I'm trying to decide whether to use Transfuse or Dagger for Android dependency injection. I've never used Transfuse, and have basic knowledge of Dagger. Thanks much.
To start, I am the primary author of Transfuse thus this answer may be a bit slanted in that direction.
Both Transfuse and Dagger handle Dependency Injection / Inversion of Control for Android in similar ways. Both use Annotation Processing at Compile time via JSR269 to generate code the supports the DI/IOC functionality. This allows them to avoid the costly runtime reflection-based analysis typically associated with DI containers found in non-Android Java. Without going into the specifics, Dagger and Transfuse do approach code generation in significantly different ways, which is reflected in the features of the libraries. Also, Transfuse and Dagger both use the common JSR330 annotations (#Inject, Provider, etc). This means they both follow a Guice-style injection scheme.
Here's how you create an object graph in Dagger:
public class DaggerActivity extends Activity {
#Inject Example example;
#Override protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
ObjectGraph.create().inject(this);
//do something else...
}
}
The equivalent code in Transfuse uses its #Factory functionality:
#Factory
public interface Injector {
Example get();
}
public class TransfuseActivity extends Activity {
Example example;
#Override protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
example = Factories.get(Injector.class).get();
//do something else...
}
}
Transfuse is meant to be used in the following way, however, utilizing POJO components, lifecycle events, etc:
#Activity
public class TransfuseActivity{
#Inject Example example;
#OnCreate public void doSomethingElse(){
//do something else...
}
}
Here's some little differences in the DI engine in Transfuse and Dagger:
Transfuse supports (as well as it can) cyclic dependencies, Dagger purposefully throws an exception in this case.
Transfuse satisfies JSR330, Dagger specifically does not. The Dagger developers wanted to err on the side of simplicity, avoiding method injection allowed them to avoid a handful of confusing cases (link).
Dagger has a reflection based engine for cases where code was not generated. Transfuse does not and requires code to be generated (annotation processor to run) in order to work.
Transfuse will inject into private fields, constructors, methods (not necessarily recommended because of the reflection overhead). Dagger throws an exception in this case.
Dagger uses Modules in a very direct way, mirroring the capabilities of Guice. Each time you create an object graph, you are given the option to configure it with a Module class, ie: ObjectGraph.create(new DripCoffeeModule()). Transfuse's configuration module is a bit different as it is incorporated into the application at compile time. Each Module in Transfuse is global to the project (this may change in future versions of Transfuse, but it has not been an issue for the use of Transfuse yet).
Singletons in Dagger are per-object-graph where Singletons in Transfuse are global to the application.
The big difference between Dagger and Transfuse is that Dagger focused on being a simple Dependency Injection library, while Transfuse's focus is to "make Android a better API using performance sensitive techniques"
Transfuse supports these capabilities as well as DI:
POJO Components
Manifest Management
Roboguice/Butterknife style injections
Lightweight event system (#Observes, #OnCreate, etc)
AOP
I'd recommend that if you're interested, give Transfuse a try. Personally, I'd love to hear about your experience contrasting it with Dagger. We have a mailing list where you can share with the community and pretty through documentation on the website.
Related
My app includes many helper classes like the next ones:
A class TMHttp with an "DownloadFile" or "IsOnline" methods (among others).
A TMEmail class with a "SendEmail" method.
A ImageHelper class with a "TakeScreenshot" or "CropBitmap" methods.
TMImport/TMExport classes with "ImportData" and "ExportData" methods.
A TMSession class with "LoginUser", "LogoutUser" and more such methods.
A TMUtils class with methods like "GenerateRandom".
A TMDialog class with methods like "CreateYesNoDialog" ... and many more.
And what I'd need to know -regarding DI- is if I have to inject them where needed -like I do with repository and services- or just create an instance of them. Do I have to inject every class in my app?
I believe there is no strict answer to this question because it depends on the architecture of your project, the principles the team is using, the purpose and the complexity of the helper/util class, etc.
