Starting a new GWT application and wondering if I can get some advice from someones experience.
I have a need for a lot of server-side functionality through RPC services...but I am wondering where to draw the line.
I can make a service for every little call or I can make fewer services which handle more operations.
Let's say I have Customer, Vendor and Administration services. I could make 3 services or a service for each function in each category.
I noticed that much of the service implementation does not provide compile-time help and at times troublesome to get going, but it provides good modularity. When I have a larger service, I don't have the modularity as I described, but I don't have to the service creation issues and reduce the entries in my web.xml file.
Is there a resource issue with using a lot of services? What is the best practice to determine what level of granularity to use?
in my opinion, you should make a rpc service for "logical" things.
in your example:
one for customer, another for vendors and a third one for admin
in that way, you keep several services grouped by meaning, and you will have a few lines to maintain in the web.xml file ( and this is a good news :-)
More seriously, rpc services are usually wrappers to call database stuff, so, you even could make a single 'MagicBlackBoxRpc' with a single web.xml entry and thousands of operations !
but making a separate rpc for admin operations, like you suggest, seems a good thing.
Read general advice on "how big should a class be?", which is available in any decent software engineering book.
In my opinion:
One class = One Subject (ie. group of functions or behaviours that are related)
A class should not deal with more than one subject. For example:
Class PersonDao -> Subject: interface between the DB and Java code.
It WILL NOT:
- cache Person instances
- update fields automatically (for example, update the field 'lastModified')
- find the database
Why?
Because for all these other things, there will be other classes doing it! Respectively:
- a cache around the PersonDao is concerned with the efficient storage of information to avoid hitting the DB more often than necessary
- the Service class which uses the DAO is responsible for modifying anything that needs to be modified automagically.
- To find the database is responsibility of the DataSource (usually part of a framework like Spring) and your Dao should NOT be worried about that. It's not part of its subject.
TDD is the answer
The need for this kind of separation becomes really clear when you do TDD (Test-Driven Development). Try to do TDD on bad code where a single class does all sorts of things! You can't even get started with one unit test! So this is my final hint: use TDD and that will tell you how big a class should be.
I think the thing to optimize for is that you can accomplish a result in one round trip to the server. I have an ad-hoc collection of methods on my service object, one for each situation the client finds itself in when it has to get something done. You do not want the client to RPC to the server several times in a row while the user is sitting there waiting.
REST makes things orthogonal, but orthogonality has a cost: there is a reason that the frequently used verbs in languages are irregular. In terms of maintaing clean orthogonal structure to your app, make sure your schema is well-designed. That is where each class should have semantics orthogonal to that of the other classes. When the semantics of each RPC call can be stated cleanly in the schema there will be no confusion as to what they mean, even if they aren't REST-fully ideal.
Related
I have a webapp with multiple Spring services (they each have their own ear and web-controllers) and they talk to each other via REST calls. Some of the different services use the same data POJOs. The existing code just has duplicates of those data objects in different services.
For example, my /users service has a myApp.users.UserData object, and my /emails service calls /users/{userId} and holds the result as a myApp.emails.UserData object. Both of these objects are identical.
The problem here is I have to keep myApp.emails.UserData and myApp.users.UserData in sync, since in reality they represent the same info and are meant to be the "same" class. Say I update the name of a field in emails.UserData, I better remember to update it in users.UserData, otherwise things will break.
I know I could make a shared dependency called something like SharedDataObjects, define myApp.sharedDataObjects.UserData there, and just have both versions refer to that. For some reason, though, my gut feeling is that this is not a good solution... (maybe it is though?)
Are there any better ways of approaching this issue? Is there something fundamentally wrong with the way the webapp structured, and if so how could that be addressed?
It's always tempting to spot duplicated (or nearly duplicated) code and try to eliminate the duplication. An obvious path in that direction is to make a shared library (or module) and have different services (or applications) depend on it.
Doing so, however, increases coupling between services/applications/modules, which has its own set of drawbacks in many contexts. Especially in a microservices architecture*, that kind of coupling often leads to headaches. It can even lead to loss of the value of using microservices* in the first place.
If two services* are supposed to be independent but integrated, they have an integration protocol of some kind. For REST services, for example, that's almost always HTTP and JSON**. By introducing a shared library, you have coupled them in a binary way, separate from (and more binding than) HTTP and JSON. Not a good situation; experience has taught many of us that painful lesson.
Instead, focus on the public interfaces that each service exposes, and use appropriate versioning to evolve those interfaces when needed. Don't worry about some duplicate-looking classes, especially if they're anemic POJOs; that's a minor concern compared to a tightly-coupled set of services that are supposed to be independent.
