I have an IConfig
object that contains settings used throughout my application. At the moment, I inject the entire object into the constructor of each object that needs it, as follows:
public interface IConfig
{
string Username { get; }
string Password { get; }
//... other settings
}
public class Foo : IFoo
{
private readonly string username;
private readonly string password;
public Foo(IConfig config)
{
this.username = config.Username;
this.password = config.Password;
}
}
The downside is that IConfig
contains a large number of settings because it's deserialised from an overall config file, so injecting the entire object is unnecessary. What I'd like to do is change the constructor to Foo(string username, string password)
so that it only receives the settings it needs. This also makes creating Foo
objects for testing much easier (not having to set up IConfig
just to create Foo
). I'd like to bind the constructor arguments directly in my NinjectModule
, something like the following:
public class MyModule : NinjectModule
{
public override void Load()
{
Bind<IConfig>().To<JsonConfig>()
.InSingletonScope();
Bind<IFoo>().To<Foo>()
.WithConstructorArgument("username", IConfig.Username)
.WithConstructorArgument("password", IConfig.Password);
}
}
Obviously this code doesn't work, but how would I go about doing what I wanted?
My original idea was to use the NinjectModule.Kernel
to get the IKernel
then get an instance of my IConfig
object and inject the properties as required, but the object returned by NinjectModule.Kernel
has no Get<T>()
method.
IConfig
object created in the lambda contains all default values (empty strings for those settings). If I just do a normal Kernel.Get<IConfig> now that I've addedusing Ninject
this does work properly. Should I invest the effort to get thecontext
version working, is there any advantage? – Atlanta