to generate the strongly-typed client. It will be used to make requests to the web service and deserialize responses.
Start by defining an interface IChuckNorrisEndpoint with a method Search() that returns a Task of type ApiResponse<ChuckNorrisData>:
```
[Headers("Content-Type: application/json")]
public interface IChuckNorrisEndpoint
{
    [Get("/jokes/search")]
    Task<ApiResponse<ChuckNorrisData>> Search(CancellationToken ct, [AliasAs("query")] string searchTerm);
}
```
Note

A class named ChuckNorrisData will be defined in the next section of this guide. It will be used to deserialize the response from the API.
The Headers attribute is used to specify the Content-Type header for the request. Get specifies the relative path for a Search request. Notice that the searchTerm parameter for the request is aliased as query using an AliasAs attribute.
ApiResponse<T> is used to deserialize the response from the API. It is defined in the Refit package.

3. Generate a data model from the API response

The next step is to generate a data model from the response we will get from the API. This will be used to deserialize the response which we'll learn is a JSON object with a total property and result that contains an array of Chuck Norris data objects.
By the end of this section, a ChuckNorrisData class will be defined in the Models namespace to represent each object in the array.

We will use Hoppscotch to make an HTTP request and inspect the response. Open it, and make a GET request to the /jokes/search endpoint with the query parameter query set to fight.

Your screen should appear similar to this:
For informational purposes, this is the full HTTP request URL:
```
GET https://api.chucknorris.io/jokes/search?query=fight
```

The response body pane will populate with a formatted JSON object that looks similar to this:

{
    "total": 111,
    "result": [
        {
            "categories": [],
            "created_at": "2020-01-05 13:42:18.823766",
            "icon_url": "https://assets.chucknorris.host/img/avatar/chuck-norris.png",
            "id": "VvGL-lRHSsOL-jj9IEDWRA",
            "updated_at": "2020-01-05 13:42:18.823766",
            "url": "https://api.chucknorris.io/jokes/VvGL-lRHSsOL-jj9IEDWRA",
            "value": "Chuck Norris doesn't beat people up he looks at them they get scared and fight their self to the death"
        },
    ...
    ]
}

Since this response is needed for the next step, copy the entire JSON object from the response body pane

Next, we need to use a language-agnostic tool that infers a data model from the response we recieved. We will open quicktype and use it to generate a data model from the JSON we copied above
- Replace any demo text in the left JSON pane with our text by pasting the JSON object into editor
- Within the properties box on the right side, make the following selections:
  - C# as the language
  - System Text Json as the serialization framework
  - Complete as the output features
- Your screen should appear similar to this:
- For the purposes of this tutorial, only select and copy the two model classes Welcome and Result from the right side containing the generated C# code.
  - What you copied should appear like this:
```
public partial class Welcome
{
    [JsonPropertyName("total")]
    public long Total { get; set; }

    [JsonPropertyName("result")]
    public List<Result> Result { get; set; }
}

public partial class Result
{
    [JsonPropertyName("categories")]
    public List<Category> Categories { get; set; }

    [JsonPropertyName("created_at")]
    public DateTimeOffset CreatedAt { get; set; }

    [JsonPropertyName("icon_url")]
    public Uri IconUrl { get; set; }

    [JsonPropertyName("id")]
    public string Id { get; set; }

    [JsonPropertyName("updated_at")]
    public DateTimeOffset UpdatedAt { get; set; }

    [JsonPropertyName("url")]
    public Uri Url { get; set; }

    [JsonPropertyName("value")]
    public string Value { get; set; }
}
```
- Go back to Visual Studio and create a new folder named Models in the shared project. Create a new file named ChuckNorrisData.cs in the Models folder and paste the code you copied into it.
- Rename the partial classes Welcome to ChuckNorrisData and Result to ChuckNorrisDataResult

4. Register the endpoint

Unlike standard HTTP endpoints, Refit endpoints you define will be registered as a service with the AddRefitClient<T>() extension method, where the type argument T corresponds to an interface you define.

This extension method can take a delegate as its argument:

protected override void OnLaunched(LaunchActivatedEventArgs args)
{
    var appBuilder = this.CreateBuilder(args)
        .Configure(hostBuilder =>
        {
            hostBuilder.UseHttp((context, services) =>
                services.AddRefitClient<IChuckNorrisEndpoint>(context, httpClient =>
                {
                    httpClient.BaseAddress = new Uri("https://api.chucknorris.io/");
                })
            );
        });
...

Another way to configure an HTTP client is to specify a configuration section name. This allows you to configure the added HTTP client using the appsettings.json file. This tutorial demonstrates use of the configuration section method.

Add the Refit client to the service collection with the following code instead:

protected override void OnLaunched(LaunchActivatedEventArgs args)
{
    var appBuilder = this.CreateBuilder(args)
        .Configure(hostBuilder =>
        {
            hostBuilder.UseHttp((context, services) =>
                services.AddRefitClient<IChuckNorrisEndpoint>(context)
            );
        });
...

Look for a file named appsettings.json in the shared project. If it does not exist, create it. Open this file and add the following JSON:
```
{
    "ChuckNorrisEndpoint": {
        "Url": "https://api.chucknorris.io/",
        "UseNativeHandler": true
    }
}
```
The Url property specifies the base URL for the API.
We set UseNativeHandler to true to use the native HTTP handler for the platform. This is recommended for production apps unless you have a specific reason to use the managed handler.
- On Windows, intercepting the network traffic of your app with tools like Fiddler is not possible when using the native handler. This is because it does not use a compatible network stack.
- In these cases, you can set UseNativeHandler to false to use the managed handler instead
That block of JSON from the above file is called a configuration section. In this case, it's used to configure the HTTP client associated with the endpoint. For more information, see the overview for configuration sections.
Observe that the name of the configuration section is similar to that of the endpoint interface we defined earlier. By default, the interface name without the leading I prefix is assumed to be the section name. This can be overridden by specifying a name while registering the service. The name you use should be the second argument to AddRefitClient<T>().

5. Use the endpoint

IChuckNorrisEndpoint can now be used in a service implementation by injecting it into the constructor. This interface will be used to make requests to the web service and deserialize responses.

For the purposes of this tutorial, we will create a view model for the page that triggers the request for a Chuck Norris fact named FactViewModel. It will have a method SearchAsync() that returns a Task of type ChuckNorrisData:

public class FactViewModel
{
    private readonly IChuckNorrisEndpoint _endpoint;

    public FactViewModel(IChuckNorrisEndpoint endpoint)
    {
        _endpoint = endpoint;
    }

    public async Task<ChuckNorrisData> SearchAsync(string searchTerm)
    {
        var response = await _endpoint.Search(CancellationToken.None, searchTerm);

        if (response.IsSuccessStatusCode && response.Content is not null)
        {
            // Return first result
            var content = response.Content;
            return content.Result[0];
        }

        _logger.LogError(response.Error, "An error occurred while retrieving the latest fact.");
        return Task.FromException<ChuckNorrisData>(response.Error);
    }
}

Now, any views that need to display a Chuck Norris fact can use the FactViewModel to retrieve one. Thanks to the use of Refit, the details of HTTP requests and deserialization are abstracted away from the view, and boilerplate code is avoided.