Stop using the HttpClient the wrong way in .NET
Вставка
- Опубліковано 14 сер 2022
- Check out my courses: dometrain.com
Become a Patreon and get source code access: / nickchapsas
Hello everybody I'm Nick and in this video I will show you why the way you might be using the HttpClient in .NET could be completely wrong and then follow it up by showing you the right way to implement it.
Workshops
NDC Oslo | 26 - 30 Sept | bit.ly/ndcoslo2022workshop
NDC Sydney | 10 - 14 Oct | bit.ly/ndcsydney2022workshop
dotnetdays | 20 - 22 Oct | bit.ly/dotnetdays2022workshop
NDC Minnesota | 15 - 18 Nov | bit.ly/ndcminnesota2022workshop
NDC London | 23-27 January 2023 | bit.ly/ndclondon2023
Don't forget to comment, like and subscribe :)
Social Media:
Follow me on GitHub: bit.ly/ChapsasGitHub
Follow me on Twitter: bit.ly/ChapsasTwitter
Connect on LinkedIn: bit.ly/ChapsasLinkedIn
Keep coding merch: keepcoding.shop
#csharp #dotnet #nickchapsas
Hello everybody, I forgot to mention something in the video. If you add a service with a specific lifetime in the DI container and then you use the typed client versions of AddHttpClient (not the other ones for the IHttpClientFactory) then the service will be rewired as a Transient. If you need a singleton, or even scoped service, use the IHttpClientFactory which will guarantee that.
If you re register the service, then the http client configs are lost… I was using aws X-ray and experienced that… the best way to deal with this is really using named clients. So you create the named client on configure, and you inject the factory to your singleton service (or whatever scope you want) and through the factory you get the client just like you did on the end of the video. Do not register the service twice.
@@devtekve1396 Exactly this. Same happened to me.
Flurl
Thanks, I was wondering through the video how it works if the WeatherClient is singleton and therefore keeps all the instances for a lifetime, this explains a lot :)
This was actually what let me scroll back in the video a dozen times until I saw your comment 😂 always this DI dark magics in background
Hey Nick, great video! I just wanted to say I love that you deliver videos on a constant schedule basically at the same time every week. I know that when I wake up on Monday mornings and am getting geared up for work I have a Nick Chapsas video to get my head in the zone, inspired to code. Thank you for creating these videos!
Note you have introduced a small unintended change in your app when you do AddHttpClient
Good point, totally forgot to mention that after I got sidetracked with the IHttpClientFactory example. Will add it in the pinned comment thanks.
@@nickchapsas Lets hope people read the comments then!
Can I contact you on any social media? Would really love to learn and understand .net
@@davideglass We do.
@@LiamLagan You are a gentelman and a scholar!
FINALLY I get a Nick Chapsas video where I can say "oh yeah, I knew that 😎"
At my last job I was the driving force of getting the factory (using named clients) as THE way to use HttpClient. At my current job, it is much more of a struggle, but we will get there.
Ive watched so many of your videos brother, this was the first time I was already doing what you suggest. Thank you so much for you content, it helps me all the time!
I've been using IHttpClientFactory since I learned about the reason behind it. So many developers are unaware of its need and the problem it solves still.
The best channel for C# I have found on UA-cam. Thankyou.
Great video! I wanted to note that, for Blazor components - with a code behind that is a partial class, instead of:
private readonly IHttpClientFactory _httpClientFactory;
public MyClass(IHttpClientFactory httpClientFactory)
{
_httpClientFactory = httpClientFactory;
}
you use:
[Inject]
public IHttpClientFactory _httpClientFactory { get; set; }
Hope this helps the Blazor folks!
In my current (newest) project I used the static verstion you descriped. I've updated it to instanciate it from the builder without any problems. Thank you for sharing!
Best explanation I've seen, this is excellent. This problem has been painful! Subscribed, can't wait to learn about Polly.
Thanks. I already had an opportunity to use the "new" method.
Your explanation of the concept is very clear.
Nice job on this explanation Nick.
You saved my life! Thank u so much! Greetings from Mexico!
Very well explained. Thank you!
