当前位置 博文首页 > 不问前世：重新整理 .net core 实践篇————依赖注入应用之援
介绍第三方依赖注入框架Autofac,看看为我们解决什么问题。
下面介绍4个点:
命名注册
属性注册
aop 注入
子容器命名
为什么我们需要使用第三方框架？第三方框架为我们做了什么？第三方框架扩展了哪一个部分？
这里主要介绍一下Autofac。
Autofac 主要是替换了我们ServiceProviderFactory 这个东西。
public interface IServiceProviderFactory<IContainerBuilder>
我们使用的时候这样用：
Host.CreateDefaultBuilder(args).UseServiceProviderFactory(new AutofacServiceProviderFactory())
.ConfigureWebHostDefaults(webBuilder =>
{
webBuilder.UseStartup<Startup>();
});
看下UseServiceProviderFactory 源码:
private IServiceFactoryAdapter _serviceProviderFactory = new ServiceFactoryAdapter<IServiceCollection>(new DefaultServiceProviderFactory());
public IHostBuilder UseServiceProviderFactory<TContainerBuilder>(IServiceProviderFactory<TContainerBuilder> factory)
{
_serviceProviderFactory = new ServiceFactoryAdapter<TContainerBuilder>(factory ?? throw new ArgumentNullException(nameof(factory)));
return this;
}
所以说是替换了我们的serviceProviderFactory。
原理篇<<从新整理1400篇>>介绍了这个东西，这里再简要介绍一下。
_serviceProviderFactory 通过代理模式下进行的，也就是一个适配过程，那么我们直接看适配器。
其实适配器有一个小的隐藏信息哈。比如说_serviceProviderFactory的命名上看，翻译过来就是serviceProvider的构建工厂，也就是为我们提供serviceProvider的生产工厂。
直接看ServiceFactoryAdapter的源码。
internal class ServiceFactoryAdapter<TContainerBuilder> : IServiceFactoryAdapter
{
private IServiceProviderFactory<TContainerBuilder> _serviceProviderFactory;
private readonly Func<HostBuilderContext> _contextResolver;
private Func<HostBuilderContext, IServiceProviderFactory<TContainerBuilder>> _factoryResolver;
public ServiceFactoryAdapter(
IServiceProviderFactory<TContainerBuilder> serviceProviderFactory)
{
IServiceProviderFactory<TContainerBuilder> serviceProviderFactory1 = serviceProviderFactory;
if (serviceProviderFactory1 == null)
throw new ArgumentNullException(nameof (serviceProviderFactory));
this._serviceProviderFactory = serviceProviderFactory1;
}
public ServiceFactoryAdapter(
Func<HostBuilderContext> contextResolver,
Func<HostBuilderContext, IServiceProviderFactory<TContainerBuilder>> factoryResolver)
{
Func<HostBuilderContext> func1 = contextResolver;
if (func1 == null)
throw new ArgumentNullException(nameof (contextResolver));
this._contextResolver = func1;
Func<HostBuilderContext, IServiceProviderFactory<TContainerBuilder>> func2 = factoryResolver;
if (func2 == null)
throw new ArgumentNullException(nameof (factoryResolver));
this._factoryResolver = func2;
}
public object CreateBuilder(IServiceCollection services)
{
if (this._serviceProviderFactory == null)
{
this._serviceProviderFactory = this._factoryResolver(this._contextResolver());
if (this._serviceProviderFactory == null)
throw new InvalidOperationException("The resolver returned a null IServiceProviderFactory");
}
return (object) this._serviceProviderFactory.CreateBuilder(services);
}
public IServiceProvider CreateServiceProvider(object containerBuilder)
{
if (this._serviceProviderFactory == null)
throw new InvalidOperationException("CreateBuilder must be called before CreateServiceProvider");
return this._serviceProviderFactory.CreateServiceProvider((TContainerBuilder) containerBuilder);
}
}
看这个CreateServiceProvider方法，和猜想的一致。再来看下HostBuilder调用情况。
private void CreateServiceProvider()
{
var services = new ServiceCollection();
#pragma warning disable CS0618 // Type or member is obsolete
services.AddSingleton<IHostingEnvironment>(_hostingEnvironment);
#pragma warning restore CS0618 // Type or member is obsolete
services.AddSingleton<IHostEnvironment>(_hostingEnvironment);
services.AddSingleton(_hostBuilderContext);
// register configuration as factory to make it dispose with the service provider
services.AddSingleton(_ => _appConfiguration);
#pragma warning disable CS0618 // Type or member is obsolete
services.AddSingleton<IApplicationLifetime>(s => (IApplicationLifetime)s.GetService<IHostApplicationLifetime>());
