I have tried to understand AOP, Dependency Injection and Inversion of Control SPRING related concepts but I am having hard time understanding it.
Can anyone explain this in simple English ?
I understand your confusion and it took me some time to understand how these concepts were related together. So here is my (somehow personal) explanation of all this:
1. Inversion of Control
Inversion of control is a design principle rather generic that refers to the decoupling of the specification of a behavior from when it is actually executed. Compare for instance,
myDependency.doThis();
with
myDependency.onEventX += doThis();
In the latter, there is no direct invocation which is more flexible. In its general form, inversion of control relates to the observer pattern, events, or callbacks.
2. Dependency inversion
Dependency inversion is another design principle. Roughly speaking, it says that higher-level abstraction should not depend directly on lower-level abstractions; this results indeed in a design where higher-level abstraction can not be reused without the lower-level abstractions.
class MyHighLevelClass {
MyLowLevelClass dep = new MyLowLeverClass();
}
class App {
void main() { new HighLevelClass().doStuff(); }
}
Here, MyHighLevelClass can not compile without access to MyLowLevelClass. To break this coupling, we need to abstract the low level class with an interface, and remove the direct instantiation.
class MyLowLevelClass implements MyUsefulAbstraction { ... }
class MyHighLevelClass {
MyUsefulAbstraction dep;
MyHighLevelClass( MyUsefulAbstraction dep ) {
this.dep = dep;
}
}
class App {
void main() { new HighLevelClass( new LowLevelClass() ).doStuff(); }
}
Note that you don't need anything special like a container to enforce dependency inversion, which is a principle. A good reading is The Dependency Inversion Principle by Uncle Bob.
3. Dependency injection
Now comes dependency injection. To me dependency injection = IoC + dependency inversion:
dependencies are provided externally so we enforce the dependency inversion principle
the container sets the dependencies (not us) so we speak of inversion of control
In the example I provided above, dependency injection can be done if a container is used to instantiate objects and automatically inject the dependency in the constructor (we speak then frequently of DI container):
class App {
void main() { DI.getHighLevelObject().doStuff(); }
}
Note that there are various form of injections. Note also that under this perspective, setter injection can be seen as a form of callback -- the DI container creates the object then calls back the setter. The flow of control is effectively inverted.
4. AOP
Strictly speaking, AOP has little to do with the 3 previous points. The seminal paper on AOP is very generic and present the idea of weaving various sources together (possibly expressed with different languages) to produce a working software.
I won't expand more on AOP. What is important here, is that dependency injection and AOP do effectively plays nicely together because it makes the weaving very easy. If an IoC container and dependency injection is used to abstract away the instantiation of objects, the IoC container can easily be used to weave the aspects before injecting the dependencies. This would otherwise requires a special compilation or a special ClassLoader.
Hope this helps.
Dependency injection was explained very well in How to explain dependency injection to a 5-year-old?:
When you go and get things out of the
refrigerator for yourself, you can
cause problems. You might leave the
door open, you might get something
Mommy or Daddy doesn't want you to
have. You might even be looking for
something we don't even have or which
has expired.
What you should be doing is stating a
need, "I need something to drink with
lunch," and then we will make sure you
have something when you sit down to
eat.
AOP - Aspect Oriented Programming - basically means that the source you write is modified with other code, based on rules located ELSEWHERE. This means that you can e.g. say "as the first line of every method I want a 'log.debug("entering method()")' in a central place and each and every method you compile with that rule in place will then have that line included. The "aspect" is the name of looking on code in other ways than simply from first source line to last.
Inversion of Control basically means that you do not have a central piece of code controlling everything (like a giant switch in main()) but have a lot of pieces of code that "somehow" get called. The subject is discussed at Wikipedia: http://en.wikipedia.org/wiki/Inversion_of_control
These three are all different concepts, but they all work well together, and so Spring apps often make use of all of it at once. I'll give you an example.
Let's say that we have a web application that can do many different things. We could construct this application in many ways, but one way is to create a class that is in charge of doing each of these things. We need to invoke and create these classes from somewhere. One option is to have a big main class that creates one of each of these services, opens up a socket, and passes calls to these services as they come in. Unfortunately, we've gone and created ourselves a god class, which has way too much logic and knows way too much about how everything in our program works. If we change anything about our program, we're probably going to need to modify this class.