Can't believe I had this right before Nick telling me. I'm improving! Thanks Nick!
Awesome, I am trying to edit already existing code and this gives me an idea why the things are the way they are.
Wow, I am amazed at how many videos you make. Respect.
Going to have to try this, I just had to build an in house authentication api and we were wondering exactly how to handle the client component of it. Thanks as always!!
The most asked question on the job interviews so far. I’ve been asked 4 times about it during this summer.
Great video Nick, thnx!
I want to warn you against using a static HttpClient. You will have problems if:
1. Multi-threaded use.
2. Outgoing connection to different endpoints.
3. Use different headers.
HttpClient has internal state (fields). And if different headers are initialized in different threads, you will either have dirty data errors or even NREs.
It's such a relief to find out I've been doing it the right way. I was a bit nervous there for a while...
beautiful examples, thanks :)
Thanks a whole lot for this clear explanation as to how to properly utilize HttpClient for API calls.
Thanks for bringing awareness to this sometimes silent issue. I would say the confusion have been due to how this type has evolved between runtimes over the years. Everything you explained can be found in the official docs around HttpClient. Microsoft have done a great job there.
Hi Nick, well done on getting signed up on all the conferences, you deserve it as you are a fantastic ambassador... i hope they are looking after you :) Business Class I mean !
Named HTTPClients with Polly here! But it took some massive headaches before I learned to do it that way.
for the first time Nick discuss a topic I know and in my birthday
💯. Thanks for the awesome content 🫶🏼
Thanks Nick for a great video.. it would be great if you could do a follow-up video about shared CookieContainer issue and how to work around that.
I'm using HttpClientFactory with named clients and poly (because I learnt it from you) in the next integration to be deployed. Made things so much cleaner, especially as I use multiple clients with different header configurations. I had no clue about the DNS issue, feels like I dodged a bullet here.
wow, that is really helpful. thanks
this is a great subject. Can't wait to learn more about polly
Glad I paid attention to the correct way years ago. Had to because I was writing custom handlers.
I was using this, works amazing.
What a coincedent, as today I was developing a reusable library which needs to connect the httpclient with a client certificate. The problem I was facing is that the certificate to use on production needs to be fetched from azure keyvault and for local development I don't want to use the keyvault (that will be done with User Secrets). Actually I don't want to have a dependency on the keyvault packages in the reusable library at all.
I was already looking into the IHttpClientFactory and your video will surely help me tomorrow finalize this job. Thanks!
Learned this from Microsoft Docs, so used it like this already😃
Γειά σου βρε Χρήστο με τους μοναδικούς σου τρόπους!!!
Thanks!
Great video. Cant wait for the microservices series (in case it is stil planned).
This came in handy, I am just about to implement HTTP functionality for my app.
Thanks a lot
Great video! One thing I've wondered when using IHttpClientFactory is what the best practices are when it comes to guarenteeing that a named client, which is required by a service, has actually been configured.
Currently I'm using an extension method which registers both the named client and the service that requires it - but I'm wondering if there is a better way?
We use the AddHttpClient method a lot in our services, although lately we have been moving to using Flurl and injecting an IFlurlClientFactory, which you then use to request a client. The concept, as far as I can tell, is the same as injecting an IHttpClientFactory. But your comment about IHttpClientFactory and its integration with Polly has me intrigued! I may have to check that out (unless you want to do a video for it next!!)
LOVE IT!
thanks, i was just creating new instance always.
In a project I coded for we used named httpclients. I refactored that to use typed httpclients and also configured a polly retry policy in the startup for one of the httpclients. The basic auth username and password are also configure in the startup for one of de httpclients that needed to login.
Great video Nick. And I was glad to hear u shout out Raw Coding in a earlier vid. I subscribe to both of ur channels as well as IAmTimCorey, Les Jackson and Shiv Kumars' So u think u know C#. Keep up the great work. Oh and very glad to see u a NDC Conferences!
great video
Every company I've worked for uses restclient. I've not looked at how it's used under the hood either. But I believe earlier iterations of restclient suffered some of these issues. You've explained and shared things here I was none the wiser. Thank you
This is properly useful information; like many I was just instantiating and using a single HttpClient instance without considering things like DNS lifetimes. This also reminds me why I have a love/hate relationship with DI - it makes things easier in a lot of instances, but sometimes I feel like it hides a bit too much and lacks intuitiveness.