#pragma warning restore CS0618 // Type or member is obsolete
services.AddSingleton<IHostApplicationLifetime, ApplicationLifetime>();
services.AddSingleton<IHostLifetime, ConsoleLifetime>();
services.AddSingleton<IHost, Internal.Host>();
services.AddOptions();
services.AddLogging();
foreach (var configureServicesAction in _configureServicesActions)
{
configureServicesAction(_hostBuilderContext, services);
}
var containerBuilder = _serviceProviderFactory.CreateBuilder(services);
foreach (var containerAction in _configureContainerActions)
{
containerAction.ConfigureContainer(_hostBuilderContext, containerBuilder);
}
_appServices = _serviceProviderFactory.CreateServiceProvider(containerBuilder);
if (_appServices == null)
{
throw new InvalidOperationException($"The IServiceProviderFactory returned a null IServiceProvider.");
}
// resolve configuration explicitly once to mark it as resolved within the
// service provider, ensuring it will be properly disposed with the provider
_ = _appServices.GetService<IConfiguration>();
}
的却如此。完全证实了这个猜想。
那么serviceProver 用来做什么的呢？
1: internal class ServiceProvider : IServiceProvider, IDisposable
2: {
3: public ServiceProvider Root { get; private set; }
4: public ServiceTable ServiceTable { get; private set; }
5: public ConcurrentDictionary<Type, Func<ServiceProvider, object>> RealizedServices { get; private set; } = new ConcurrentDictionary<Type, Func<ServiceProvider, object>>();
6: public IList<IDisposable> TransientDisposableServices { get; private set; } = new List<IDisposable>();
7: public ConcurrentDictionary<IService, object> ResolvedServices { get; private set; } = new ConcurrentDictionary<IService, object>();
8:
9: public ServiceProvider(IServiceCollection services)
10: {
11: this.Root = this;
12: this.ServiceTable = new ServiceTable(services);
13: }
14:
15: public object GetService(Type serviceType)
16: {
17: Func<ServiceProvider, object> serviceAccessor;
18: if (this.RealizedServices.TryGetValue(serviceType, out serviceAccessor))
19: {
20: return serviceAccessor(this);
21: }
22:
23: IServiceCallSite serviceCallSite = this.GetServiceCallSite(serviceType, new HashSet<Type>());
24: if (null != serviceCallSite)
25: {
26: var providerExpression = Expression.Parameter(typeof(ServiceProvider), "provider");
27: this.RealizedServices[serviceType] = Expression.Lambda<Func<ServiceProvider, object>>(serviceCallSite.Build(providerExpression), providerExpression).Compile();
28: return serviceCallSite.Invoke(this);
29: }
30:
31: this.RealizedServices[serviceType] = _ => null;
32: return null;
33: }
34:
35: public IServiceCallSite GetServiceCallSite(Type serviceType, ISet<Type> callSiteChain)
36: {
37: try
38: {
39: if (callSiteChain.Contains(serviceType))
40: {
41: throw new InvalidOperationException(string.Format("A circular dependency was detected for the service of type '{0}'", serviceType.FullName);
42: }
43: callSiteChain.Add(serviceType);
44:
45: ServiceEntry serviceEntry;
46: if (this.ServiceTable.ServieEntries.TryGetValue(serviceType,
47: out serviceEntry))
48: {
49: return serviceEntry.Last.CreateCallSite(this, callSiteChain);
50: }
51:
52: //省略其他代码
53:
54: return null;
55: }
56: finally
57: {
58: callSiteChain.Remove(serviceType);
59: }
60: }
61:
62: public void Dispose()
63: {
64: Array.ForEach(this.TransientDisposableServices.ToArray(), _ => _.Dispose());
65: Array.ForEach(this.ResolvedServices.Values.ToArray(), _ => (_ as IDisposable)?.Dispose());
66: this.TransientDisposableServices.Clear();
67: this.ResolvedServices.Clear();
68: }
69: //其他成员
70: }
看到GetService 是否特别的眼熟？这个就是通过我们的注册信息来产生不同的对象的。
那么Autofac的作用就是替换了我们的ServiceProvider。也就是替换了，往容器注入的方式了。
那么他给我们扩展了什么功能，后面就不介绍具体源码了，应为是实践篇，主要解释用法。
为什么有命名注册的方式呢？有什么痛点呢？
比如说:
services.AddSingleton<IMySingletonService>(new MySingletonService());