Also, it's difficult to test. We can't test any class in isolation if it runs around instantiating and invoking the other classes directly. Unit tests become much, much harder to write.
A way to get around this is to use inversion of control. We say "okay, these are service classes. Who instatiates them? Not me." Usually, each one defines an interface, like LoginService or BillingService. There might be more than one implementation of that interface, but your app doesn't care. It just knows that it can ask for a certain kind of a service or a service with a certain name, and it'll get something nice back.
Dependency injection allows us to wire all of our litle pieces together. Classes have accessible fields, constructor arguments, or setter methods that are references to the other components that they'll need to access. That makes unit testing much easier. You can create the object under test, throw a mock or stub dependency at it, and then test that the object behaved correctly in isolation.
Now, our real application is a complex jumble of pieces that all need to be wired together just so. There are many ways to accomplish this, including allowing the application to make guesses ("this class wants a UserService, there is exactly one other class I'm in charge of that implements UserService") or by carefully explaining how they wire together in XML or Java. Spring, at its core, is a service that takes care of wiring these classes together.
Now we get to AOP. Let us say that we have all of these classes that are wired to each other in elaborate ways. There are some cross-cutting concerns that we might want to describe in very generic ways. For instance, perhaps you'd like to start a database transaction whenever any service is invoked, and commit that transaction so long as the service doesn't throw an exception. It turns out that Spring is in a unique position to perform such a task. Spring can create proxy classes on the fly that implement whatever interface your classes need, and it can wrap your class in its proxy. Now, IoC and dependency injection certainly aren't necessary to do aspect-oriented programming, but it's an extremely convenient way to accomplish it.
The difference between Dependency Injection and Inversion of Control is explained very well in
http://martinfowler.com/articles/dipInTheWild.html
("You mean Dependency Inversion, Right?" section)
The summary:
DI is about how one object acquires a dependency. When a dependency
is provided externally, then the system is using DI.
IoC is about who initiates the call. If your code initiates a call,
it is not IoC, if the container/system/library calls back into code
that you provided it, is it IoC.
let me tell you some word about AOP, hope it make it simplier to understand.
The very base principle of AOP is finding common tasks/aspects which returns in
many places in the code and dont belong to the concerete business of the code. For example
write to log on every entrance to any function, or when an object is created wrapp it, or send email to admin when calling to specific function.
So instead of the programmers will handle these non businuss aspect we take it from them and
we manage these aspects behond the scene.
That all the basic of AOP on 1 leg....
A simple comparison from Spring in Action:
Whereas DI helps you decouple your application
objects from each other, AOP helps you decouple cross-cutting concerns from the
objects that they affect.
Related
I just started getting into DI(dependency injection) frameworks and I was wondering should I inject every and any class or specific classes? How would I know which classes you should inject and which you shouldn't?
A lot of the tutorials(Dagger 2, Hilt, etc) that I've found don't seem to talk about this or don't explain it very well.
I don't have any code examples cause I'm trying out things as I read along with tutorials. Just trying to get a feeling of things.
Explain it as if I was 4 year old :)
My answer is no, you do not have to do that use dependency injection every time you need to create an object.
This is the intent of dependency injection according to Wikipedia
The intent behind dependency injection is to achieve separation of concerns of construction and use of objects. This can increase readability and code reuse.
So you should use dependency injection it would improve your code reuse and readability.
This article mentions some cases where you should use dependency injection to create objects
You need to inject configuration data into one or more components.
You need to inject the same dependency into multiple components.
You need to inject different implementations of the same dependency.
You need to inject the same implementation in different configurations.
You need some of the services provided by the container.
Imagine that you have username and password fields in an app that you would use to capture a user's detail, you do not have to use dependency injection to create a User object from the details, you can just create it directly, its not something that would be reused across your application, there are other object creation patterns that you should look into like factory pattern, abstract factory pattern, builder pattern e.t.c.
There's a certain number of classes which I always inject and never ask myself the question: the application context, shared preferences, DAOs, Retrofit, Json object. As you can see, these are usually related the third-party libraries or the Android framework.
Then there are classes which depend on those: the repository is a common example, but there are many more. For example you could have a preference manager which depends on shared preferences and context. Or you could have a class JsonExporter used to export user data which depends on DAOs and the Json mapper.
