I understand that a MEF CompositionContainer creates and keeps instances of classes. I don't know under what circumstances a CompositionContainer has a class instance in its bowels.
Can anybody list operations performed on the CompositionContainer or methods of the CompositionContainer class that cause the CompositionContainer to store an instance of a class within the CompositionContainer.
Is it possible to view class instances held within a CompositionContainer in the debugger or any other fashion?
The CompositionContainer will keep references to all shared parts for the lifetime of the CompositionContainer. (The default CreationPolicy is Any for both imports and exports, which means by default all parts will be shared unless otherwise specified.)
References to NonShared parts will be kept if the part implements IDisposable. The reference will be released when the root export that was pulled from the container is released (if that export was from a NonShared part). Exports can be released either by calling CompositionContainer.ReleaseExport, or ExportLifetimeContext.Dispose for exports created with an ExportFactory.
I don't think there's any simple way to view what's held by the CompositionContainer. The source code is available so you could theoretically dive into it and figure out exactly where it's stored.
In regards to your second question (#2 above) ...
Using the QuickWatch Window (Shift + F9) or a regular Watch window, copy the following in there:
((System.ComponentModel.Composition.Hosting.CompositionContainer)(this.Container))._catalogExportProvider._activatedParts
The line above assumes that the object you are stopped on has a "this.Container" property, which is the CompositionContainer of the scope you are referring to.
From there, you'll get an array of ActivatedParts. You then navigate the dictionary of Parts. Find the Part Definition you want to find the instance of, and expand its "Non-Public Members". There you will find the CachedInstance, and this will be the instance of your "Shared" exported part that's been created.
I think Parts that are exported NonShared and not IDisposable aren't cached or held onto at all. At least that's the behavior I've seen.
Related
I am developing a modular WPF application with Prism in .Net Core 5.0 (using MVVM, DryIoc) and I would like to have a module that is not a WPF module, i.e., a module with functionality that can be used by any other module. I don't want any project reference, because I want to keep the loosely coupled idea of the modules.
My first question is: is it conceptually correct? Or is it mandatory that a module has a screen? I guess it should be ok.
The second and more important (for me) is, what would be the best way to create the instance?
This is the project (I know I should review the names in this project):
HotfixSearcher is the main class, the one I need to get instantiated. In this class, for example, I subscribe to some events.
And this is the class that implements the IModule interface (the module class):
namespace SearchHotfix.Library
{
public class HotfixSearcherModule : IModule
{
public HotfixSearcherModule()
{
}
public void OnInitialized(IContainerProvider containerProvider)
{
//Create Searcher instance
var searcher = containerProvider.Resolve<IHotfixSearcher>();
}
public void RegisterTypes(IContainerRegistry containerRegistry)
{
containerRegistry.RegisterSingleton<IHotfixSearcher, HotfixSearcher>();
}
}
}
That is the only way I found to get the class instantiated, but I am not a hundred per cent comfortable with creating an instance that is not used, I think it does not make much sense.
For modules that have screens, the instances get created when navigating to them using the RequestNavigate method:
_regionManager.RequestNavigate(RegionNames.ContentRegion, "ContentView");
But since this is only a library with no screens, I can't find any other way to get this instantiated.
According to Prism documentation, subscribing to an event shoud be enough but I tried doing that from within my main class HotfixSearcher but it does not work (breakpoints on constructor or on the event handler of the event to which I subscribe are never hit).
When I do this way, instead, the instance is created, I hit the constructor breakpoint, and obviously the instance is subscribed to the event since it is done in the constructor.
To sum up, is there a way to get rid of that var searcher = containerProvider.Resolve<IHotfixSearcher>(); and a better way to achieve this?
Thanks in advance!
Or is it mandatory that a module has a screen?
No, of course not, modules have nothing to do with views or view models. They are just a set of registrations with the container.
what would be the best way to create the instance?
Let the container do the work. Normally, you have (at least) one assembly that only contains public interfaces (and the associated enums), but no modules. You reference that from the module and register the module's implementations of the relevant interfaces withing the module's Initialize method. Some other module (or the main app) can then have classes that get the interfaces as constructor parameters, and the container will resolve (i.e. create) the concrete types registered in the module, although they are internal or even private and completely unknown outside the module.
This is as loose a coupling as it gets if you don't want to sacrifice strong typing.
is there a way to get rid of that var searcher = containerProvider.Resolve<IHotfixSearcher>(); and a better way to achieve this?
You can skip the var searcher = part :-) But if the HotfixSearcher is never injected anywhere, it won't be created unless you do it yourself. OnInitialized is the perfect spot for this, because it runs after all modules had their chance to RegisterTypes so all dependencies should be registered.
If HotfixSearcher is not meant to be injected, you can also drop IHotfixSearcher and resolve HotfixSearcher directly:
public void OnInitialized(IContainerProvider containerProvider)
{
containerProvider.Resolve<HotfixSearcher>();
}
I am not a hundred per cent comfortable with creating an instance that is not used, I think it does not make much sense.