You need to decide it for each such helper/util class particularly, guided by the following advantages of Dependency injection:
Reusability of classes and decoupling of dependencies: It's easier to swap out implementations of a dependency. Code reuse is improved because of inversion of control, and classes no longer control how their dependencies are created, but instead work with any configuration.
Ease of refactoring: The dependencies become a verifiable part of the API surface, so they can be checked at the object-creation time or at compile time rather than being hidden as implementation details.
Ease of testing: A class doesn't manage its dependencies, so when you're testing it, you can pass in different implementations to test all of your different cases.
For example, I would argue like this:
A TMUtils class with methods like "GenerateRandom"
Likely, it's a quite simple helper method that basically can have a standard single implementation; it doesn't depend on other libraries, so I can freely use it without DI.
A class TMHttp with an "DownloadFile" or "IsOnline" methods (among others).
It's heavy and it depends on a lib and Android SDK then it makes sense to inject.
And so on.
So I am currently in the process of learning dagger 2, and from the tutorials that I've read so far, for a dependency to be injected, the #Inject annotation gets placed inline with fields (for Activities/Fragments) or constructors. However I see that as an issue if I'm not the owner of parts of the code and can't add the required annotations for this technique to work, or if I don't want other parts of the code to know that dagger exists.
The application structure I have at the moment is:
App Module - where I'd like to put my DI code in (e.g. dagger modules, etc).
Presentation Module - Views/ViewModels etc.
Domain Module - Use Cases etc.
Data Module - Repositories etc.
With pretty much this style of classes contained in my application:
class ExampleViewModelImpl(useCase: ExampleUseCase): ViewModel() in Presentation (gets initialised from an Activity or similar).
class ExampleUseCaseImpl(repository: ExampleRepository): ExampleUseCase in Domain
class ExampleRepositoryImpl(dao: ExampleDao): ExampleRepository in Data
With the structure above, what is the minimum number of classes outside of the App Module that I need to touch in order to utilize dagger with as much automated dependency injection as possible? Code examples of how this is achieved would be great.
I am unsure of some terminologies, and wasn't able to find a solution online. If there are good resources which explains what I'm asking, that would also be great.
if I don't want other parts of the code to know that dagger exists.
#Inject is a standard (JSR 330) which Dagger implements. Adding those annotations doesn't have anything to do with Dagger and can be used the same way with other DI frameworks. If it's your code you should just add those #Inject annotations where appropriate. Think of them as documentation: What constructor/field/method must be injected to create & use this object?
The only place where your classes will know that Dagger exists is at the same place where you'd be creating the objects otherwise, too.
Going down that path, of course you could use Dagger without any #Inject annotations, but you'd be writing a lot of unnecessary boilerplate and missing out on the most powerful feature of Dagger at the same time (code generation).
#Inject annotation gets placed inline with fields (for Activities/Fragments) or constructors. However I see that as an issue if I'm not the owner of parts of the code and can't add the required annotations for this technique to work
That's what #BindsInstance with the #Component.Builder is for (add an object to the component) and what #Provides annotated methods are for (create and initialize an object from a module)
If you really want to write code without #Inject, then you'd do exactly this for all of your objects. This means a lot of modules, and even more #Provides annotated methods. It will work, but I don't see the point in writing all those methods if a single #Inject on the constructor has the same effect.
In my opinion the best thing about Dagger is that I can add / remove / change constructor parameters and don't have to touch any other parts of my code since Dagger will generate new code with the new arguments. In your case you'd have to also change the parameters to the #Provides method as well as the constructor invocation.
Next let's look at how to remove #Inject from fields. Basically you don't want to do field injection, so instead of writing an injection method in the component, you'd write provision methods.
#Component
class MyComponent {
fun inject(activity: MyActivity)
}
class MyActivity {
#Inject lateinit var myDep: Dependency
fun onCreate() {
component.inject(this)
}
}
Removing the #Inject we need to use the provision methods instead.
