getApplication() vs. getApplicationContext()

Question

I couldn't find a satisfying answer to this, so here we go: what's the deal with Activity/Service.getApplication() and Context.getApplicationContext()?

In our application, both return the same object. In an ActivityTestCase however, mocking the application will make getApplication() come back with the mock, but getApplicationContext will still return a different context instance (one injected by Android). Is that a bug? Is it on purpose?

I don't even understand the difference in the first place. Are there cases outside a test suite where both calls may come back with different objects? When and why? Moreover, why is getApplication defined on Activity and Service, but not on Context? Shouldn't there always be a valid application instance available from anywhere?

Answer 1

Very interesting question. I think it's mainly a semantic meaning, and may also be due to historical reasons.

Although in current Android Activity and Service implementations, getApplication() and getApplicationContext() return the same object, there is no guarantee that this will always be the case (for example, in a specific vendor implementation).

So if you want the Application class you registered in the Manifest, you should never call getApplicationContext() and cast it to your application, because it may not be the application instance (which you obviously experienced with the test framework).

Why does getApplicationContext() exist in the first place ?

getApplication() is only available in the Activity class and the Service class, whereas getApplicationContext() is declared in the Context class.

That actually means one thing : when writing code in a broadcast receiver, which is not a context but is given a context in its onReceive method, you can only call getApplicationContext(). Which also means that you are not guaranteed to have access to your application in a BroadcastReceiver.

When looking at the Android code, you see that when attached, an activity receives a base context and an application, and those are different parameters. getApplicationContext() delegates it's call to baseContext.getApplicationContext().

One more thing : the documentation says that it most cases, you shouldn't need to subclass Application:

There is normally no need to subclass Application. In most situation, static singletons can provide the same functionality in a more modular way. If your singleton needs a global context (for example to register broadcast receivers), the function to retrieve it can be given a Context which internally uses Context.getApplicationContext() when first constructing the singleton.

I know this is not an exact and precise answer, but still, does that answer your question?

Answer 2

It seems to have to do with context wrapping. Most classes derived from Context are actually a ContextWrapper, which essentially delegates to another context, possibly with changes by the wrapper.

The context is a general abstraction that supports mocking and proxying. Since many contexts are bound to a limited-lifetime object such as an Activity, there needs to be a way to get a longer-lived context, for purposes such as registering for future notifications. That is achieved by Context.getApplicationContext(). A logical implementation is to return the global Application object, but nothing prevents a context implementation from returning a wrapper or proxy with a suitable lifetime instead.

Activities and services are more specifically associated with an Application object. The usefulness of this, I believe, is that you can create and register in the manifest a custom class derived from Application and be certain that Activity.getApplication() or Service.getApplication() will return that specific object of that specific type, which you can cast to your derived Application class and use for whatever custom purpose.

In other words, getApplication() is guaranteed to return an Application object, while getApplicationContext() is free to return a proxy instead.