It is used, I suppose, although not through calling one of its methods. It's used by sending it an event. That's just fine. Think of it like Task.Run - it's fine for the task to exist in seeming isolation, too.
I've edited the question to make what I want to obtain clearer.
Here's the original question:
I'm working on a class that inherits from Control which will be used in my View and includes some Dependency Properties.
One of these DPs is an IEnumerable(Of RfidTag) and will be bound to an ObservableCollection(Of RfidTag) inside the VM.
The class RfidTag is defined as public class in the same file where the VM's class resides.
The questions are:
1. is it a good practice to expose a VM-related class to a Control class?
2. is the VM source file a good place for the RfidTag class?
UPDATE 1
In my application logic (I think I could say in my Model) there is an event published throught an Eventaggregator. The event's payload is a List of ModelRfidTag (that is the model class).
My VM subscribes to this event and I made the RfidTag class to wrap my Model class and provides some additional properties related only to the VM.
When the event handler inside the VM is executed, it makes an ObservableCollection(Of RfidTag) bindable from the view.
Then in the View I've a bounch of my control instances like that
<c:RfidTagPresenter
TagPosition="1"
Collection="{Binding RfidTagList, Mode=OneWay}" />
Then in my RfidTagPresenter (the class that inherits from Control) I've a DP of type RfidTag (called RfidTagResult) that returns the object in the OC which has the Position property (property available inside the RfidTag class) equal to the value set by the TagPosition DP.
In this way, the ControlTemplate of the RfidTagPresenter can bind its elements to the desired object.
This is the simplification of what I want to make. In the actual application there are some other DPs on which the RfidTagResult selection is performed
UPDATE 2
After a bit of research, seem that I can solve one problem with an indexed property that return (in the get method) the object from the collection I want to bind.
However a problem still exist: My control need to have a DP of type of RfidTag so that the relative ControlTemplate can bind to the property declared in the RfidTag class.
So: Is it possible (read: a good practice) to have a DP of a type that is a VM related class?
In other words: Can a custom control know about the class type used by the VM?
I will try to go all over your question (if I miss something let me know) but first you should explain the purpose of binding a Collection in a control as a DP.
is it a good practice to expose a VM-related class to a Control class?
RfidTag, I suppose, is a Model. What you are really doing here is binding a Model in your control which go against the MVVM pattern. You should think about the next question... Do you really need all your RfidTag to be shown in the View ? If you need to show a name, an ID... you could just create an IEnumerable<string> as DP (which is correct) and then in your VM instead of an ObservableCollection<RfidTag> you would have an ObservableCollection<string>.
Some simple theory. In MVVM, VM adapts the Model to the View. So your VM should have everything that will be shown in your View.
is the VM source file a good place for the RfidTag class?
RfidTag is a Model so there's no better place for it :)
I have an object instance, and I want to end up with a MEF catalog that contains that object instance, exported as a specific interface type. How can I do this?
TypeCatalog doesn't seem workable here, because (a) it creates a new instance instead of using an existing one, and (b) it requires the type to have an [Export] attribute. In my case, the instance comes from MEF's metadata system, so MEF creates the underlying type and I can't add attributes to it.
As far as I can tell, the usual advice is, if you've got an existing instance, you should add it to the container (e.g. via CompositionBatch), not to the catalog. But when I add this instance, I'm also adding an entire AssemblyCatalog worth of types, all in the same operation. I'll also want to be able to remove all of these types later. It makes more sense to me to bundle everything into an AggregateCatalog. That way, I can add both the assembly and the instance in one atomic operation, and I can remove them all again the same way.
For example:
// Bootstrapper code to initialize MEF:
public void Configure() {
_selectedGameCatalog = new AggregateCatalog();
var globalCatalog = new AggregateCatalog(_selectedGameCatalog);
_container = new CompositionContainer(globalCatalog);
// ... more MEF initialization ...
}
// Sometime later, I want to add more stuff to the MEF ecosystem:
public void SelectGame(Lazy<Game, IGameMetadata> entry) {
var newCatalog = new AggregateCatalog();
// Make the assembly available to import:
newCatalog.Catalogs.Add(new AssemblyCatalog(entry.Value.GetType().Assembly));
// I also want the metadata to be available to import:
IGameMetadata metadata = entry.Metadata;
newCatalog.Catalogs.Add(MakeCatalogFromInstance<IGameMetadata>(metadata));
// Replace whatever game was selected before:
_selectedGameCatalog.Catalogs.Clear();
_selectedGameCatalog.Catalogs.Add(newCatalog);
}
The part I don't know how to do is "MakeCatalogFromInstance". How can I create a catalog that contains an existing instance (registered as a specific type)?
