Configuration

Flurl.Http includes a robust set of options and techniques for configuring its behavior at various levels.

Settings

Flurl is configurable primarily through a Settings property available on IFlurlClient, IFlurlRequest, IFlurlClientBuilder, and HttpTest. Here are the available settings:

Setting Type Default Value
Timeout TimeSpan? 100 seconds
HttpVersion string "1.1"
AllowedHttpStatusRange string null ("2xx,3xx", effectively)
Redirects.Enabled bool true
Redirects.AllowSecureToInsecure bool false
Redirects.ForwardHeaders bool false
Redirects.ForwardAuthorizationHeader bool false
Redirects.MaxAutoRedirects int 10
JsonSerializer ISerializer DefaultJsonSerializer
UrlEncodedSerializer ISerializer UrlEncodedSerializer

All properties are read/write and can be set directly:

// set default on the client:
var client = new FlurlClient("https://some-api.com");
client.Settings.Timeout = TimeSpan.FromSeconds(600);
client.Settings.Redirects.Enabled = false;

// override on the request:
var request = client.Request("endpoint");
request.Settings.Timeout = TimeSpan.FromSeconds(1200);
request.Settings.Redirects.Enabled = true;

You can also configure them fluently:

clientOrRequest.WithSettings(settings => {
    settings.Timeout = TimeSpan.FromSeconds(600);
    settings.AllowedHttpStatusRange = "*";
    settings.Redirects.Enabled = false;
})...

Or even more fluently in many cases:

clientOrRequest
    .WithTimeout(600)
    .AllowAnyHttpStatus()
    .WithAutoRedirect(false)
    ...

When using clients, requests, and fluent configuration together, you'll want to pay close attention to which objects you're configuring, as it can determine whether subsequent calls with that client are affected:

await client
    .WithSettings(...) // configures the client, affects all subsequent requests
    .Request("endpoint") // creates and returns a request
    .WithSettings(...) // configures just this request
    .GetAsync();

The clientless pattern supports all the same extension methods. Configure and send a request without ever referencing a client explicitly:

var result = await "https://some-api.com/endpoint"
    .WithSettings(...) // configures just this request
    .WithTimeout(...) // ditto
    .GetJsonAsync<T>();

All of the above settings and extension methods are also available on IFlurlClientBuilder, which is useful for configuring things at startup:

// clientless pattern, all clients:
FlurlHttp.Clients.WithDefaults(builder =>
    builder.WithSettings(...));

// clientless pattern, for a specific site/service:
FlurlHttp.ConfigureClientForUrl("https://some-api.com")
    .WtihSettings(...);

// DI pattern:
services.AddSingleton<IFlurlClientCache>(_ => new FlurlClientCache()
    // all clients:
    .WithDefaults(builder =>
        builder.WithSettings(...))
    // specific named client:
    .Add("MyClient", "https://some-api.com", builder =>
        builder.WithSettings(...))

The fourth and final object that supports Settings (and all the related fluent goodness) is HttpTest, and it takes precedence over everything:

using var test = new HttpTest.AllowAnyHttpStatus();
await sut.DoThingAsync(); // no matter how things are configured in the SUT,
                          // test settings always win 

Serializers

The JsonSerializer and UrlEncodedSerializer settings deserve special attention. As you might expect, they control the details of (de)serializing JSON requests and responses, and URL-encoded form posts, respectively. Both implement ISerializer, which defines 3 methods:

string Serialize(object obj);
T Deserialize<T>(string s);
T Deserialize<T>(Stream stream);

Flurl provides default implementations for both. It's very unlikely you'll need to use a different UrlEncodedSerializer, but there's a couple reasons you may want to swap out the JsonSerializer:

  1. You prefer the Newtonsoft-based version from 3.x and earlier. (4.0 replaced this with a System.Text.Json-based version.) This is available with the the Flurl.Http.Newtsonsoft companion package.

  2. You want to use the default implementation, but with custom JsonSerializerOptions. This can be accomplished by providing your own instance:

clientOrRequest.Settings.JsonSerializer = new DefaultJsonSerializer(new JsonSerializerOptions {
    PropertyNameCaseInsensitive = true,
    IgnoreReadOnlyProperties = true
});

Event Handlers

Keeping cross-cutting concerns like logging and error handling separated from your normal logic flow often results in cleaner code. Flurl.Http defines 4 events - BeforeCall, AfterCall, OnError, and OnRedirect - and an EventHandlers collection on IFlurlClient, IFlurlRequest, and IFlurlClientBuilder. (Unlike Settings, event handlers are not available on HttpTest.)

Similar to Settings, everything with an EventHandlers property brings with it some fluent shortcuts:

clientOrRequest
    .BeforeCall(call => DoSomething(call)) // attach a synchronous handler
    .OnError(call => LogErrorAsync(call))  // attach an async handler

In the example above, call is an instance of FlurlCall, which contains a robust set of information and options related to all aspects of the request and response:

IFlurlRequest Request
HttpRequestMessage HttpRequestMessage
string RequestBody
IFlurlResponse Response
HttpResponseMessage HttpResponseMessage
FlurlRedirect Redirect
Exception Exception
bool ExceptionHandled
DateTime StartedUtc
DateTime? EndedUtc
TimeSpan? Duration
bool Completed
bool Succeeded

OnError fires before AfterCall, and gives you an opportunity to decide whether to allow the exception to bubble up:

clientOrRequest.OnError(async call => {
    await LogTheErrorAsync(call.Exception);
    call.ExceptionHandled = true; // otherwise, the exeption will bubble up
});

