I'm currently switching from .net framework to .net core 3.1.
Defining Api Controllers inside the namespace is all fine and works.
Now I have the case, that I need to declare the Api Controllers within another class, like this:
namespace Api.Controllers
{
public class MainClass : BaseClass
{
public MainClass()
{
}
[ApiController]
[Route("Test")]
public class TestController : ControllerBase
{
[HttpGet]
public int GetResult()
{
return 0;
}
}
}
}
The result is, that the Api Controller can't be found after calling the "AddControllers" method inside the "ConfigureServices" method of the startup class.
The MainClass is instantiated before the Startup class will be called.
I've tried to change the global route and defining an area like "{area=Test}/{controller=Test}", or set the ApiController attribute above the MainClass, but none of them worked.
Is there a way to get this working?
Looks like the default ControllerFeatureProvider does not treat nested controller types as controller. You can add (don't need to replace) your custom provider to change that behavior, like this:
public class NestedControllerFeatureProvider : ControllerFeatureProvider
{
protected override bool IsController(TypeInfo typeInfo)
{
if(!typeInfo.IsClass) return false;
if(typeInfo.IsAbstract) return false;
var isNestedType = typeInfo.DeclaringType != null;
var isPublic = true;
var parentType = typeInfo.DeclaringType;
while(parentType != null){
isPublic = parentType.IsPublic;
parentType = parentType.DeclaringType;
}
return isNestedType && isPublic;
}
}
Then add that provider to the ApplicationPartManager in Startup.ConfigureServices like this:
services
.AddMvc()
.ConfigureApplicationPartManager(appPart => {
appPart.FeatureProviders.Add(new NestedControllerFeatureProvider());
});
If you want to replace the default ControllerFeatureProvider, just find it in the FeatureProviders and remove it. Of course then you need to ensure that your custom one should handle everything just like what done by the default logic, something like this:
//for IsController
return base.IsController(typeInfo) || <...your-custom-logic...>;
NOTE: You can refer to the default implementation of ControllerFeatureProvider to learn some standard logic to implement your own logic correctly. The code above is just a basic example. To me, as long as the classes inherits from ControllerBase and not abstract, they can work fine as a controller to serve requests. There would be no serious troubles except some weird conventions (e.g: class name not ending with Controller is still a controller or some standard attributes applied on the controller class are not working ...).
We should not use nested controller classes. Each controller class should be put in a separate file (as a good practice). However the point of this answer (the most interesting part that I'm pretty sure not many know about, is the use of ControllerFeatureProvider which can help you customize the features set in other scenarios). And really if you really have to stick with your design somehow, you of course have to use this solution, no other way.
Related
I'm investigating how to use OData in ASP.Core.
I've created a BooksController, subclassed from ODataController, within which I've defined two Actions: Get(), and Get(int id).
/odata/books resolves to the first Action, but /odata/books(1) does not find the second Action.
Once the Models are defined, it is able to find the following Controller:
[ODataRoutePrefix("Books")]
public class BooksController : ODataController
{
private BookStoreContext _db;
public BooksController(BookStoreContext context)
{
_db = context;
}
[ODataRoute]
[EnableQuery]
public IActionResult Get()
{
return Ok(_db.Books);
}
[EnableQuery]
[ODataRoute("({key})")]
public IActionResult Get([FromODataUri] int key)
{
return Ok(_db.Books.FirstOrDefault(c => c.Id == key.ToGuid()));
}
}
The site has the default Convention rules for all routes (see below).
But I think this is not in play, as the BooksController is decorated with [ODataRoutePrefix("Books")] and the actions with [ODataRoute] (and [EnableQuery]) -- which I think, being Attribute based routing, take precedence (is that a correct assumption?).
My dto models are are registered using Reflection...), but the key part is where Startup invokes UseMvc(...), and defines the routes, which ends up calling here:
private void CreateODataRoutes(IRouteBuilder routeBuilder)
{
// register the convention routes for MVC first...
routeBuilder.MapRoute(
name: "default",
template: "{controller=Home}/{action=Index}/{id?}");
...
