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.
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 :)
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.
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!
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!
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!
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.
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.
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!
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!!)
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.
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 !
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!!
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.
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!
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!
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!
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!
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!
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
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 ;)
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?
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?
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?)
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.
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.
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?
And how do you make that Wheather service testable? E.g Test that once we invoked the method that returns the data, the http got called once for example?
I'll heavily advocate for using Refit, as long as you don't mind the extra package (which is a completely valid reason to not use it), I don't see any common case where you couldn't use it. (Please correct me if you do !) It's completely abstracting all these concerns, which IMO should be abstracted by the Framework in the first place and gives you less code to write to handle deserialization / throwing exceptions. + It adds no complexity when pairing it with Polly. It's just great all around.
@@nickchapsas It is, for sure ! But not having to think about it on every new project is such a time/worry saver. Anyways, thanks Nick for your great work !
I'm a realive new Junior System Engineer. I've used some of this stuff in my code to connect to an API. (It works, but there is always room for improvement, right?) And now I'm sitting here, don't understanding a single word. I don't even understand which problem this solves. I feel dumb. ._.
Ok, now i got it. I showed it to my colleague, and EXPLAINED what the video is about. ...so I reallized, I understood the point. Creating an instance of a httpClient ONCE and REUSE it, instead of building everytime a new client can improve the performance. PS: Seems I'm not dumb, yay :3 xD
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
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?
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
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!
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.
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, 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."
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?
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.
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.
It really depends. For example you might have one client for the whole api and that’s fine but then you might wanna have two named clients for the same api but one configured with service to service access while the other one has some other form of access. As long as the httpclient is short lived, you can go as flexible as you want
I usually use the latter approach of injecting IHttpClientFactory, but I don't like defining the BaseUrl and default request headers in a completely different place as the rest of the logic, so I'll do that in the constructor. I also have a nitpicking code analyzer that complains about hardcoded urls so that will be coming from IConfiguration.
Thanks for the great content. What if I have like 20 services? I cannot inject the client in each one of them one by one. So, in that case, what is the best approach?
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?
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?
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
Thanks. I already had an opportunity to use the "new" method.
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.
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!
The best channel for C# I have found on UA-cam. Thankyou.
Thanks!
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!
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!
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.
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!
Can't believe I had this right before Nick telling me. I'm improving! Thanks Nick!
The most asked question on the job interviews so far. I’ve been asked 4 times about it during this summer.
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...
Best explanation I've seen, this is excellent. This problem has been painful! Subscribed, can't wait to learn about Polly.
Wow, I am amazed at how many videos you make. Respect.
Your explanation of the concept is very clear.
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.
Awesome, I am trying to edit already existing code and this gives me an idea why the things are the way they are.
You saved my life! Thank u so much! Greetings from Mexico!
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!
for the first time Nick discuss a topic I know and in my birthday
Thanks a whole lot for this clear explanation as to how to properly utilize HttpClient for API calls.
damn, what knowledge and presentation, SUBSCRIBED!
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!!)
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.
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 !
Nice job on this explanation Nick.
Learned this from Microsoft Docs, so used it like this already😃
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!!
Named HTTPClients with Polly here! But it took some massive headaches before I learned to do it that way.
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.
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!
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!
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.
this is a great subject. Can't wait to learn more about polly
Γειά σου βρε Χρήστο με τους μοναδικούς σου τρόπους!!!
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!
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!
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
Glad I paid attention to the correct way years ago. Had to because I was writing custom handlers.
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.
I was using this, works amazing.
Thanks, I have never thought earlier to use the httpClient the way you taught us. Thanks once again.
Great video. Cant wait for the microservices series (in case it is stil planned).
How can the same be achieved in .net framework 4.8 and below ..
Any suggestions
Ouuugh.. lucky you started to mention the Factory @4:44 (feared that Best Practice has changed again;)
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 ;)
How can we do this for .Net framework projects?
This came in handy, I am just about to implement HTTP functionality for my app.
I'm big into typed clients. That is the best way to do it most of the time in my opinion.
💯. Thanks for the awesome content 🫶🏼
Very well explained. Thank you!
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.
Great video Nick, thnx!
beautiful examples, 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?
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?
8:39 Could you dispose the HttpClient returned by CreateClient?
using var client = _httpClientFactory.CreateClient("weatherapi");
Yeah that’s totally fine
you can but you must not. HttpMessageHandler will take care the lifetime better than what you will.
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?
Great video, thanks. When you create the named client from the factory, shouldn't it have a USING? (9:07 line 25)
Fine. But what if client is separate application in different project?
thanks, i was just creating new instance always.
Excellent Video. Is it possible to provide the code as well.
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.
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.
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?
wow, that is really helpful. thanks
Please create a video on proper usage of Polly and how you are using it.
Is there an equivalent for smtpclient that lets you specify max pooled connections?
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?
And how do you make that Wheather service testable? E.g Test that once we invoked the method that returns the data, the http got called once for example?
I'll heavily advocate for using Refit, as long as you don't mind the extra package (which is a completely valid reason to not use it), I don't see any common case where you couldn't use it. (Please correct me if you do !)
It's completely abstracting all these concerns, which IMO should be abstracted by the Framework in the first place and gives you less code to write to handle deserialization / throwing exceptions.
+ It adds no complexity when pairing it with Polly.
It's just great all around.
Yeah I use Refit as well, but it’s good to know how things are wired out or should be wired up behind the scenes
@@nickchapsas It is, for sure !
But not having to think about it on every new project is such a time/worry saver.
Anyways, thanks Nick for your great work !
Curious how would I unit test that client class if the httpclient is a private property?
I'm a realive new Junior System Engineer. I've used some of this stuff in my code to connect to an API.
(It works, but there is always room for improvement, right?)
And now I'm sitting here, don't understanding a single word. I don't even understand which problem this solves. I feel dumb. ._.
Ok, now i got it. I showed it to my colleague, and EXPLAINED what the video is about.
...so I reallized, I understood the point.
Creating an instance of a httpClient ONCE and REUSE it, instead of building everytime a new client can improve the performance.
PS: Seems I'm not dumb, yay :3 xD
I was using it already :D
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
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?
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
LOVE IT!
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!
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.
Thanks a lot. but I have a question. If an API requires a Token, How can I add one in Startup file ?
great video
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?
at 3:55 you said "If the DNS DTL is expired..." what is DTL referring to here? Many thanks!
Sorry I said TTL, not DTL. TTL is Time To Live
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."
@nick How HttpClient was used in times of .net mvc when .net core is not introduced
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?
What is “circuit exhaustion”? Or am I mishearing
socket* exhaustion
I wonder if there is a clean way to access the config in AddHttpClient method.
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.
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
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.
What are best practices for the named clients? Should there always be a specified name? Should it be 1 name per base URL?
It really depends. For example you might have one client for the whole api and that’s fine but then you might wanna have two named clients for the same api but one configured with service to service access while the other one has some other form of access. As long as the httpclient is short lived, you can go as flexible as you want
I usually use the latter approach of injecting IHttpClientFactory, but I don't like defining the BaseUrl and default request headers in a completely different place as the rest of the logic, so I'll do that in the constructor. I also have a nitpicking code analyzer that complains about hardcoded urls so that will be coming from IConfiguration.
What if you have different baseUrl's calling different API's?
Thanks for the great content. What if I have like 20 services? I cannot inject the client in each one of them one by one. So, in that case, what is the best approach?
Inject the IHttpClientFactory and get them as named clients as shown in the video
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?
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?
Thanks a lot