#Component
class MyComponent {
fun provisionMyDependency() : Dependency
}
class MyActivity {
lateinit var myDep: Dependency
fun onCreate() {
myDep = component.provisionMyDependency()
}
}
It will work and everything, but again, you will miss out on the single best feature of Dagger: Code generation. The example above looks alright because I only added a single dependency, but think about what happens to those 2 different implementations when you add / remove / change dependencies, how well it will scale. If you prefer to do things manually any refactoring will become arduous.
With the structure above, what is the minimum number of classes outside of the App Module that I need to touch in order to utilize dagger with as much automated dependency injection as possible?
Your question (especially the title) is in direct conflict with your goal. If you don't want to use those annotations, then you can't use Dagger code generation & injection but have to resort to do it manually as highlighted above.
with as much automated dependency injection as possible
To best utilize Dagger you add #Inject on the constructor and/or fields of every class that should end up on your dependency graph and let Dagger do its thing.
I'm working on an Android App which follows the MVP architecture pattern. In all my fragments I'm injecting a Presenter. Therefore my Fragments (Views) need to have a Component in which I declare the injects. For example:
#ActivityScope
#Component(
dependencies = AppComponent.class,
modules = {
PresenterModule.class,
InteractorModule.class
}
)
public interface ViewInjectorComponent {
void inject(SelectEventOccurrenceFragment fragment);
void inject(CreateOpponentFragment fragment);
void inject(SelectOpponentFragment fragment);
void inject(TeammatesInvitedFragment fragment);
...
}
Every new View that I add into my App (Fragment) needs to have its entry declared here. I was wondering if It's possible to generate this code automatically with some kind of annotation processor. The App has already several fragments, this component file has easily more than 300 entries. It'd be awesome if I could do something like:
#Injectable
public class MyNewFragment implements MyNewView {
...
}
And then automatically generate the entry in the ViewInjectorComponent file. It's possible? Where should I look at?
The situation you are experiencing may be a consequence of organising your Modules and Components in an unusual way. In particular, grouping laterally (one Component injects all the Presenters) rather than vertically (one component injects the functionality related to SelectOpponentActivity) is problematic.
A good example to follow is in the Google Android Architecture Blueprints GitHub repo. If you peruse the code there, you will see that they have organised functionality related to Tasks inside one Java package together with a separate Component, Module, Presenter etc. This has the nice advantage of being able to restrict accessibility of the constructors of the classes contained therein and fulfilling Effective Java Item 13: Minimize the accesibility of classes and members.
Likewise, you've grouped all your modules together into a Presenter Module and an Interactor Module. The advice from the Dagger 2 official documentation is to organise Modules first for testability and then along functional lines. Again, you can refer to the Blueprint example for how to do this.
Finally, note that there is unavoidably some boilerplate involved in using most DI frameworks like Dagger 2. In a sense, you are exchanging a bigger problem ("how do I deal with all of these constructors?") with much smaller and more manageable problems ("how do I group my Components" etc.).
Update
There is a library called Auto Dagger2 that can generate components for you. See this Github repo. Here is an example of an annotation:
#AutoComponent
#Singleton
public class ExampleApplication extends Application {
}
Which generates the following code:
#Component
#Singleton
public interface ExampleApplicationComponent {
}
Also check out Google Auto if you are interested in code generation tools.
I'm not entirely sure that what I'm going to say is appropriate for an answer, but I'll take a chance here.
Even if you find a way to do what you want, don't do this in production (it is fine if you just want to learn code generation techniques).
The benefits you get from such an approach are small (not writing several lines of trivial code), but consider the drawbacks:
The logic for code generation needs to be written/debugged/maintained
Such an approach will be a violation of "principle of least astonishment"
Code generation is BAD
Note the third point - I really mean it. Usage of code generation always leads to maintenance overhead, therefore it should be used only as a last resort.
Dagger by itself uses code generation, but for a good reason - performance. However, performance is not an issue in your case.
To summarize: your idea is very interesting, but approaches like this should not be used for production applications (unless this functionality is added to Dagger natively).