// then do the OData stuff...
routeBuilder.Count().Expand().Filter()
.MaxTop(100).OrderBy().Select();
// Use method further down the page
// to create a Build Model by reflection, using
// all OData Model definitions (ie, classes that implement
// IAllModulesOdataModelBuilderConfiguration)
var oDataConventionModelBuilder = BuildODataModelUsingReflectionAcrossAllAssemblies();
// Use the modelBuilder as the basis of defining routes:
RegisterRoutesToODataController(routeBuilder, oDataConventionModelBuilder);
}
Where BuildODataModelUsingReflectionAcrossAllAssemblies uses reflection to find individual model definitions, each one pretty simple, only defining their id (relying on convention for the rest).
Note that I'm not defining Actions as that used to be by convention (see further down).
public class BookODataModelBuilderConfigurationBase<T> : IAllModulesOdataModelBuilderConfiguration
where T : class, IHasGuidId, new()
{
public virtual void Apply(ODataModelBuilder builder ...)
{
var _controllerName = this.GetControllerNameByConvention(typeof(Book));
var entity = builder.EntitySet<T>(this._controllerName).EntityType;
entity.HasKey(x => x.Id);
//Note...no Actions defined, as planning to rely on default conventions (routing by Verb to method starting with Get...)
}
}
When the model is created, it is registered as follows;
private void RegisterRoutesToODataController(IRouteBuilder routeBuilder,
ODataConventionModelBuilder oDataConventionModelBuilder)
{
string routePrefix = $"{App.Modules.Core.Shared.Constants.ModuleSpecific.Module.AssemblyNamePrefix}.";
// Build the Edm model used to parse commands:
var edmModel = oDataConventionModelBuilder.GetEdmModel();
// Register the Odata paths
routeBuilder.MapODataServiceRoute(
routeName: $"{routePrefix}odataDefault",
routePrefix: "odata",
edmModel,
pathHandler:new DefaultODataPathHandler(),
// By convention? So that Get verb goes to Get action, etc.
routingConventions: ODataRoutingConventions.CreateDefault()
);
}
When the path is /odata/book(1) it returns HTTP ERROR 404, the page does not exist.
Thank you!
Other things I've tried include:
commented out configuration of SwaggerAPI
Removed [FromODataUri] on the key param (is it necessary?)
Added/removed [ODataRoute("({key})")]
Registered the Controller as BooksController in plural/singular
Changed the name of the Action to GetBook and back again to Get
Added/removed ODataRoutePrefix
registering the OData routes before registering the default convention routes (think that should be the case all the time, right?).
...all of which is starting to look more like desperation than coding :-( ...
Still looking. Thanks for any guidance.
OMG. (sheepishly) solved.
It wasn't the Framework, Nuget, the Controller base class, routeprefix, routes, or anything glorious, it was ...me.
The only place I was not looking was the Model itself, which defined the Id as a Guid. The Action was using an int, converting it to a Guid.
ASP.Core could not find it because it was building routes based on the Model (not the Controller), so ignored the Action as it made no sense to its convention based route building as an int != Guid. duh.
If you were wondering why the heck I used an int...it was because when I had seeded the Db I wanted a Guid Key, but for testing purposes I wanted some records to have specific Ids that I could refer back to, and I was lazy, and didn't want to type in a full Guid.
That was a DUMB idea, in retrospect...:-(
But thank you for looking into it! Appreciate the time spent.
In a Sitecore project I've integrated Simple Injector using this article
It uses sitecore pipelines and then uses a method in App_start
namespace BBC.App_Start
{
public class SimpleInjector : IPackage
{
public void RegisterServices(Container container)
{
GetContainer.RegisterServices(container);
container.Register(() => new SitecoreContext(), Lifestyle.Scoped);
container.Register(() => new Container(), Lifestyle.Singleton);
}
}
}
Simply I can inject container into controller constructor but can't have container in View files.