In turn, there are other classes which depend on those new classes: a view model/presenter in MVVM/MVP architecture could depend on both the repository and JsonExporter for example. At this point you have two choices:
Instantiate the class yourself. But that means you need to have access to all of the class' dependencies, which are only available through dependency injection...
so you might as well inject it. The cost of injection at this point is often very low, just adding #Inject to the constructor. Only the base dependencies have to be provided in a module.
From a certain number of injected base classes, the decision to inject those more nested in the dependency graph really comes automatically.
There are also cases where you'll have a class that depends on nothing. Perhaps you decided to extract code from a big class into another one. Or perhaps the class just has no dependencies. I probably wouldn't inject such a class, there's no point.
Singletons are often handy to inject because the dependency injection framework (e.g. Dagger) can make sure there's always only one instance automatically.
One reason why we do dependency injection is so that classes depend on abstractions, not concretions (aka inversion of control). It's useful to have your class depend on a Repository interface, because you can decide to provide it with the implementation you want, for example RealRepository in the app, and MockRepository in tests. Another example is flavors or build variants: an injected FlavorBehavior interface could have different implementations in two different flavors. It's not the class' responsability to decide which to use.
Note that this is not a definitive answer, and I'm not an expert on the subject. It's an opinion-based subject too.
I'm fairly new to Java and joining a project that leverages the DDD pattern (supposedly). I come from a strong python background and am fairly anal about unit test driven design. That said, one of the challenges of moving to Java is the testability of Service layers.
Our REST-like project stack is laid out as follows:
ServiceHandlers which handles request/response, etc and calls specific implementations of IService (eg. DocumentService)
DocumentService - handles auditing, permission checking, etc with methods such as makeOwner(session, user, doc)
Currently, something like DocumentService has repository dependencies injected via guice. In a public method like DocumentService.makeOwner, we want to ensure the session user is an admin as well as check if the target user is already an owner (leveraging the injected repositories). This results in some dupe code - one for both users involved to resolve the user and ensure membership, permissions, etc etc. To eliminate this redundant code, I want make a sort of super simpleisOwner(user, doc) call that I can concisely mock out for various test scenarios (such as throwing the exception when the user can't be resolved, etc). Here is where my googling fails me.
If I put this in the same class as DocumentService, I can't mock it while testing makeOwner in the same class (due to Mockito limitations) even though it somewhat feels like it should go here (option1).
If I put it in a lower class like DocumentHelpers, it feels slightly funny but I can easily mock it out. Also, DocumentHelpers needs the injected repository as well, which is fine with guice. (option 2)
I should add that there are numerous spots of this nature in our infant code base that are untestable currently because methods are non-statically calling helper-like methods in the same *Service class not used by the upper ServiceHandler class. However, at this stage, I can't tell if this is poor design or just fine.
So I ask more experienced Java developers:
Does introducing "Service Helpers" seem like a valid solution?
Is this counter to DDD principals?
If not, is there are more DDD-friendly naming convention for this aside from "Helpers"?
3 bits to add:
My googling has mostly come up with debates over "helpers" as static utility methods for stateless operations like date formatting, which doesn't fit my issue.
I don't want to use PowerMock since it breaks code coverage and is pretty ugly to use.
In python I'd probably call the "Service Helper" layer described above as internal_api, but that seems to have a different meaning in Java, especially since I need the classes to be public to unit test them.
Any guidance is appreciated.
That the user who initiates the action must be an admin looks like an application-level access control concern. DDD doesn't have much of an opinion about how you should do that. For testability and separation of concerns purposes, it might be a better idea to have some kind of separate non-static class than a method in the same service or a static helper though.
Checking that the future owner is already an owner (if I understand correctly) might be a different animal. It could be an invariant in your domain. If so, the preferred way is to rely on an Aggregate to enforce that rule. However, it's not clear from your description whether Document is an aggregate and if it or another aggregate contains the data needed to tell if a user is owner.
Alternatively, you could verify the rule at the Application layer level but it means that your domain model could go inconsistent if the state change is triggered by something else than that Application layer.
As I learn more about DDD, my question doesn't seem to be all that DDD related and more just about general hierarchy of the code structure and interactions of the layers. We ended up going with a separate DocumentServiceHelpers class that could be mocked out. This contains methods like isOwner that we can mock to return true or false as needed to test our DocumentService handling more easily. Thanks to everyone for playing along.