Not that shared code libraries are inherently bad, they're not and they have true value in many ways. Rather, my point is you need to make sure each service maintains its own definition of what important things are, so that each can remain as independent as possible - and evolve independently as much as possible.
By the way, this is somewhat related to the concept of Bounded Contexts; you might want to read more about it if you're working with microservices*.
*or whatever you call your architecture of independently-managed services/modules/apps
** Could also be some kind of messaging/queueing platform, as another example
With organizations which are slow to adapt to modern technology finally junking EJBs and getting ready to transform to SpringBoot, Microservices, REST, Angular, there are some some questions application design. One being about TransferObjects and Business Objects
When the call comes to the REST Controller, is it still popular to populate a TO (POJO) and then make Service call, which in turn populates a BusinessObject and then calls a Repository service?
OR
At the REST Controller layer do we directly populate the BO and send it to the Service (This does not make any sense to me, because a BO is populated only during the execution business logic).
If nowadays its still Option 1, then how do we avoid writing to exactly similar POJO classes in most cases (in order to use BeanUtils.copyProperties()), with the BO decorated with #Id, #Column etc.
To elaborate on #Turing85's comments...
Option 1 usually (see the end of my answer) makes infinitely more sense. It's a question of responsibility (purpose) and change; the two logical components you refer to, a REST API and a repository / system service:
Responsibility: a REST service cares about working with its callers, so ideally when designing a REST API you should be involving someone from the client-side (client as in caller), because if the API doesn't work for them it's not going to be an effective API. On the other hand, repositories are somewhat self-centered, and may need to consider things that are or no interest to API callers (and vis-versa).
Change: if you pay attention to design principles, like SOLID, you'll know that part of a system should do one job - as a way of limiting the reasons why it needs to change (see: SRP). Trying to use one object across both outward-facing API's, and inward-facing repositories, is asking for trouble because it's trying to do too much - its trying to help solve problems in two very different parts of the wider solution, and thee both have very different change drivers working against them. Turning85's comment about the persistence layer stems from the same idea.
"Option 1 usually makes infinitely more sense":
One case where the REST API's objects will / can bear a very close resemblance to those that hit the actual repository (or even be reused, I guess) is when the REST API is a System API - i.e. a dedicated façade / proxy to the repository. In this case, the System API is largely driven by the repository i.e. the main change driver is just the repository.
After researching a bit I agree, keeping things simple will result in simpler code. I found a nice simple way to take care of this manual work.
https://www.baeldung.com/entity-to-and-from-dto-for-a-java-spring-application
I am building a service that depends on another service. A typical Service oriented architecture. The service i am dependent on exposes some API and data types. I am confused should i be converting the object types exposed by that service into specific objects which my service understands. I do expect their service to change with time as these are two different services. I have two options:
Directly use those data types in my service and pass those in methods.
Transform those into specific data types which only my service understands. ( objects will look exactly same if i do this with 0 changes ).
I tried to answer these questions but still could not make the final call. I need help in making this decision.
Why should I have encapsulated/transformed types ?
To prevent building every time they build changes in the service.
To prevent widespread changes ( adapter pattern ) : Changes to the wire
format will lead me to change only the encapsulating classes.
Why should I not have the changes for the types encapsulated ?
The classes will look exactly same as the wire format classes. ( Useless effort to maintain extra classes )
As i understand the impact will be same if i go with either approach. Help ?
I am no architect or SOA specialist, so excuse me if I am saying anything stupid :-)
But I really think the way here is to keep your services simple.
In your shoes, I'd just directly use the existent API. I would not spent any time wrapping or adapting the methods into another API. Your second service (that uses the existent first service) business logic should take care of this convertion, IMO, except if you're being forced to do something that is really expensive with the existent API.
Remember that services are mutable. They're software. They have bugs, business logic changes as time goes and you'll have to change the API and sometimes you'll have to keep older methods compatible for other service consumers. You probably don't want to maintain two APIs that provide the same information without any good practical reason. Not for twice the maintenance work.
Creating another API just to adapt the data format sounds to me a little like that old "DTOs are evil" flame war. And I think a very few people write about the advantages of using DTO nowadays :-)
This is sort of opinion based question, so my opinion is, you should make your own data-types to let your piece of code understand what should be contained in which variable.
I think of services as a data provider, which accepts certain request and fulfill our needs and in return may give us some data. I think role of service is just providing services to client.
It should be responsiblity of client to accept the data returned by service and store them in certain data-structure as there can be n different clients for single service and they can have n different requirements which may lead them to design client specific data-structure to contain data.
Also, as you said client service is subject to change over the period of time, then if you make your own data-structure, then you will need to make change in one single place, and rest of your code will be safe.
