DotNetCore深入了解之三HttpClientFactory类

当需要向某特定URL地址发送HTTP请求并得到相应响应时，通常会用到HttpClient类。该类包含了众多有用的方法，可以满足绝大多数的需求。但是如果对其使用不当时，可能会出现意想不到的事情。

using(var client = new HttpClient())

对象所占用资源应该确保及时被释放掉，但是，对于网络连接而言，这是错误的。

原因有二，网络连接是需要耗费一定时间的，频繁开启与关闭连接，性能会受影响；再者，开启网络连接时会占用底层socket资源，但在HttpClient调用其本身的Dispose方法时，并不能立刻释放该资源，这意味着你的程序可能会因为耗尽连接资源而产生预期之外的异常。

所以比较好的解决方法是延长HttpClient对象的使用寿命，比如对其建一个静态的对象：

private static HttpClient Client = new HttpClient();

但从程序员的角度来看，这样的代码或许不够优雅。

所以在.NET Core 2.1中引入了新的HttpClientFactory类。

它的用法很简单，首先是对其进行IoC的注册：

1 public void ConfigureServices(IServiceCollection services)
2 {
3     services.AddHttpClient();
4     services.AddMvc();
5 }

然后通过IHttpClientFactory创建一个HttpClient对象，之后的操作如旧，但不需要担心其内部资源的释放：

 1 public class LzzDemoController : Controller
 2 {
 3     IHttpClientFactory _httpClientFactory;
 4 
 5     public LzzDemoController(IHttpClientFactory httpClientFactory)
 6     {
 7         _httpClientFactory = httpClientFactory;
 8     }
 9 
10     public IActionResult Index()
11     {
12         var client = _httpClientFactory.CreateClient();
13         var result = client.GetStringAsync("http://myurl/");
14         return View();
15     }
16 }

AddHttpClient的源码：

 1 public static IServiceCollection AddHttpClient(this IServiceCollection services)
 2 {
 3     if (services == null)
 4     {
 5         throw new ArgumentNullException(nameof(services));
 6     }
 7 
 8     services.AddLogging();
 9     services.AddOptions();
10 
11     //
12     // Core abstractions
13     //
14     services.TryAddTransient();
15     services.TryAddSingleton();
16 
17     //
18     // Typed Clients
19     //
20     services.TryAdd(ServiceDescriptor.Singleton(typeof(ITypedHttpClientFactory<>), typeof(DefaultTypedHttpClientFactory<>)));
21 
22     //
23     // Misc infrastructure
24     //
25     services.TryAddEnumerable(ServiceDescriptor.Singleton());
26 
27     return services;
28 }

它的内部为IHttpClientFactory接口绑定了DefaultHttpClientFactory类。

再看IHttpClientFactory接口中关键的CreateClient方法：

 1 public HttpClient CreateClient(string name)
 2 {
 3     if (name == null)
 4     {
 5         throw new ArgumentNullException(nameof(name));
 6     }
 7 
 8     var entry = _activeHandlers.GetOrAdd(name, _entryFactory).Value;
 9     var client = new HttpClient(entry.Handler, disposeHandler: false);
10 
11     StartHandlerEntryTimer(entry);
12 
13     var options = _optionsMonitor.Get(name);
14     for (var i = 0; i < options.HttpClientActions.Count; i++)
15     {
16         options.HttpClientActions[i](client);
17     }
18 
19     return client;
20 }

HttpClient的创建不再是简单的new HttpClient()，而是传入了两个参数：HttpMessageHandler handler与bool disposeHandler。disposeHandler参数为false值时表示要重用内部的handler对象。handler参数则从上一句的代码可以看出是以name为键值从一字典中取出，又因为DefaultHttpClientFactory类是通过TryAddSingleton方法注册的，也就意味着其为单例，那么这个内部字典便是唯一的，每个键值对应的ActiveHandlerTrackingEntry对象也是唯一，该对象内部中包含着handler。

下一句代码StartHandlerEntryTimer(entry); 开启了ActiveHandlerTrackingEntry对象的过期计时处理。默认过期时间是2分钟。

 1 internal void ExpiryTimer_Tick(object state)
 2 {
 3     var active = (ActiveHandlerTrackingEntry)state;
 4 
 5     // The timer callback should be the only one removing from the active collection. If we can't find
 6     // our entry in the collection, then this is a bug.
 7     var removed = _activeHandlers.TryRemove(active.Name, out var found);
 8     Debug.Assert(removed, "Entry not found. We should always be able to remove the entry");
 9     Debug.Assert(object.ReferenceEquals(active, found.Value), "Different entry found. The entry should not have been replaced");
10 
11     // At this point the handler is no longer 'active' and will not be handed out to any new clients.
12     // However we haven't dropped our strong reference to the handler, so we can't yet determine if
13     // there are still any other outstanding references (we know there is at least one).
14     //
15     // We use a different state object to track expired handlers. This allows any other thread that acquired
16     // the 'active' entry to use it without safety problems.
17     var expired = new ExpiredHandlerTrackingEntry(active);
18     _expiredHandlers.Enqueue(expired);
19 
20     Log.HandlerExpired(_logger, active.Name, active.Lifetime);
21 
22     StartCleanupTimer();
23 }

