Firstly i like what Guice does however i have also like to make as few classes in my system public. All services in my system are package private w/ only the interface public. This presents a few problems for Guice if one does not use a provider. I would prefer some alternative that allows me to provide a strategy to do the following:
for component C
requires dependencies X, Y, Z
call this method w/ X, Y, Z to create instance.
.. repeat again for all components.
In step 2 of the above psuedo code Guice now scans for #inject annotations. If this strategy was available I could decide how id like to find my component's dependencies in a generic way.
Unfortunately Module.bind just doesnt work because it works w/ classes literals.
The strategy interface for the above might look like this
interface ComponentDiscoverer {
Set<ComponentDefinition> components();
}
interface ComponentDefinition<T> {
List<RequiredDependency> requiredDependencies(); // similar to Guice's Key
<T> create( List<SatisfiedDependency> dependencies ); // factory method.
}
Guice gets all components by calling a registered ComponentDiscoverer.
It then calls the registered ComponentDiscoverer.components();
For each ComponentDefinition.requiredDependencies()
on demand call ComponentDefinition.create
Is this any built into Guice way or third party library to achieve the above ?
SAMPLE
package companyX;
public interface Service {
void execute();
}
package companyX;
package private class ServiceImpl implements Service {
static ServiceImpl create( Dependency dependency){
return new ServiceImpl( dependency );
}
private ServiceImpl( Dependency dependency ){
}
void execute();
}
package companyX;
package private class ServiceImpl2 implements Service2{
}
package companyX;
public class Factory {
Service alpha( Dependency ){
return ServiceImpl.create( dependency );
}
Service beta( DifferentDependency ){
// boring left out
}
Service2 gamma( AnotherDifferentDependency);
}
My module lives in a different package and all creation goes thru Fatory because everything else is package private.
You shouldn't have just one Module unless your application is very simple. You should have a module for each package that you have package-private classes you want to use in. This allows you to bind those package-private classes to the public interfaces. This is a standard practice for using Guice, and it's mentioned in the Keep Constructors Hidden section of the documentation.
With a module in the package, you just need normal #Injectable constructors or provider methods:
#Provides #Alpha Service alpha(Dependency dependency) {
return ServiceImpl.create(dependency);
}
Related
I'm trying to understand how to handle conditionally creating new instances of a class that uses #Inject. In the below example I have a factory that instantiates classes based on a parameter.
AnimalFactory does not have access to the injector the main class of my application has, so I can't use injector.getInstance(Cat.class)
class AnimalFactory {
public IAnimal create(AnimalType type) {
if (type.equals(AnimalType.CAT)) {
return new Cat(); // Cat uses #Inject, so this won't work of course. But ???
} else if (type.equals(AnimalType.DOG)) {
return new Dog();
}
}
}
In the rest of my app, classes are injected into my constructors because I always need them. Guice creates an instance/singleton for each. But in this scenario, I do not want to create and inject instances for each animal because all but one are needed.
You can use a MapBinder as described here:
public class AnimalModule extends AbstractModule {
public void configure() {
MapBinder<AnimalType, IAnimal> animalBinder= MapBinder.newMapBinder(binder(), AnimalType.class, IAnimal.class);
animalBinder.addBinding(AnimalType.DOG).to(Dog.class);
...
}
}
And than use it in your factory:
class AnimalFactory {
#Inject
Map<AnimalType, IAnimal> animals;
public IAnimal create(AnimalType type) {
return animals.get(type);
}
}
Actually I worked on exactly the same issue and I wrote a feature that allows you to create self-populating factory. Meaning that if you work in Spring/Spring-boot environment you can create a factory that can access and provide any interface implementing class that is managed by Spring-boot without injecting in the factory. You can also give the instances custom names. So, it seems like it fits your case exactly. Here is a link to an article that describes the feature in great detail: Non-intrusive access to "Orphaned" Beans in Spring framework. Also, in MgntUtils library Javadoc there is a good description of the feature here enter link description here. The library itself including source code could be found on Github here and in the package com.mgnt.lifecycle.management.example there is a working example. Maven artifacts are here
I have a Guice based project using vanilla Guice;
no Assisted-Inject, no AOP, no extra plugin extending Guice, etc.
To run it more easily on Android, Dagger seems like a better solution.
Every class has a dependency and a constructor with #Inject annotation.
No field or method injection is used.
The modules are quite simple (making Guice an overkill) and mostly contain bindings like the following:
class SomethingModule extends AbstractModule {
protected void configure() {
Bind(Handler.class)
.annotatedWith(Names.named("something"))
.to(SomeImplementation.class);
}
}
}
And later used like the following:
Injector inj = Guice.createInjector(new SomethingModule());
... = inj.getInstance(SampleInterface.class);
// and rest of the code.