Great video, and loved your NDC talk about Minimal APIs by the way - I'm a convert!
Why? An object that consumes a HTTPClient shoulnd't be concerned with all that stuff.
Thanks, I have never thought earlier to use the httpClient the way you taught us. Thanks once again.
Great video. Would be interesting to see your proposed solution for implementing nswag api client.
I have used the IHttpClient method with named http clients.
I use a slightly different method. I use an explicit factory lambda (mainly for compile time type checking) and in that factory use the httpclientfactory to create the named httpclient to pass to the class keeps all the factories together. Works well.
Nick great video but you missed the most important thing. Why you can get error pool is exhausted. It will definitely help people understand why it's important to reuse HttpClient and what exactly happened when you do not ;)
You're the official author of C# compiler source and you're the only one because you wrote the compiler yourself! That explains a lot no other way around that you know so much!
I was using it already :D
Damn this was well invested 10 minutes of watch time. Subscribing and i'll probably become a patreon to look at the code in more detail.
Thank you for this Great video. For me i use Refit with HttpClientFactory instead, it's a library provides a type-safe wrapper for interacting with HTTP-based APIs.
Hi Nick, thanks for sharing the great tutorial to handle HttpClient.
I am wondering, is there anyway to add new Http header into the existing HttpClient after the DI is initialized?
Great video, thanks. When you create the named client from the factory, shouldn't it have a USING? (9:07 line 25)
thanks. i've been using the typed HttpClient. i didn't know that it is implemented using the factory, and that there could be problems when using my own instance
Ouuugh.. lucky you started to mention the Factory @4:44 (feared that Best Practice has changed again;)
Please create a video on proper usage of Polly and how you are using it.
Thanks for sharing! So I can have as many http-named or typed clients in my program.cs and use them wherever I need them in my app?
How can the same be achieved in .net framework 4.8 and below ..
Any suggestions
That DNS issue on static or singleton normal new() clients has a nasty failure mode. Everything working fine for weeks and then suddenly not anymore. Bit me in the rear once when I knew a little less about this stuff.
I'm using the HttpClient with the factory by coincidence because I wanted to mock it in unit tests. I'm glad this is also the recommended way, thanks for explaining why!
Good video, I've been using the IHttpClientFactory already but a non specified one, just using the AddHttpClient(). Is there any drawback except that you need to configure base url etc for it each time you use it?
So what's the set-up when you want to use custom HttpClientHandlers? Is it simply more configuration for your services? Or would you be using a custom factory for that?
Hi @Nick. Great video. Just had a quick question on this. You don't need that AddSingleton when you added the line AddHttpClient line in program.cs, right? Because that AddHttpClient will register the OpenweatherClient type in DI as transient, right? Can you please confirm? Also, if I have a separate class library services project which my API project calls and all my HttpClient calls are in the services layer, what is the best way to configure/ inject HttpClient to the service layer classes? Thanks!
But how is this unit testable? Does the CreateClient method in the factory return an interface or the HttpClient itself? I use Refit which provides a Library that uses IHttpclientfactory behind the scenes. What do you think is Refit a good way to go this time?
nice one
Good one nick, wondering is there any best way to call SOAP service like similar fashion ? taking advantage of Httpclientfactory ? I believe calling SOAP services will have the same impact on connections if its not handled properly.
If you are requesting data from multiple APIs, would you create separate HttpClients each with their own base url, or still a single HttpClient that can call any url?
amazing
Great video. Polly is such a great tool and time saver. Are you using it in any of your courses?
I will be using it in my REST API course
How do you see basic implementation of classes like HttpClient in this video? I have seen you cycle through code.
Or is this an extension on Rider of how things work?
How can we do this for .Net framework projects?
If you are retrying an API post or put you should make sure it supports idempotent requests. You don't want to retry make payment with stripe 10 times because the client timed out
Perfect! I'm just working on a project where I need the HttpClient or two actually. Exactly what I needed :D
Edit: Actually I'm using RestSharp, I wonder if this works also this way...
