.net core 2.1中单例注入方法的区别

Question

What's the difference between these injection methods?

services.TryAdd(ServiceDescriptor.Singleton<IMyClass, MyClass>());

services.AddSingleton<IMyClass, MyClass>();

Answer 1

 services.TryAdd(ServiceDescriptor.Singleton<IMyClass, MyClass>()); 

Tries to register IMyClass , if its already registered, nothing will happen and the previous registration remains the only one. This results in one IMyClass registration.

 services.AddSingleton<IMyClass, MyClass>(); 

Will register IMyClass anyways, even if there is a different registration already such as <IMyClass, DefaultMyClass> . This results in two IMyClass registrations.

The difference comes, when using provider.GetService / provider.GetRequriedService or injecting IMyClass vs IEnumerable<IMyClass> .

When using provider.GetService / provider.GetRequriedService or injecting IMyClass , the last registration will be used (correct me if wrong, can't find the reference/GitHub issue on the quick).

So if you had

 services.AddSingleton<IMyClass, DefaultMyClass>();
 services.AddSingleton<IMyClass, MyClass>();

MyClass instance will be returned when resolved.

Whereas

 services.AddSingleton<IMyClass, DefaultMyClass>();
 services.TryAdd(ServiceDescriptor.Singleton<IMyClass, MyClass>());

would return DefaultMyClass , because <IMyClass,MyClass> would not be registered.

On the other side, when you use provider.GetServices(...) (notice the plural version, with an s at the end) or inject IEnumerable<IMyClass> , then you would return both instance when using

 services.AddSingleton<IMyClass, DefaultMyClass>();
 services.AddSingleton<IMyClass, MyClass>();

and you could iterate over them.