One of the best advantages of using DI is it makes testing a lot easier (What is dependency injection? backs it too). Most of DI frameworks I've worked with on other programming languages (MEF on .NET, Typhoon on Obj-C/Swift, Laravel's IoC Container on PHP, and some others) allows the developer do register dependencies on a single entry point for each component, thus preventing the "creation" of dependency on the object itself.
After I read Dagger 2 documentation, it sounds great the whole "no reflection" business, but I fail to see how it makes testing easier as objects are still kind of creating their own dependencies.
For instance, in the CoffeMaker example:
public class CoffeeApp {
public static void main(String[] args) {
// THIS LINE
CoffeeShop coffeeShop = DaggerCoffeeShop.create();
coffeeShop.maker().brew();
}
}
Even though you're not explicitly calling new, you still have to create your dependency.
Now for a more detailed example, let's go to an Android Example.
If you open up DemoActivity class, you will notice the onCreate implementation goes like this:
#Override
protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
// Perform injection so that when this call returns all dependencies will be available for use.
((DemoApplication) getApplication()).component().inject(this);
}
You can clearly see there is no decoupling from the DI component, to the actual code. In summary, you'd need to mock/stub ((DemoApplication) getApplication()).component().inject(this); on a test case (if that's even possible).
Up to this point, I am aware Dagger 2 is pretty popular, so there is got to be something I am not seeing. So how does Dagger 2 makes testing classes easier? How would I mock, let's say a network service class that my Activity depends on? I would like the answer to be as simple as possible as I'm only interested in testing.
Dagger 2 doesn't make testing easier
...beyond encouraging you to inject dependencies in the first place, which naturally makes individual classes more testable.
The last I heard, the Dagger 2 team were still considering potential approaches to improving support for testing - though whatever discussions are going on, they don't seem to be very public.
So how do I test now?
You're correct to point out that classes which want to make explicit use of a Component have a dependency on it. So... inject that dependency! You'll have to inject the Component 'by hand', but that shouldn't be too much trouble.
The official way
Currently, the officially-recommended approach to swapping dependencies for testing is to create a test Component which extends your production one, then have that use custom modules where necessary. Something like this:
public class CoffeeApp {
public static CoffeeShop sCoffeeShop;
public static void main(String[] args) {
if (sCoffeeShop == null) {
sCoffeeShop = DaggerCoffeeShop.create();
}
coffeeShop.maker().brew();
}
}
// Then, in your test code you inject your test Component.
CoffeeApp.sCoffeeShop = DaggerTestCoffeeShop.create();
This approach works well for the things you always want to replace when you are running tests - e.g. Networking code where you want to run against a mock server instead, or IdlingResource implementations of things for running Espresso tests.
The unofficial way
Unfortunately, it the official way can involve a lot of boilerplate code - fine as a one-off, but a real pain if you only want to swap out a single dependency for one particular set of tests.
My favourite hack for this is to simply extend whichever Module has the dependency you want to replace, then override the #Provides method. Like so:
CoffeeApp.sCoffeeShop = DaggerCoffeeShop.builder()
.networkModule(new NetworkModule() {
// Do not add any #Provides or #Scope annotations here or you'll get an error from Dagger at compile time.
#Override
public RequestFactory provideRequestFactory() {
return new MockRequestFactory();
}
})
.build();
Check this gist for a full example.
"allows the developer do register dependencies on a single entry point for
each component" - analogues in Dagger 2 are the Modules and Components where you define the dependencies. The advantage is that you don't define the dependencies directly in your component thus decoupling it so later when writing unit tests you may switch the Dagger 2 component with a test one.
"it sounds great the whole "no reflection" business" - the "no reflection" thing is not the "big deal" about dagger. The "big deal" is the full dependency graph validation at compile time. Others DI frameworks don't have this feature and if you fail to define how some dependency is satisfied you will get an error late at runtime. If the error is located in some rarely used codepath your program may look like it is correct but it will fail at some point in the future.
"Even though you're not explicitly calling new, you still have to create your dependency." - well, you always have to somehow initiate dependency injection. Other DI may "hide"/automate this activity but at the end somewhere building of the graph is performed. For dagger 1&2 this is done at app start. For "normal" apps (as you shown in the example) in the main(), For android apps - in the Application class.