Or, alternatively, if I'm going about this all wrong, is there a better way to plug an entire catalog and an existing instance all into MEF at the same time, with the ability to unplug them all again later and replace them with something else?
I think it's probably best to add the types to the catalog and then add the instance to the container.
Catalogs contain part definitions. Part definitions are used to create parts. (The types for this are ComposablePartDefinition and ComposablePart.) So you could theoretically write your own catalog and a part definition that always returned a part corresponding to the instance when CreatePart was called. But catalogs weren't really designed to be used this way.
For prosperity...
MEF devivides the chores of what type info is to be used (catalog) from the actual running object instances (container). To me it is a logical descicion, especially when you setup a more complex MEF environment in your application.
If you want the ability to 'change' containers on the fly, I would suggest you try to use hierarchical containers. The root catalog/container is filled with static types and any of the child containers can be filled with each specific set of meta types you need for your game.
Hope it helps,
Marc
I'm learning prism V4 using MEF to load my modules. Loading modules does work, but in one module I want to load a View/ViewModel (MVVM) and don't really know how I get MEF to resolve all this stuff for me.
First: how do I need to mark the ViewModel (I follow the StockTraderRI example) so it is not loaded on startup but instead can be loaded during runtime into a region?
Second: how do I load the ViewModel using MEF so it gets connected to the corresponding interfaces?
MEF does this very nicely for things on startup which are marked as [Export], but I got no idea how to achieve this during runtime.
You can use what is known as a Lazy Export so that the interface is not resolved until you explicitly use it.
If you need to create multiple instances, MEF doesn't support this particularly well. You can either do your own discovery and instantiation, or you can define the Export like this:
[PartCreationPolicy(CreationPolicy.NonShared)]
[Export(typeof(ISomething)]
public class Something : ISomething { }
The downside is that now wherever you need to create the instance, you need to have a reference to the actual Container instance. Then you can do:
var something = _container.GetExportedObject<ISomething>();
EDIT: Okay, I think I understand better what you're after. Here is how I've typically resolved this issue:
I implement my View objects as UserControl instances and don't set a DataContext anywhere in their code or XAML.
I create a DataTemplate that binds from the Type of the ViewModel to the UserControl.
On my MainViewModel (or whatever corresponds to the View hosting the regions), I expose a general RegionX Object (possibly typed to an interface if all of my ViewModels will share some common functionality, but Object works fine).
I create a ContentPresenter with Content bound to the RegionX property.
Now my MainViewModel can import different ViewModel instances corresponding to the types of ViewModels that might be hosted by the RegionX. When I want to switch the 'active' View in the region, I simply set RegionX to the corresponding ViewModel.
i'm playing with MEF and in the example i see this code ( i call it the MEF compose code):
var catalog = new AssemblyCatalog(System.Reflection.Assembly.GetExecutingAssembly());
var container = new CompositionContainer(catalog);
container.ComposeParts();
In most examples this is done in the same source file as the Startup Program class lives, as well as all of the other interfaces and classes.
Now i want to use the MEF, but i wonder what this container does. As far as i guessed it does the export / import mapping, but what if i have this code (from a windows forms app:
[STAThread]
static void Main()
{
Application.Run(new Form1());
}
and in that Form1 i want to create an object of type Auditer (which has a property of type ILogger marked import, and i have a class implementing ILogger marked export).
Do i have to place the MEF compose code in the void Main(), in the Load event of Form1, or in the constructor of the Auditer class?
I only get it to work when i put it (the MEF compose code) in the constructor of the Auditer class, but the examples i read somehow give me the idea that you only have to call the compose code once.
The CompositionContainer is the class that actually composes your parts in MEF.
When you want to use MEF, you need to always compose the part that's attributed with the import definitions. If you have a property in Form1 that's marked with [Import(typeof(ILogger))], at some point, you'll need to compose your Form1 instance.
The CompositionContainer is the class that actually performs this composition. It finds the appropriate exported ILogger based off the Catalog(s) contained within the container, and constructs types, matches exports to the imports, etc.
The reason that the MEF samples only "compose" one time is that, often, with DI, you'll have a single instance of the container constructed and setup at the beginning of the application, and it will compose your "main" window. All other classes will be composed automatically if they're being used as part of the main window. (For example, if Form1 composes an ILogger, but your ILogger implementation has an [Import] of it's own, it too will get composed in that pass.)
That being said, there is no fixed rule that specifies you can't compose more than once. In WPF and Silverlight, for example, it's frequent that MEF can't construct your object, which means it can't automatically compose your object for you. In these situations, a common pattern is to use the CompositionInitializer (in the box in Silverlight, not in the desktop) to have parts compose themselves, based off a static catalog. I recently blogged about this approach for WPF.
With Windows Forms, this may be less necessary, since there isn't a third party product (the XAML parser) constructing your types. However, you could still use this same approach, if you so choose.