Hi,
What's the difference between using GetFromJsonAsync vs using
ReadAsStreamAsync and then working with the stream either via stream-reader or by deserializing the stream directly with JsonSerializer.Deserialize(stream)
Is there any performance/memory usage difference between the two approaches?
Thanks
What's the best way to make HTTP requests from within a library (distributed as a .dll or on NuGet), where the application using it is not necessarily a web application and might not have dependency injection available? I imagine there's some way to let the consumer of the library provide their own objects for request handling, but I'm not sure what the most sensible defaults should be if they don't do so (or, should they be forced to do so by making it a required argument?)
Did you find a suitable approach for this?
Hi Nick, great video, I have on question tough. About the dns problem when the connection expires, what if the disconnection didn't come from a timeout, but were disconnected by the server, is there any way to get notified that before doing a new request ?
please create video on blazor vs react native. I love to work with c#. I am also learning javascript for nodjs . how would take the market demand c# as compare to javascript?
My scenario. There are more different URLs with different authentication headers in database. Of course, users can add/update/remove these endpoints. Each endpoint is POST method and the content which is sent is also the same.
When a new message arrives in the application (and is processed), then it should be sent to all defined endpoints (which are currently defined in database).
How to efficiently send this message to endpoints in parallel using HttpClient?
Thank you for suggestions.
Fine. But what if client is separate application in different project?
So many stuff to do to make just one api call in C#... Lots of videos showing different 20 line examples on how to do a get call. JS is a ninja
How would I do this for projects like Specflow? Do I define a httpClientFactory and register it in the built in DI container and then return a httpClient so it can be injected in my classes?
Thanks a lot. but I have a question. If an API requires a Token, How can I add one in Startup file ?
How to add a outgoing request body logger and incoming response logger?
Great video, I've used IhttpclientFactory multiple times , but I have one issue that I encounter when I'm using it inside a wpf application to call an Api, it seems that the http client cant detect when the internet connection is restored, for now the only solution is to reinstantiate http client each time when the connection is restored, is there any way to configure httpclient to detect internet connection when it is restored while using IhttpclientFactory
Question: In some cases I had to bypass some ssl certificate using a client handler, the handler goes in the constructor of the httpclient, so, how to incorporate that in this example
I have always used HttpClient with the using pattern. Good to know that there is a factory solution for that... can this be used with Winforms / WPF as well?
Hi Nick,
Do you know if setting the RequestTimeout for a HttpRequestMessage object will always be honored by the server or not?
Sample code:
var request = new HttpRequestMessage();
request.Method = HttpMethod.Post;
request.RequestUri = new Uri(url);
request.Properties["RequestTimeout"] = timeout;
I was working on some task and I stubmbled on this when I was doing some research regarding request timeout:
GPT:
"Yes, the method MakePost you've defined will use the timeout specified in the TimeSpan argument to set a timeout for the HTTP request. Specifically, it sets the timeout using the request.Properties["RequestTimeout"] property.
The timeout value is passed as a TimeSpan and is used to set the request timeout. When making the HTTP request, the timeout will be applied, and the request will timeout if it exceeds the specified duration.
Keep in mind that setting a timeout on the HttpRequestMessage doesn't guarantee that the server will honor it. The server may choose to ignore the client's timeout and handle the request in its own time."
Microsoft should have just put the HttpClientFactory functionality into the HttpClient code to manage "circuit exhaustion" and DNS caching. Imagine if every library in .NET had to be instantiated with a Factory to deal with problems built into the default library.
At this point, HttpClient should be deprecated, and a replacement API with less foot-shooting likelihood promoted. Alas...
Unfortunately, this is far from the only IDisposable that's not really (appropriately) disposable, e.g. Task.
What is “circuit exhaustion”? Or am I mishearing
🥵🤯😓 Sooooo much to know... Im a beginner watching this blowing my head.
Same here 😂
Nick, Can you make a video regarding thé header propagation. In my company de try to use that with the method AddHeaderPropagation. Take Care. 👌👌