I am trying to have an annotation #FeatureDependent be used on methods to signal that the method requires certain things to be enabled in order for it to work. And I was wondering if it was possible to have a method called everytime a method with #FeatureDependent was called which would check if the criteria were met for the method to be called.
It sounds like you are describing Aspect Oriented Programming (AOI). This technique allows you to address "cross-cutting" concerns, tasks like logging, security, and transaction management which tend to affect many methods in the same manner. Your use case sounds like it would be a good fit for AOP.
There are two common approaches to AOP. The first mechanism is to create objects in a container (e.g. a Spring container). The container can then scan the class, detect any advice that needs to be applied, and apply the advice via dynamic proxies (Googling Spring and AOP is a good place to start with this). The downside is that your components will need to be constructed by a container so it makes sense for larger components.
The second approach is an extra compilation step (sometimes done at compilation, sometimes done as a separate compilation phase, and sometimes done by a weaving class loader) to wire in the additional methods. This is typically called "weaving" and AspectJ is a common library to look into for this.
Both approaches will allow you to apply "advice" (code run before and after a method invocation) based on annotations on an object. Explaining either in more detail would be beyond the scope of a SO answer but I hope it can get you started.
I should warn that AOP has gotten a bit of a reputation for complicating the flow of an application and tending to be difficult to understand and debug. As a result it has declined in popularity lately.
Another approach is to use something like Servlet Filters, basically a single choke point for all requests and workflows where you can apply various logging & security mechanisms. Such an approach tends to be a little easier to understand and involve a bit less "black magic".
I have been trying to follow a domain driven design approach in my new project. I have always generally used Spring for dependency injection, which nicely separates my application code from the construction code, however, with DDD I always seem to have one domain object wanting to create another domain object, both of which have state and behaviour.
For example, given a media file, we want to encode it to a different format - the media asset calls on a transcode service and receives a callback:
class MediaAsset implements TranscodingResultListener {
private NetworkLocation permanentStorage;
private Transcoder transcoder;
public void transcodeTo(Format format){
transcoder.transcode(this,format);
}
public void onSuccessfulTranscode(TranscodeResult result){
Rendition rendition = new Rendition(this, result.getPath(), result.getFormat());
rendition.moveTo(permanentStorage);
}
}
Which throws two problems:
If the rendition needs some dependencies (like the MediaAsset requires a "Transcoder") and I want to use something like Spring to inject them, then I have to use AOP in order for my program to run, which I don't like.
If I want a unit test for MediaAsset that tests that a new format is moved to temporary storage, then how do I do that? I cannot mock the rendition class to verify that it had its method called... the real Rendition class will be created.
Having a factory to create this class is something that I've considered, but it is a lot of code overhead just to contain the "new" keyword which causes the problems.
Is there an approach here that I am missing, or am I just doing it all wrong?
I think that the injection of a RenditionFactory is the right approach in this case. I know it requires extra work, but you also remove a SRP violation from your class. It is often tempting to construct objects inside business logic, but my experience is that injection of the object or a objectfactory pays off 99 out of 100 times. Especially if the mentioned object is complex, and/or if it interacts with system resources.
I assume your approach for unit testing is to test the MediaAsset in isolation. Doing this, I think a factory is the common solution.
Another approach is to test the whole system (or almost the whole system). Let your test access the outer interface[1] (user interface, web service interface, etc) and create test doubles for all external systems that the system accesses (database, file system, external services, etc). Then let the test inject these external dependencies.
Doing this, you can let the tests be all about behaviour. The tests become decoupled from implementation details. For instance, you can use dependency injection for Rendition, or not: the tests don't care. Also, you might discover that MediaAsset and Rendition are not the correct concepts[2], and you might need to split MediaAsset in two and merge half of it with Rendition. Again, you can do it without worrying about the tests.
(Disclaimer: Testing on the outer level does not always work. Sometimes you need to test common concepts, which requires you to write micro tests. And then you might run into this problem again.)
[1] The best level might actually be a "domain interface", a level below the user interface where you can use the domain language instead of strings and integers, and where you can talk domain actions instead of button clicks and focus events.
[2] Perhaps this is actually your problem: Are MediaAsset and Rendition the correct concepts? If you ask your domain expert, does he know what these are? If not, are you really doing DDD?