"You can clearly see there is no decoupling from the DI component, to the actual code" - Yes, you are 100% correct. That arises from the fact that you don't control directly the lifecycle of the activities, fragments and services in Android, i.e. the OS creates these objects for you and the OS is not aware that you are using DI. You need manually to inject your activities, fragments and services. At first this seem seems awkward but in real life the only problem is that sometimes you may forget to inject your activity in onCreate() and get NPE at runtime.
#SuppressWarnings("unsued")
#Override
#SuppressLint({ "InflateParams", "SimpleDateFormat" })
I don't get why we need to declare annotations.
We want to facilitate the writing and the maintenance of Android applications.
We believe that simple code with clear intents is the best way to achieve those goals.
Robert C. Martin wrote:
The ratio of time spent reading [code] versus writing is well over 10 to 1 [therefore] making it easy to read makes it easier to write.
While we all enjoy developing Android applications, we often wonder: Why do we always need to write the same code over and over? Why are our apps harder and harder to maintain? Context and Activity god objects, complexity of juggling with threads, hard to discover API, loads of anonymous listener classes, tons of unneeded casts... can't we improve that?
How?
Using Java annotations, developers can show their intent and let AndroidAnnotations generate the plumbing code at compile time.
Features
Dependency injection: inject views, extras, system services, resources, ...
Simplified threading model: annotate your methods so that they execute on the UI thread or on a background thread.
Event binding: annotate methods to handle events on views, no more ugly anonymous listener classes!
REST client: create a client interface, AndroidAnnotations generates the implementation.
No magic: As AndroidAnnotations generate subclasses at compile time, you can check the code to see how it works.
AndroidAnnotations provide those good things and even more for less than 50kb, without any runtime perf impact!
Is your Android code easy to write, read, and maintain?
Look at that:
#EActivity(R.layout.translate) // Sets content view to R.layout.translate
public class TranslateActivity extends Activity {
#ViewById // Injects R.id.textInput
EditText textInput;
#ViewById(R.id.myTextView) // Injects R.id.myTextView
TextView result;
#AnimationRes // Injects android.R.anim.fade_in
Animation fadeIn;
#Click // When R.id.doTranslate button is clicked
void doTranslate() {
translateInBackground(textInput.getText().toString());
}
#Background // Executed in a background thread
void translateInBackground(String textToTranslate) {
String translatedText = callGoogleTranslate(textToTranslate);
showResult(translatedText);
}
#UiThread // Executed in the ui thread
void showResult(String translatedText) {
result.setText(translatedText);
result.startAnimation(fadeIn);
}
// [...]
}
Java annotations bind specific conditions to be satisfied with code. Consider a scenario where we think we are overriding a method from anther class and we implemented code that (we think) is overriding the method. But if we somehow missed to exactly override one (e.g. we misspelled name. In superclass it was "mMethodOverridden" and we typed "mMethodoverridden"). The method will still compile and execute but it will not be doing what it should do.
So #Override is our way of telling Java to let us know if we are doing right thing. If we annotate a method with #override and it is not overriding anything, compiler will give us an error.
Other annotations work in a very similar way.
For more information, read docs Lesson: annotations
Annotations are basically syntactic metadata that can be added to Java source code.Classes, methods, variables, parameters and packages may be annotated .
Metadata is data about data
Why Were Annotations Introduced?
Prior to annotation (and even after) XML were extensively used for metadata and somehow a particular set of Application Developers and Architects thought XML maintenance was getting troublesome. They wanted something which could be coupled closely with code instead of XML which is very loosely coupled (in some cases almost separate) from code. If you google “XML vs. annotations”, you will find a lot of interesting debates. Interesting point is XML configurations were introduced to separate configuration from code. Last two statements might create a doubt in your mind that these two are creating a cycle, but both have their pros and cons.
For eg:
#Override
It instructs the compiler to check parent classes for matching methods.