紧跟上一篇文章。通过路由和动作匹配后,最终会得到跟当前请求最匹配的一个ActionDescriptor,然后通过IActionInvoker执行动作。
我们先来看一下IActionInvoker如何得到,代码如下:
context.Handler = (c) => { var routeData = c.GetRouteData(); //根据actiondescriptor实例化ActionContext对象 var actionContext = new ActionContext(context.HttpContext, routeData, actionDescriptor); if (_actionContextAccessor != null) { _actionContextAccessor.ActionContext = actionContext; } //创建IActionInvoker var invoker = _actionInvokerFactory.CreateInvoker(actionContext); if (invoker == null) { throw new InvalidOperationException( Resources.FormatActionInvokerFactory_CouldNotCreateInvoker( actionDescriptor.DisplayName)); } //执行invoker处理请求 return invoker.InvokeAsync(); };
从上面的代码可以看到,一个IActionInvoker是通过IActionInvokerFactory创建,框架里该接口实现类是ActionInvokerFactory,该类CreateInvoker方法实现代码如下:
public IActionInvoker CreateInvoker(ActionContext actionContext) { var context = new ActionInvokerProviderContext(actionContext); foreach (var provider in _actionInvokerProviders) { provider.OnProvidersExecuting(context); } for (var i = _actionInvokerProviders.Length - 1; i >= 0; i--) { _actionInvokerProviders[i].OnProvidersExecuted(context); } return context.Result; }
在方法中,首先实例化一个ActionInvokerProviderContext,然后调用IActionInvokerProvider来设置context.Result,context.Result就是一个IActionInvoker,所以我们跟踪下框架中IActionInvokerProvider实现类,来看看它里面是如何工作的。框架中的提供了一个实现类ControllerActionInvokerProvider,在OnProvidersExecuting方法中创建了IActionInvoker对象,代码如下:
public void OnProvidersExecuting(ActionInvokerProviderContext context) { if (context == null) { throw new ArgumentNullException(nameof(context)); } var actionDescriptor = context.ActionContext.ActionDescriptor as ControllerActionDescriptor; if (actionDescriptor != null) { var controllerContext = new ControllerContext(context.ActionContext); // PERF: These are rarely going to be changed, so let's go copy-on-write. controllerContext.ValueProviderFactories = new CopyOnWriteList<IValueProviderFactory>(_valueProviderFactories); controllerContext.ModelState.MaxAllowedErrors = _maxModelValidationErrors; var cacheState = _controllerActionInvokerCache.GetState(controllerContext); context.Result = new ControllerActionInvoker( _controllerFactory, _argumentBinder, _logger, _diagnosticSource, controllerContext, cacheState.Filters, cacheState.ActionMethodExecutor); } }
从上面的代码我们可以看到,最终就是实例化了一个ControllerActionInvoker对象,该类构造方法前几个参数不再介绍了,重点是最后两个参数,一个是IFilterMetadata[],表示跟当前动作相关联的过滤器信息集合,一个是ObjectMethodExecutor,从名字上可以看出,这个就是控制器方法的执行器。而这两个参数来自cacheState,这个对象是通过调用_controllerActionInvokerCache.GetState(controllerContext)得到的,它是一个ControllerActionInvokerState结构体类型,定义如下:
public struct ControllerActionInvokerState { public ControllerActionInvokerState( IFilterMetadata[] filters, ObjectMethodExecutor actionMethodExecutor) { Filters = filters; ActionMethodExecutor = actionMethodExecutor; } //动作过滤器集合 public IFilterMetadata[] Filters { get; } //方法执行器 public ObjectMethodExecutor ActionMethodExecutor { get; set; } }
cacheState创建的过程如下:
public ControllerActionInvokerState GetState(ControllerContext controllerContext)
{
var cache = CurrentCache;
var actionDescriptor = controllerContext.ActionDescriptor;
IFilterMetadata[] filters;
Entry cacheEntry;
if (!cache.Entries.TryGetValue(actionDescriptor, out cacheEntry))
{
var filterFactoryResult = FilterFactory.GetAllFilters(_filterProviders, controllerContext);
filters = filterFactoryResult.Filters;
var executor = ObjectMethodExecutor.Create(
actionDescriptor.MethodInfo,