As you many know when you proxy an object, like when you create a bean with transactional attributes for Spring/EJB or even when you create a partial mock with some frameworks, the proxies object doesn't know that, and internal calls are not redirected, and then not intercepted either...
That's why if you do something like that in Spring:
#Transactionnal
public void doSomething() {
doSomethingInNewTransaction();
doSomethingInNewTransaction();
doSomethingInNewTransaction();
}
#Transactional(propagation = Propagation.REQUIRES_NEW)
public void doSomethingInNewTransaction() {
...
}
When you call doSomething, you expect to have 3 new transactions in addition to the main one, but actually, due to this problem you only get one...
So i wonder how do you do to handle these kind of problems...
I'm actually in a situation where i must handle a complex transactional system, and i don't see any better way than splitting my service into many small services, so that I'm sure to pass through all the proxies...
That bothers me a lot because all the code belongs to the same functional domain and should not be split...
I've found this related question with interesting answers:
Spring - #Transactional - What happens in background?
Rob H says that we can inject the spring proxy inside the service, and call proxy.doSomethingInNewTransaction(); instead.
It's quite easy to do and it works, but i don't really like it...
Yunfeng Hou says this:
So I write my own version of CglibSubclassingInstantiationStrategy and
proxy creator so that it will use CGLIB to generate a real subclass
which delegates call to its super rather than another instance, which
Spring is doing now. So I can freely annotate on any methods(as long
as it is not private), and from wherever I call these methods, they
will be taken care of. Well, I still have price to pay: 1. I must list
all annotations that I want to enable the new CGLIB sub class
creation. 2. I can not annotate on a final method since I am now
generating subclass, so a final method can not be intercepted.
What does he mean by "which spring is doing now"? Does this mean internal transactional calls are now intercepted?
What do you think is better?
Do you split your classes when you need some transactional granularity?
Or do you use some workaround like above? (please share it)
I'll talk about Spring and #Transactional but the advise applies for many other frameworks also.
This is an inherent problem with proxy based aspects. It is discussed in the spring documentation here:
http://static.springsource.org/spring/docs/3.0.x/spring-framework-reference/html/aop.html#aop-understanding-aop-proxies
There are a number of possible solutions.
Refactor your classes to avoid the self-invocation calls that bypass the proxy.
The Spring documentation describes this as "The best approach (the term best is used loosely here)".
Advantages of this approach are its simplicity and that there are no ties to any framework. However, it may not be appropriate for a very transactional heavy code base as you'd end up with many trivially small classes.
Internally in the class get a reference to the proxy.
This can be done by injecting the proxy or with hard coded " AopContext.currentProxy()" call (see Spring docs above.).
This method allows you to avoid splitting the classes but in many ways negates the advantages of using the transactional annotation. My personal opinion is that this is one of those things that is a little ugly but the ugliness is self contained and might be the pragmatic approach if lots of transactions are used.
Switch to using AspectJ
As AspectJ does not use proxies then self-invocation is not a problem
This is a very clean method though - it is at the expense of introducing another framework. I've worked on a large project where AspectJ was introduced for this very reason.
Don't use #Transactional at all
Refactor your code to use manual transaction demarcation - possibly using the decorator pattern.
An option - but one that requires moderate refactoring, introducing additional framework ties and increased complexity - so probably not a preferred option
My Advice
Usually splitting up the code is the best answer and can also be good thing for seperation of concerns also. However, if I had a framework/application that heavily relied on nested transactions I would consider using AspectJ to allow self-invocation.
As always when modelling and designing complex use cases - focus on understandable and maintainable design and code. If you prefer a certain pattern or design but it clashes with the underlying framework, consider if it's worth a complex workaround to shoehorn your design into the framework, or if you should compromise and conform your design to the framework where necessary. Don't fight the framework unless you absolutely have to.
My advice - if you can accomplish your goal with such an easy compromise as to split out into a few extra service classes - do it. It sounds a lot cheaper in terms of time, testing and agony than the alternative. And it sure sounds a lot easier to maintain and less of a headache for the next guy to take over.
I usually make it simple, so I split the code into two objects.
The alternative is to demarcate the new transaction yourself, if you need to keep everything in the same file, using a TransactionTemplate. A few more lines of code, but not more than defining a new bean. And it sometimes makes the point more obvious.