Unfortunately,
I can not get my head around Daggers terminology.
Can you guide me with a direct translation / transformation of a Guice module to a Dagger module?
Dagger has:
Dagger's Components.
Dagger's Modules.
#Provides
#Inject
Guice has:
#Inject
#Named (or any custom annotation, if implemented correctly).
Our modules extending AbstractModule.
#Provides in the modules.
Guice Injector created from modules.
How do these relate?
Update: In addition to the nice answer by EpicPandaForce, these slides can help too.
Bind(Handler.class)
.annotatedWith(Names.named("something"))
.to(SomeImplementation.class);
Would translate to
#Module
public class SomethingModule {
#Provides
#Named("something")
//scope if needed
public Handler handler() {
return new SomeImplementation();
}
}
Which would be bound to an "Injector" (component):
#Component(modules={SomethingModule.class})
//scope if needed
public interface SomethingComponent {
#Named("something")
Handler handler();
void inject(ThatThingy thatThingy);
}
Which is an "injector" that you have to create with the APT-generated builder:
SomethingComponent somethingComponent = DaggerSomethingComponent.builder()
.somethingModule(new SomethingModule()) //can be omitted, has no params
.build();
somethingComponent.inject(thatThingy);
Where that thingy has
public class ThatThingy {
#Inject
#Named("something")
Handler handler;
}
Components typically exist per scope, so for example #ApplicationScope has one "injector" (component). Scoping can be achieved with subcomponents and component dependencies.
Important fact, a component has provision methods (which are the dependencies that are inherited to subscoped components if you use component dependencies), and void inject(X x); formatted methods. This is required for field injection per concrete type. A base class for example can only inject itself, and not its subclasses. You can however write a method called protected abstract void injectThis() which would call the .inject(this) on the subclass as well.
As I haven't really used Guice, I'm not sure if I missed out on anything. I think I forgot constructor injection, which is an issue because while Dagger does support it, it cannot be reconfigured. For reconfiguration, you have to use modules, and do the injection in the constructors yourself.
#Module(includes={ThoseModule.class, TheseModule.class})
public abstract class SomethingModule {
#Binds
abstract Whatever whatever(WhateverImpl impl);
}
#Singleton
public class WhateverImpl implements Whatever {
Those those;
These these;
#Inject
public Whatever(Those those, These these) {
this.those = those;
this.these = these;
}
}
#Component(modules={SomethingModule.class})
#Singleton
public interface SomethingComponent {
These these();
Those those();
Whatever whatever();
}
I'm working with a Guice enabled framework.
When using classes that were created by the framework (or subclasses that override existing bindings), I can instantiate framework provided variables very easily. Whatever I need, it's just a matter of
#Inject
FrameworkProvidedType variable;
However, in my custom created classes, that doesn't work. All of the injected variables are null.
It's my understanding that in order to use injection, my class has to have a binding.
If I'm subclassing an existing framework class, I can override the binding in my module class. That's pretty straightforward.
But I have a new class and I don't know how to bind it to the underlying framework.
public Class myCustomClass {
private String iNeedthis;
private Context thisToo;
#Inject
FrameWorkThing magic;
public myCustomClass(String iNeedThis, Context thisToo){
this.iNeedThis = iNeedThis;
this.thisToo = thisToo;
}
public void DoMagic(){
//null pointer error because magic was not injected
magic.doMagic(this.iNeedthis);
}
}
How do I Guice-enable this new class?
I tried this in my Runtime Module
public Class<myCustomClass> bindMyCustomClass(){
return MyCustomClass.class;
}
and failed miserably.
No thanks to #bmorris591 who dismissed and downvoted the question out of the gate, I found an answer.
#Inject-ing a field into a class means that the class instance needs to be created by Guice.
Step 1 is creating a factory for the class. This may not be necessary, but it worked for me.
public interface MyCustomClassFactory {
public MyCustomClass create(String iNeedThis, Context thisToo);
}
Step 2 is installing the factory into Guice
#Override
public void configure(Binder binder) {
super.configure(binder);
binder.install(new FactoryModuleBuilder().build(MyCustomClass.class));
}
In my particular case - the framework I'm working with provides a Module class that is an implementation of com.google.inject.Module.
Within that class is a "configure(Binder binder)" function that is called on startup.
Step 3 is actually annotating the constructor
#Inject
public myCustomClass(String iNeedThis, Context thisToo){
this.iNeedThis = iNeedThis;
this.thisToo = thisToo;
}
Useful and related web page that put me on the right track:
http://beust.com/weblog/2012/08/21/advanced-dependency-injection-with-guice/
This talks about assisted injection, but it gave enough information and a simple enough to understand example that taking the next step was pretty easy.
Lot's of times, classes need to be instantiated (constructed), and then "wired" (configured) before they can be used. For instance:
// Construction.