OnRedirect allows precision handling of 3xx responses:

clientOrRequest.OnRedirect(call => {
    if (call.Redirect.Count > 5) {
        call.Redirect.Follow = false;
    }
    else {
        log.WriteInfo($"redirecting from {call.Request.Url} to {call.Redirect.Url}");
        call.Redirect.ChangeVerbToGet = (call.Response.Status == 301);
        call.Redirect.Follow = true;
    }
});

At a lower level, event handlers are objects that implement IFlurlEventHandler, which defines 2 methods:

void Handle(FlurlEventType eventType, FlurlCall call);
Task HandleAsync(FlurlEventType eventType, FlurlCall call);

Typically you'll only need to implement one or other, so Flurl provides a default implementation, FlurlEventHanler, that makes a handy base class - both methods are virtual no-ops, so just override the one you need. Handlers are assignable like this:

clientOrRequest.EventHandlers.Add((FlurlEventType.BeforeCall, new MyEventHandler()));

Note that EventHanlers items are of type Tuple<EventType, IFlurlEventHandler>. Keeping handlers decoupled from even types means a given handler could be reused for different event types.

One reason you might favor the object-based approach over the simpler lambda-based one described earlier is if you are using DI and your handlers need certain dependencies injected:

public interface IFlurlErrorLogger : IFlurlEventHandler { }

public class FlurlErrorLogger : FlurlEventHandler, IFlurlErrorLogger
{
    private readonly ILogger _logger;

    public FlurlErrorLogger(ILogger logger) {
        _logger = logger;
    }
}

Here's how this might be wired up using Microsoft's DI framework:

// register ILogger:
services.AddLogging();
// register service that implements IFlurlEventHander and has dependency on ILogger
services.AddSingleton<IFlurlErrorLogger, FlurlErrorLogger>();

// register event hanlder with Flurl, using IServiceProvider to wire up dependencies:
services.AddSingleton<IFlurlClientCache>(sp => new FlurlClientCache()
    .WithDefaults(builder =>
        builder.EventHandlers.Add((FlurlEventType.OnError, sp.GetService<IFlurlErrorLogger>()))

Message Handlers

Flurl.Http is built on top of HttpClient, which uses HttpClientHandler (by default) for most of its heavy lifting.IFlurlClientBuilder exposes methods for configuring both:

// clientless pattern:
FlurlHttp.Clients.WithDefaults(builder => builder
    .ConfigureHttpClient(hc => ...)
    .ConfigureInnerHandler(hch => {
        hch.Proxy = new WebProxy("https://my-proxy.com");
        hch.UseProxy = true;
    }));

// DI pattern: 
services.AddSingleton<IFlurlClientCache>(_ => new FlurlClientCache()
    .WithDefaults(builder => builder.
        .ConfigureHttpClient(hc => ...)
        .ConfigureInnerHandler(hch => ...)));

A few word of caution:

  1. Flurl disables AllowAutoRedirect and UseCookies in order to re-implement its own notion of these features. If you set either to true via ConfigureInnerHander, it may break these features in Flurl.
  2. Flurl also implements its own notions of DefaultRequestHeaders and Timeout, likely leaving few reasons that you'd ever need to configure the HttpClient.

Perhaps you've read about SocketsHttpHandler and are wondering how to use it in Flurl. You may be surprised to learn that you probably already are. As mentioned, FlurlClient delegates most of its work to HttpClient, which in turn delegates most of its work to HttpClientHandler. But what's lesser known is that, since .NET Core 2.1, HttpClientHandler delegates virtually all of its work to SocketsHttpHandler on all platforms that support it, which is basically everything except browser-based (i.e. Blazor) platforms. (Browse the source if you need convincing.)

FlurlClient →
    HttpClient →
        HttpClientHandler →
            SocketsHttpHandler (on all supported platforms)

Still, there is a reason that you may want to bypass HttpClientHandler and use SocketsHttpHander directly: some of its configurability is not available on HttpClientHandler. So long as you don't need to support Blazor, you can do this:

// clientless pattern:
FlurlHttp.Clients.WithDefaults(builder =>
    builder.UseSocketsHttpHandler(shh => {
        shh.PooledConnectionLifetime = TimeSpan.FromMinutes(10);
        ...
    }));

// DI pattern: 
services.AddSingleton<IFlurlClientCache>(_ => new FlurlClientCache()
    .WithDefaults(builder => builder.UseSocketsHttpHandler(shh => {
        ...
    })));

Note: Calling both ConfigureInnerHandler and UseSocketsHttpHandler on the same builder will result in a runtime exception. Use one or the other, never both.

The final type of message handler Flurl supports directly is DelegatingHandler, a chainable handler type commonly referred to as middleware and often implemented by third-party libraries.

// clientless pattern:
FlurlHttp.Clients.WithDefaults(builder => builder
    .AddMiddleare(new MyDelegatingHandler()));

// DI pattern: 
services.AddSingleton<IFlurlClientCache>(sp => new FlurlClientCache()
    .WithDefaults(builder => builder
        .AddMiddleware(sp.GetService<IMyDelegatingHandler>())

This example uses Polly, a popular resiliency library, along with configuring Flurl in a DI scenario:

using Microsoft.Extensions.Http;

var policy = Policy
    .Handle<HttpRequestException>()
    ...

services.AddSingleton<IFlurlClientCache>(_ => new FlurlClientCache()
    .WithDefaults(builder => builder
        .AddMiddleware(() => new PolicyHttpMessageHandler(policy))));

The above example requires installing Microsoft.Extensions.Http.Polly.


Spot an error or omission? Edit this page.