I tried to declare a static property in App-start and save container to it. but still I'm getting no registration type in Views
What is the best way to have container object in views?
As Stephen suggests in his comment, the literal answer to your question is "you shouldn't do that - because it's not really the way MVC and DI are supposed to work". The more detailed answer goes something like this:
The job of your view is to present data that it has been passed via the Model. Views should not really contain logic. Very simple stuff like "if flag is false, hide this block of mark-up" is ok, but the more complex code to work out what the value of the flag is shouldn't be in the view.
MVC tries to make our website code better by encouraging you to separate presentation (the View) from data (the Model) and logic (the Controller). This should make our code easier to work with - So if you have processing that needs doing, then it should really be happening when your controller method runs.
If your view requires some special data, best practice suggests it should work it out in the controller method and pass it to the view in the model. The code might look more like this:
public class MyModel
{
public string SpecialData { get; set; }
}
public class MyController : Controller
{
public ActionResult DoSomething()
{
// do whatever processing is needed
var somethingCalculate = resultFromYourOtherObject();
// do other stuff
var model = new MyModel() { SpecialData = somethingCalculated };
return View(model);
}
}
And then the View just needs to accept the MyModel class as its model, and render the SpecialData property - no logic required.
I think also it's considered a bad idea to have calls to fetch objects from your DI container spread about your codebase. For MVC apps, generally your DI container gets wired in to the process of creating a controller for a request when the app starts up. Rather than passing about a DI Container into your controllers, the DI framework extends the Controller-creation process, and the container isn't exposed outside of this. When the MVC runtime needs to create a controller, the controller-creation logic uses the DI framework to fetch objects for all the controller's dependencies.
Without more detail about what you actually want to achieve, it's difficult to say what the "right" approach to creating your object(s) here is, but the two most common patterns are probably:
1) Constructor injection: Your controller has a parameter which accepts the object required. The DI container creates this object for you at the point where it creates the controller, so your controller gets all its dependencies when it is created. Good for: scenarios where you know how to create the object at the beginning of the request.
public interface IMySpecialObject
{
string DoSomething();
}
public class MyController : Controller
{
private IMySpecialObject _specialObject;
public MyController(IMySpecialObject specialObject)
{
_specialObject = specialObject;
}
public ActionResult RenderAView()
{
// do some stuff
var data = _specialObject.DoSomething();
return View(data);
}
}
As long as IMySpecialObject and a concrete implementation for it are registered with your DI container when your app starts up, all is well.
2) Factory classes: Sometimes, however, the object in question might be optional, or it might require data that's not available at controller-creation time to create it. In that case, your DI framework could pass in a Factory object to your controller, and this is used to do the construction of the special object later.
public interface ISpecialFactory
{
ISpecialObject CreateSpecialObject(object data);
}
public class MyController : Controller
{
private IMySpecialFactory _specialFactory;
public MyController(IMySpecialFactory specialFactory)
{
_specialFactory = specialFactory;
}
public ActionResult RenderAView()
{
// do some stuff
if( requireSpecialObject )
{
var data = getSomeData();
var specialObject = _specialFactory.CreateSpecialObject(data);
var data = _specialObject.DoSomething();
return View(data);
}
return View("someOtherView");
}
}
But a good book on using DI may suggest other approaches that fit your specific problem better.
I have a base controller which is globally marked as [Authorize]. Is there a way to prevent Controllers which inherit it from overriding the authorization requirement by simply adding the [AllowAnonymous] attribute?