services.AddSingleton<IMySingletonService>(ServiceProvider =>
{
return new MySingletonService();
});
那么可以肯定一点的是，这时候通过IMySingletonService 可以获取第一个，也可以获取全部。
但是假如我要获取指定的一个呢？那么要给他们加上标记，但是.net core 自带的并没有并没有。
这时候使用autofac。
在startUp 中加入：
public void ConfigureContainer(ContainerBuilder builder)
{
builder.RegisterType<TestService>().As<ITestService>();
builder.RegisterType<TestService>().Named<ITestService>("service2");
}
然后在startUp 的Configure 中加入:
this.AutofacContainer = app.ApplicationServices.GetAutofacRoot();
var testServiceNameDefault = this.AutofacContainer.Resolve<ITestService>();
Console.WriteLine(testServiceNameDefault.GetHashCode());
var testServiceName2= this.AutofacContainer.ResolveNamed<ITestService>("service2");
Console.WriteLine(testServiceName2.GetHashCode());
发现他们的hashcode并不一致，故而获取了不同对象。
把这个TestService 修改一下:
public interface ITestService
{
public void ShowAttributeState();
}
public class TestService:ITestService,IDisposable
{
public AttributeService Attribute { get; set; }
public void ShowAttributeState()
{
Console.WriteLine($"attribute is null?{(Attribute == null ? "true":"false")}");
}
public void Dispose()
{
Console.WriteLine($"DisposableTestService Disposed:{this.GetHashCode()}");
}
}
AttributeService 如下：
public class AttributeService
{
}
AttributeService 的注册信息:
然后添加:
services.AddTransient<AttributeService>();
builder.RegisterType<TestService>().As<ITestService>();
注:这里我特意用services 来注册AttributeService 是为了证明autofac 兼容了.net core 原生的注册信息，证明前面的替换serviceProvider 的推导过程，这样我们就可以在我们的老项目中直接使用。
获取信息:
this.AutofacContainer = app.ApplicationServices.GetAutofacRoot();
var testServiceNameDefault = this.AutofacContainer.Resolve<ITestService>();
testServiceNameDefault.ShowAttributeState();
TestService的注册信息修改一下:
builder.RegisterType<TestService>().As<ITestService>().PropertiesAutowired();
设置自动注入属性，这样就可以了。如果对象属性有注册信息的话，会帮我们自动填充。
aop 如果没怎么看的话，可以简单理解可以理解为拦截器。
一般我们看到属性的方式来写加入拦截器:
[Attribuite]
public void method(){
}
这样是显式注入拦截器，那么autofac 为我们隐式注入。
安装一下Castle.core。然后我们写一个拦截器。
public class TestInterceptor : IInterceptor
{
public void Intercept(IInvocation invocation)
{
Console.WriteLine($"Invocation before Method:{invocation.Method.Name}");
invocation.Proceed();
Console.WriteLine($"Invocation after Method:{invocation.Method.Name}");
}
}
然后我们注册的时候注入拦截器。
安装一下:Autofac.Extras.DynamicProxy
写入注册信息:
builder.RegisterType<TestInterceptor>();
builder.RegisterType<TestService>().As<ITestService>().PropertiesAutowired().InterceptedBy(typeof(TestInterceptor)).EnableInterfaceInterceptors();
然后调用:
this.AutofacContainer = app.ApplicationServices.GetAutofacRoot();
var testServiceNameDefault = this.AutofacContainer.Resolve<ITestService>();
testServiceNameDefault.ShowAttributeState();
看下结果：
builder.RegisterType<TestService>().As<ITestService>().InstancePerMatchingLifetimeScope("selfScope");
然后调用：
this.AutofacContainer = app.ApplicationServices.GetAutofacRoot();
using (var myScope = AutofacContainer.BeginLifetimeScope())
{
var obj = myScope.Resolve<ITestService>();
Console.WriteLine(obj==null?"true":"false");
}
这里会报错myScope.Resolve
要这样写:
using (var myScope = AutofacContainer.BeginLifetimeScope("selfScope"))
{
var obj = myScope.Resolve<ITestService>();
Console.WriteLine(obj==null?"true":"false");
}
这么写好像是不能体现出这个容器命名有什么作用。
画一张概念图:
上述是原先.net core 的一个隔离思路。
如果给容器命名的话，相当于每个scope可以继续套娃，起一个隔离作用。
代码演示:
this.AutofacContainer = app.ApplicationServices.GetAutofacRoot();
using (var myScope = AutofacContainer.BeginLifetimeScope("selfScope"))
{
var obj = myScope.Resolve<ITestService>();
using (var myChildScope = myScope.BeginLifetimeScope())
{
var obj1 = myChildScope.Resolve<ITestService>();
Console.WriteLine(obj.GetHashCode());
Console.WriteLine(obj1.GetHashCode());
}
}
这种隔离机制做项目的时候就能体现出来，因为可能有几个服务共同用到了某个类，这样解决管理困难问题。
上述只是个人整理，如有问题，望请指出，谢谢。
下一节:配置之盟约。