I have an application with 3 layers: GWT-RPC, Business and DAO
In every layer I have different beans. For example for a User I'd have UserRPC (for the UI), User (business) and UserDTO (to persist). At every layer change I reconvert the objects.
The main problem are the enumerations. The enumerations are exactly the same across the layers but I need to replicate them in order to keep the layer independence.
Any suggestion?
(Short answer)
If they are truly the same you want to create a commons project / jar . Inside here are all the things that are common to the entire application, such as utility classes, enumerations, etc.
(Long answer)
Consider the overall architecture of your system, there should not be any need to have duplicated data in any tier. If this is occurring it means there is a flaw in the design of the system and a scenario of tight coupling may be on the horizon. This is part of the reason why when developing software engineers often (should) document the interfaces (contracts / api) before doing any implementation. Once these interfaces are approved and there is limited to no chance of duplication the actual implementation can begin. This would catch the scenario of X number of enumerations, which are constants, being created at each individual tier of the application. Also, keep in mind that with enumerations that if something in your business tier changes that would impact an enumeration you must recompile the code and redeploy. You must take care to not have an enumeration being used as a catch all for all of your system constants as this is an entirely different issue you may have to deal with.
The enumerations are exactly the same across the layers but I need to
replicate them in order to keep the layer independence.
Data duplication is never a good idea. How do you make sure that updates to one enumeration are reflected in other layers? You should probably create a Utility layer where you put all common classes there.
I am building a small website for fun/learning using a fairly standard Web/Service/Data Access layered design.
To save me from constantly having to create instances of my service layer/data access layer classes, I have made the methods in them all static. I shouldn't get concurrency issues as they use local variables etc and do not share any resources (things are simple enough for this at the moment).
As far as I can see the only trade-off for this is that I am not really following a true OO approach, but then again it keeps the code much cleaner.
Is there any reason this would not be a viable approach? What sort of problems might arise later on? Would it be better to have a "factory" class that can return me instances of the service and data layer classes as needed?
You know those rides at the amusement park where they say "please keep your hands and feet inside the ride at all times"? It turns out the ride is a lot more fun if you don't. The only real trade-off is that you're not really following a true keeping-your-hands-and-feet-inside-the-ride-at-all-times approach.
The point is this -- there is a reason you should follow a "true OO approach", just as there's a reason to keep your hands and feet inside the ride -- it's great fun until you start bleeding everywhere.
The way you describe it, this isn't the "wrong" approach per se but I don't really see the problem you're trying to avoid. Can't you just create a single instance of these business objects when the server starts up and pass them to your servlets as needed?
If you're ready to throw OO out the window you might want to check out the Singleton pattern as well.
Disadvantages:
You will be unable to write unit tests as you will be unable to write mock data access/business logic objects to test against.
You will have concurrency problems as different threads try to access the static code at the same time - or if you use synchronized static methods you will end up with threads queuing up to use the static methods.
You will not be able to use instance variables, which will become a restriction as the code becomes more complex.
It will be more difficult to replace elements of the business or data access layers if you need to.
If you intend to write your application in this manner you would be better off using a language designed to work in this way, such as PHP.
You would be better off going for non-static business/data access layer classes by either:
Using the singleton pattern (creating a single instance of each class and sharing them among threads)...
Or creating instances of the classes in each thread as and when they are needed.
Keep in mind that each user/session connected to your application will be running in it's own thread - so your web application is inherently multi-threaded.
I don't really see the advantage to your design, and there are many things that could go wrong. You are saving a line of code, maybe? Here's some disadvantages to your approach:
You cannot easily replace implementations of your business logic
You cannot defined instance variables to facilitate breaking up logic into multiple methods
Your assumption that multi-threaded issues will not arise is almost certainly wrong
You cannot easily mock them for testing
I really don't see that the omission of one line of code is buying you anything.
It's not really an "OO Design" issue, but more of an appropriateness. Why are you even using Java in such a procedural way? Surely PHP would be more appropriate to this kind of design (and actually save you time by not having to compile and deploy).
I would just make your business layer non-static; it will make it so much easier for to maintain, change, and evolve your application.
You may have difficulty unit-testing your objects with this type of architecture. For example, if you have a layer of business objects that reference your static data access layer, it could be difficult to test the business layer because you won't be able to easily use mock data access objects. That is, when testing your business layer, you probably won't want to use the "real" methods in the data access layer because they will make unwanted changes to your database. If your data access layer was not static, you could provide mock data access objects to your business layer for testing purposes.
I would think that you will have concurrency issues with all static methods with multiple users. The web layer will thread out concurrent users. Can all your static methods handle this? Perhaps, but won't they constantly be locked in queuing the requests in single file? I'm not sure, never tried your idea.