Here is my exact scenario: I have three base controllers: one is for anonymous users, and two are for logged in users, both of which are globally decorated with [Authorize]. Each new controller that is created inherits from one of the base three, depending on which functionality is needed. One of the [Authorize] controllers contains "highly secure" functionality which absolutely should not be run by anonymous users. A developer inheriting from this "secure" base controller accidentally decorated some methods with [AllowAnonymous] which enabled anonymous users to potentially access the "secure" functionality in the base controller. It was caught in testing but I thought it would be a good idea to prevent that type of mistake, and I'm wondering if there is a simple way to do that. For now, I have taken all of the code inside of the secure base controller and wrapped it in blocks of:
if (Request.IsAuthenticated)
{
// do stuff
}
else
{
// redirect to login page, basically simulating what [Authorize] does
}
The above accomplishes what I want, however it kind of defeats the purpose of the global [Authorize] decoration in the first place. I'm envisioning something along the lines of:
[Authorize(AllowAnonymousOverride=false)] // this doesn't exist, but might be helpful
Is there a better way to accomplish this functionality?
The correct way to do this is to derive your own AuthorizeAttribute. The default AuthorizeAttribute looks like:
namespace System.Web.Mvc
{
public class AuthorizeAttribute : FilterAttribute, IAuthorizationFilter
{
public virtual void OnAuthorization(AuthorizationContext filterContext)
{
if (filterContext == null)
{
throw new ArgumentNullException("filterContext");
}
if (OutputCacheAttribute.IsChildActionCacheActive(filterContext))
{
throw new InvalidOperationException(MvcResources.AuthorizeAttribute_CannotUseWithinChildActionCache);
}
// This is the Important part..
bool flag = filterContext.ActionDescriptor
.IsDefined(typeof(AllowAnonymousAttribute), true)
|| filterContext.ActionDescriptor.ControllerDescriptor
.IsDefined(typeof(AllowAnonymousAttribute), true);
if (flag)
{
return;
}
if (this.AuthorizeCore(filterContext.HttpContext))
{
HttpCachePolicyBase cache = filterContext.HttpContext.Response.Cache;
cache.SetProxyMaxAge(new TimeSpan(0L));
cache.AddValidationCallback(
new HttpCacheValidateHandler(this.CacheValidateHandler), null);
return;
}
this.HandleUnauthorizedRequest(filterContext);
}
}
}
Derive your own:
public class CustomAuthorizeAttribute : AuthorizeAttribute
{
public bool IsAllowAnonymousEnabled { get; set; }
public virtual void OnAuthorization(AuthorizationContext filterContext)
{
if (filterContext == null)
{
throw new ArgumentNullException("filterContext");
}
if (OutputCacheAttribute.IsChildActionCacheActive(filterContext))
{
throw new InvalidOperationException(MvcResources.AuthorizeAttribute_CannotUseWithinChildActionCache);
}
// This is the Important part..
bool flag = IsAllowAnonymousEnabled
&& (filterContext.ActionDescriptor
.IsDefined(typeof(AllowAnonymousAttribute), true)
|| filterContext.ActionDescriptor.ControllerDescriptor
.IsDefined(typeof(AllowAnonymousAttribute), true));
if (flag)
{
return;
}
if (this.AuthorizeCore(filterContext.HttpContext))
{
HttpCachePolicyBase cache = filterContext.HttpContext.Response.Cache;
cache.SetProxyMaxAge(new TimeSpan(0L));
cache.AddValidationCallback(
new HttpCacheValidateHandler(this.CacheValidateHandler), null);
return;
}
this.HandleUnauthorizedRequest(filterContext);
}
}
Usage:
[CustomAuthorizeAttribute(IsAllowAnonymousEnabled = false)]
public class MyController : Controller
{
[AllowAnonymous]
public ActionResult Index()
{
// This will still execute Authorization regardless of [AllowAnonymous]
return View();
}
}
You can't call base.OnAuthorization() because it will Allow AllowAnonymous.
First, you seem to have a controller with some methods and then your approach is to inherit it so that the same methods are exposed. I wonder what's the point of having two or more controllers that expose the same data. Was is the mistake of that developer or rather it is a custom routine in your approach?
Then, you expect to have an attribute that prevents other attributes but this is clearly not possible in the language nor in the mvc framework.