先是将ActiveHandlerTrackingEntry对象传入新的ExpiredHandlerTrackingEntry对象。

1 public ExpiredHandlerTrackingEntry(ActiveHandlerTrackingEntry other)
2 {
3     Name = other.Name;
4 
5     _livenessTracker = new WeakReference(other.Handler);
6     InnerHandler = other.Handler.InnerHandler;
7 }

在其构造方法内部，handler对象通过弱引用方式关联着，不会影响其被GC释放。

然后新建的ExpiredHandlerTrackingEntry对象被放入专用的队列。

最后开始清理工作，定时器的时间间隔设定为每10秒一次。

 1 internal void CleanupTimer_Tick(object state)
 2 {
 3     // Stop any pending timers, we'll restart the timer if there's anything left to process after cleanup.
 4     //
 5     // With the scheme we're using it's possible we could end up with some redundant cleanup operations.
 6     // This is expected and fine.
 7     // 
 8     // An alternative would be to take a lock during the whole cleanup process. This isn't ideal because it
 9     // would result in threads executing ExpiryTimer_Tick as they would need to block on cleanup to figure out
10     // whether we need to start the timer.
11     StopCleanupTimer();
12 
13     try
14     {
15         if (!Monitor.TryEnter(_cleanupActiveLock))
16         {
17             // We don't want to run a concurrent cleanup cycle. This can happen if the cleanup cycle takes
18             // a long time for some reason. Since we're running user code inside Dispose, it's definitely
19             // possible.
20             //
21             // If we end up in that position, just make sure the timer gets started again. It should be cheap
22             // to run a 'no-op' cleanup.
23             StartCleanupTimer();
24             return;
25         }
26 
27         var initialCount = _expiredHandlers.Count;
28         Log.CleanupCycleStart(_logger, initialCount);
29 
30         var stopwatch = ValueStopwatch.StartNew();
31 
32         var disposedCount = 0;
33         for (var i = 0; i < initialCount; i++)
34         {
35             // Since we're the only one removing from _expired, TryDequeue must always succeed.
36             _expiredHandlers.TryDequeue(out var entry);
37             Debug.Assert(entry != null, "Entry was null, we should always get an entry back from TryDequeue");
38 
39             if (entry.CanDispose)
40             {
41                 try
42                 {
43                     entry.InnerHandler.Dispose();
44                     disposedCount++;
45                 }
46                 catch (Exception ex)
47                 {
48                     Log.CleanupItemFailed(_logger, entry.Name, ex);
49                 }
50             }
51             else
52             {
53                 // If the entry is still live, put it back in the queue so we can process it 
54                 // during the next cleanup cycle.
55                 _expiredHandlers.Enqueue(entry);
56             }
57         }
58 
59         Log.CleanupCycleEnd(_logger, stopwatch.GetElapsedTime(), disposedCount, _expiredHandlers.Count);
60     }
61     finally
62     {
63         Monitor.Exit(_cleanupActiveLock);
64     }
65 
66     // We didn't totally empty the cleanup queue, try again later.
67     if (_expiredHandlers.Count > 0)
68     {
69         StartCleanupTimer();
70     }
71 }

上述方法核心是判断是否handler对象已经被GC，如果是的话，则释放其内部资源，即网络连接。

回到最初创建HttpClient的代码，会发现并没有传入任何name参数值。这是得益于HttpClientFactoryExtensions类的扩展方法。

1 public static HttpClient CreateClient(this IHttpClientFactory factory)
2 {
3     if (factory == null)
4     {
5         throw new ArgumentNullException(nameof(factory));
6     }
7 
8     return factory.CreateClient(Options.DefaultName);
9 }

Options.DefaultName的值为string.Empty。

DefaultHttpClientFactory缺少无参数的构造方法，唯一的构造方法需要传入多个参数，这也意味着构建它时需要依赖其它一些类，所以目前只适用于在ASP.NET程序中使用，还无法应用到诸如控制台一类的程序，希望之后官方能够对其继续增强，使得应用范围变得更广。

1 public DefaultHttpClientFactory(
2     IServiceProvider services,
3     ILoggerFactory loggerFactory,
4     IOptionsMonitor optionsMonitor,
5     IEnumerable filters)