actionDescriptor.ControllerTypeInfo);
cacheEntry = new Entry(filterFactoryResult.CacheableFilters, executor);
cacheEntry = cache.Entries.GetOrAdd(actionDescriptor, cacheEntry);
}
else
{
// Filter instances from statically defined filter descriptors + from filter providers
filters = FilterFactory.CreateUncachedFilters(_filterProviders, controllerContext, cacheEntry.FilterItems);
}
return new ControllerActionInvokerState(filters, cacheEntry.ActionMethodExecutor);
}
主要看下第10行到第15行代码。首先通过FilterFactory获取到跟当前动作相关联的过滤器信息集合,然后通过ObjectMethodExecutor.Create创建一个ObjectMethodExecutor对象,创建好后进行缓存。
其实到这里我们已经知道IActionInvokder就是一个ControllerActionInvoker,然后调用该对象的InvokeAsync方法开始动作执行。最新版本实现采用了状态机(如果概念上错误,欢迎大家拍砖指正)的特点,执行时在不同状态之间进行切换,最终完成处理。而这部分工作就是在Next方法里实现的。方法真正执行实在State.ActionInside状态时执行,在此状态下通过调用InvokeActionMethodAsync完成控制器方法的执行,主要代码如下:
//下面的代码就是根据方法返回值类型不同,实现不同的逻辑 var returnType = executor.MethodReturnType; if (returnType == typeof(void)) { executor.Execute(controller, orderedArguments); result = new EmptyResult(); } else if (returnType == typeof(Task)) { await (Task)executor.Execute(controller, orderedArguments); result = new EmptyResult(); } else if (executor.TaskGenericType == typeof(IActionResult)) { result = await (Task<IActionResult>)executor.Execute(controller, orderedArguments); if (result == null) { throw new InvalidOperationException( Resources.FormatActionResult_ActionReturnValueCannotBeNull(typeof(IActionResult))); } } else if (executor.IsTypeAssignableFromIActionResult) { if (_executor.IsMethodAsync) { result = (IActionResult)await _executor.ExecuteAsync(controller, orderedArguments); } else { result = (IActionResult)_executor.Execute(controller, orderedArguments); } if (result == null) { throw new InvalidOperationException( Resources.FormatActionResult_ActionReturnValueCannotBeNull(_executor.TaskGenericType ?? returnType)); } } else if (!executor.IsMethodAsync) { var resultAsObject = executor.Execute(controller, orderedArguments); result = resultAsObject as IActionResult ?? new ObjectResult(resultAsObject) { DeclaredType = returnType, }; } else if (executor.TaskGenericType != null) { var resultAsObject = await executor.ExecuteAsync(controller, orderedArguments); result = resultAsObject as IActionResult ?? new ObjectResult(resultAsObject) { DeclaredType = executor.TaskGenericType, }; } else { // This will be the case for types which have derived from Task and Task<T> or non Task types. throw new InvalidOperationException(Resources.FormatActionExecutor_UnexpectedTaskInstance( executor.MethodInfo.Name, executor.MethodInfo.DeclaringType)); } _result = result; }
无论是哪种返回值类型,都是通过调用executor.Execute完成控制器方法调用,并获取结果,executor就是上面提到的ObjectMethodExecutor,我们只看下Execute方法实现:
public object Execute(object target, object[] parameters) { return _executor(target, parameters); }
这里面又是一个_executor,它是一个delegate object ActionExecutor(object target, object[] parameters)委托类型,它是在实例化ObjectMethodInvoker时通过GetExecutor方法创建的,采用的是动态lambda表达式,有了这个委托对象,就可以执行控制器方法了。
方法执行后,会得到方法返回结果,然后进入Result相关状态流程,直到State.ResultInside状态时,调用InvokeResultAsync方法执行动作结果,具体代码如下:
protected async Task InvokeResultAsync(IActionResult result) { var actionContext = _actionContext; _diagnosticSource.BeforeActionResult(actionContext, result); try { await result.ExecuteResultAsync(actionContext); } finally { _diagnosticSource.AfterActionResult(actionContext, result); } }
到这里一个控制器动作就算执行完了,需要指出的是文章只介绍了主体流程,还有很多其他执行状态文章中没提到。