Third, someone wrote a controller without unit tests or maybe with tests but no one verified these tests so that the issue was caught during manual testing phase. This indicates that the issue is wider and not only restricted to inheritance - suppose your developer wrote a controller that doesn't inherit anything and still exposes some critical data because of allow anonymous mark. Then what? Even if you have a remedy for your original issue, it wouldn't be able to catch the new possible issue.
My advice would be to have a custom analyzer attached to the post build event that scans all possible controllers and makes a list of all anonymous and restricted actions and compares it to a previously generated list. If there is a change, then an alert is created and someone has to resolve the issue manually, either by approving newly created actions or rejecting changes because a bug has been introduced.
Looking for some guidance in designing my new MVC 4 app.
I would like to have a url parameter s=2011 on every page of the app to let me know what year of data I'm working with. Obviously, the user will have a way to change that parameter as needed.
I will need that parameter in every controller and wondering the best way to do this. I was thinking of creating a base controller that reads Request.QueryString and puts the year into a public property. However, considering all the extensability points in MVC, I'm wondering if there's a better way to do this?
This very much depends on the design of your app, but just to give you two alternatives
IActionFilter
If you are doing data context per request you can use a global IActionFilter to hook pre-action execution globally and apply a query filter to your data context behind the scenes.
Major down-side of this is that to test the controller you will need to have the full MVC pipeline setup so that the actionfilter gets applied properly.
Dependency Injection
Instead of using sub-classing (base controller as you say) you can use dependency injection . Keeping things more loose will allow you to pull the filter from query string, cookie, user setting in the database or whatever else - without your controller knowing where it comes from.
Here is some pseudo code how I would do it if I was using something like Entity Framework or Nhibernate (also I am sure applicable with other technologies as well)
public Car
{
public string Year { get; set; }
}
public class CarsDataContext : DbContext
{
private IQuerable<Cars> _cars = null;
private Func<Car, bool> _carsFilter = null;
public IQuerable<Car> Cars {
get {
if (_carsFitler != null)
return _cars.Where(_carsFitler);
return _cars;
}
set { _cars = value; }
}
public void ApplyCarsFilter(Func<Car, bool> predicate)
{
_carsFilter = predicate;
}
}
Assuming you have dependency injection setup already (NInject or whichever other framework) in you can configure how the context to be intialized
Bind<CarsDataContext>().ToMethod(() => {
string yearFilter = GetYearFilter(); // can be coming from anywhere
CarsDataContext dataContext = new CarsDataContext();
dataContext.Applyfilter(car => car.Year == yearFilter);
return dataContext;
}).InRequestScope();
Then my controller knows nothing about the data filtering and I can easily test it:
class MyController : Controller
{
public MyController(CarsDataContext dataContext)
{
}
...
}
However I would only do this is filtering the dataset was across many controllers and important part of my software. Otherwise it's pure over-engineering.
I'm trying to use the new MVC4 DbDataController to expose a restful data api.
My problem is trying to secure this. I have created custom authorization attributes that derive from Authorize Attribute
public class AdminOnlyAttribute : AuthorizeAttribute
{
public override void OnAuthorization(AuthorizationContext filterContext)
{
if (!IsAllowed()) {
filterContext.Result = new HttpUnauthorizedResult("Not logged in");
}
...
}
And that works fine when applied to my normal controller actions. I'm trying to use the same thing in my data service like this:
[AdminOnlyAttribute]
public class DataServiceController : DbDataController<AppBuilderDataContext>
{
[AdminOnlyAttribute]
public IQueryable<Thing> GetThings()
{
return DbContext.AllMyThings();
}
}
You can see I've tried my attribute on both the controller and the action, but it's not firing for either one. I've set a breakpoint inside my authorize attribute function, and it's not getting called.
I'm pretty sure Scott Guthrie said this was going to work. Am I doing it wrong, or do I need a completely different method to secure these?
To work with an DataController or any other type derived from ApiController your attribute must derive from System.Web.Http.AuthorizeAttribute