EventBus bus = new EventBus();
FizzEventHandler fizzHandler = new FizzHandler();
BuzzEventHandler buzzHandler = new BuzzHandler();
// Wiring.
bus.register(fizzHandler);
bus.register(buzzHandler);
In Guice, we accomplish the first part (construction; injection) with a Binder:
public class MyModule extends AbstractModule {
#Override
public void configure() {
bind(EventBus.class).to(SimpleEventBus.class);
bind(FizzEventHandler.class).to(DefaultFizzEventHandler.class);
bind(BuzzEventHandler.class).to(DefaultBuzzEventHandler.class);
}
}
But where does the wiring take place? When my Guice-based app starts up, we engage the DI "bootstrapping" process:
public class MyApp {
private EventBus bus;
private FizzEventHandler fizzHandler;
// ...etc.
public static void main(String[] args) {
MyApp app = new MyApp();
app.run();
}
public MyApp() {
// Bootstrap DI.
MyModule myModule = new MyModule();
Injector injector = Guice.createInjector(myModule);
bus = injector.inject(EventBus.class);
fizzHandler = injector.inject(FizzEventHandler.class);
// ...etc.
// Wire
bus.register(fizzHandler);
}
}
This works OK for the top-level (root) DI classes. But as we get further "down" the dependency tree, and get into all the other objects used by the application, putting the wiring logic in constructors like this is ugly and (I believe) is a discouraged practice.
So I ask: where doe battle-weary Guice veterans place their wiring/config code?
I work on a reasonably big system (~3000 classes) which uses Guice. I would say that our approach is to do everything with constructors. There aren't distinct "construction" and "wiring" activities as you describe, there's only construction.
In your example, the event handlers would be constructor parameters to the bus, which would register them in its constructor.
If you want to have fairly flexible injection of all the components of a given type (here, you would want to inject all event listeners into the bus), you could use multibindings. However, i don't think we actually use this in our codebase; we just write out manual lists of everything that needs injecting, which turns out not to be all that arduous in practice.
I generally use multiple modules, separated out by logical function. So one module might have authentication in it, another has data repositories, another the messaging system that I'm using, etc. This allows you to have different modules for mocking, caching Vs. non-caching, or just different implementations of the same service, and to switch out chunks of dependencies quickly and easily.
To make things even more flexible you could have a configuration file which declares the modules that should be used when the injector starts up.
When I have some logic to be done right after I instantiate my object I usually do it in methods annotated with #Provides. Your example might looks like this :
public class MyModule extends AbstractModule {
#Override
protected void configure() {
bind(FizzEventHandler.class).to(DefaultFizzEventHandler.class);
bind(BuzzEventHandler.class).to(DefaultBuzzEventHandler.class);
}
#Provides
public EventBus getEventBus(SimpleEventBuss simpleBus/* this here is going to be injected as it is a class not an interface and Guice is clever and it know how to do it ;) */
, FizzEventHandler fizz, BuzzEventHandler buzz) {
simpleBus.register(fizz);
simpleBus.register(buzz);
return simpleBus;
}
}
I have a Spring bean, let's say:
#TransactionAttribute(TransactionAttributeType.REQUIRED)
public class AImpl implements A {
public void setSomeDependency(D dependency) {
// This setter DOES NOT BELONG to interface A
}
}
<bean id="aImpl" class="AImpl"/>
Now I want to integration test it, but first I need to mock the dependency D, because it does too much stuff. Since the AImpl implements an interface and contains a transactional annotation, the generated proxy is only compatible with the interface A, so I can do this:
#Inject #Named("aImpl")
private A a;
but cannot:
#Inject #Named("aImpl")
private AImpl a;
As a result, I can't mock my dependency.
Please note that adding void setSomeDependency(D dependency) to interface A is not an option, as it has no business meaning. Neither it is using the proxy-target-class="true", as it breaks a whole lot of other beans (this attribute affects all beans in the context).
Is there a way to unproxy the injected bean A, so I could cast it to AImpl?
Try this:
if(AopUtils.isAopProxy(a) && a instanceof Advised) {
Object target = ((Advised)a).getTargetSource().getTarget();
AImpl ai = (AImpl)target;
}
Bonus: in Scala I am using the following equivalent function for the very same purpose:
def unwrapProxy(a: AnyRef) = a match {
case advised: Advised if(AopUtils.isAopProxy(advised)) =>
advised.getTargetSource.getTarget
case notProxy => notProxy
}
With the introduction of Spring 4.2.RC1, there is now a dedicated utility class in the spring-test module that handles this case for you.
The class is called AopTestUtils and provides the methods:
getTargetObject (unwraps only the top-level proxy)
getUltimateTargetObject (unwraps multiple levels of proxies if they exist).
Check out the relevant commit as well as the respective issue.