We are following Domain Driven Design for the implementation of a large website.
However by putting the behaviour on the domain objects we are ending up with some very large classes.
For example on our WebsiteUser object, we have many many methods - e.g. dealing with passwords, order history, refunds, customer segmentation. All of these methods are directly related to the user. Many of these methods delegate internally to other child object but
this still results in some very large classes.
I'm keen to avoid exposing lots of child objects
e.g. user.getOrderHistory().getLatestOrder().
What other strategies can be used to avoid this problems?
The issues you are seeing aren't caused by Domain Driven Design, but rather by a lack of separation of concerns. Domain Driven Design isn't just about placing data and behavior together.
The first thing I would recommend is taking a day or so and reading Domain Driven Design Quickly available as a free download from Info-Q. This will provide an overview of the different types of domain objects: entities, value objects, services, repositories, and factories.
The second thing I would recommend is to go read up on the Single Responsibility Principle.
The third thing I would recommend is that you begin to immerse yourself in Test Driven Development. While learning to design by writing tests first won't necessarily make you designs great, they tend to guide you toward loosely coupled designs and reveal design issues earlier.
In the example you provided, WebsiteUser definitely has way too many responsibilities. In fact, you may not have a need for WebsiteUser at all as users are generally represented by an ISecurityPrincipal.
It's a bit hard to suggest exactly how you should approach your design given the lack of business context, but I would first recommend doing some brain-storming by creating some index cards representing each of the major nouns you have in your system (e.g. Customer, Order, Receipt, Product, etc.). Write down candidate class names at the top, what responsibilities you feel are inherent to the class off to the left, and the classes it will collaborate with to the right. If some behavior doesn't feel like it belongs on any of the objects, it's probably a good service candidate (i.e. AuthenticationService). Spread the cards out on the table with your colleges and discuss. Don't make too much of this though, as this is really only intended as a brainstorming design exercise. It can be a little easier to do this at times than using a whiteboard because you can move things around.
Long term, you should really pick up the book Domain Driven Design by Eric Evans. It's a big read, but well worth your time. I'd also recommend you pick up either
Agile Software Development, Principles, Patterns, and Practices or Agile Principles, Patterns, and Practices in C# depending on your language preference.
Although real humans have lots of responsibilities, you're heading towards the God object anti-pattern.
As others have hinted, you should extract those responsibilities into separate Repositories and/or Domain Services. E.g.:
SecurityService.Authenticate(credentials, customer)
OrderRepository.GetOrderHistoryFor(Customer)
RefundsService.StartRefundProcess(order)
Be specific with naming conventions (i.e. use OrderRepository or OrderService, instead of OrderManager)
You've run into this problem because of convenience. i.e. it's convenient to treat a WebsiteUser as an aggregate root, and to access everything through it.
If you place more emphasis on clarity instead of convenience, it should help separate these concerns. Unfortunately, it does mean that team members must now be aware of the new Services.
Another way to think of it: just as Entities shouldn't perform their own persistence (which is why we use Repositories), your WebsiteUser should not handle Refunds/Segmentation/etc.
Hope that helps!
A very simple rule of thumb to follow is "most of the methods in your class HAVE to use most of the instance variables in your class" - if you follow this rule the classes will be automatically of the right size.
I ran into the same problem, and I found that using child "manager" objects was the best solution in our case.
For example, in your case, you might have:
User u = ...;
OrderHistoryManager histMan = user.getOrderHistoryManager();
Then you can use the histMan for anything you want. Obviously you thought of this, but I don't know why you want to avoid it. It seperates concerns when you have objects which seem to do too much.
Think about it this way. If you had a "Human" object, and you had to implement the chew() method. Would you put it on the Human object or the Mouth child object.
You may want to consider inversing some things. For example, a Customer doesn't need to have an Order property (or a history of orders) - you can leave those out of the Customer class. So instead of
public void doSomethingWithOrders(Customer customer, Calendar from, Calendar to) {
List = customer.getOrders(from, to);
for (Order order : orders) {
order.doSomething();
}
}
you could instead do:
public void doSomethingWithOrders(Customer customer, Calendar from, Calendar to) {
List = orderService.getOrders(customer, from, to);
for (Order order : orders) {
order.doSomething();
}
}
This is 'looser' coupling, but still you can get all the orders belonging to a customer. I'm sure there's smarter people than me that have the right names and links referring to the above.
I believe that your problem is actually related to Bounded Contexts. For what I see, "dealing with passwords, order history, refunds, customer segmentation", each one of these can be a bounded context. Therefore, you might consider splitting your WebsiteUser into multiple entities, each one corresponding to a context. There may arise some duplication, but you gain focus on your domain and get rid off very large classes with multiple responsibilities.
Related
I'm curious as to which is the better practice and the reasoning behind it, for this example I'm going to be using a social application which contains a 'friends' and a 'ignore' list with some custom logic based on them, (For sending messages directly, etc)
Which would be the better practice, and why?
Scenario 1:
class user {
List<> friends;
List<> ignores;
...
logical methods here
}
Scenario 2:
class User {
Social social;
...
}
class Social {
List<> friends;
List<> ignores;
...
logical methods here
}
I've seen both scenarios used throughout numerous applications and I'm curious as to which is the "Correct" way to lay it out in java, these will have methods such as
#addFriend(User user)
check ignore
check valid user
check other info
add to list
end
#getFriend(int id)
find friend by id
check online status
return friend
It seems like while have a 'Social' class may be a cleaner approach, does it really follow good practices? Seems like it'd use more memory/user just for cleaner code.
The reason why you have such constructs as your Social, most of the time, is that they represent a logical set of data and operations which is needed for different entities in your application.
If nothing other than User has those properties and actions, then there is no point in doing it separately from User. But you may design it separately anyway, for future uses (for example, if you want to be able to expand it later and you believe there will be other entities which will need Social functionality).
Looking at this from an object-oriented viewpoint, it means that the Social is a type. And then you have to ask yourself, is whether your User is_a Social or whether your User has_a Social. Does it make sense to say that the user has a "social subsystem" or is the user a "social object"? If the correct relation is is_a, then User should extend Social. If not, it should have a Social member, such as you described.
However, in Java, since you can't have multiple inheritance of implementation, sometimes your type may inherit from several types, and you have to decide which of them to extend. Many times, you simulate multiple inheritance of implementation, by having a member of what should have been the "second parent class", declare all the methods in your class, and delegate them to that member.
So the general guidelines are, more or less:
If in your application's domain, the only class where it will make sense to have friends and ignores and their operations is User, and no other conceivable entity would ever need them, then implement them directly in User.
If other entities may need similar functionality, and not all of them extend User anyway, you may consider this functionality to be an entity or class in its own right, and then you should have every class which has an is_a relationship to this entity extend it.
If Java's limitations of multiple inheritance don't allow extending directly, as it makes more sense for the class to extend some other class, you should embed an object and delegate the operations.
There may be other practical reasons to separate the Social entity from User, despite User being the only class to use them. For example, if you have several different possible implementations of "social" behavior, you may want to be able to use various Social subclasses as "plug-ins" inside User, rather than subclassing User.
Don't worry about memory so early. Go for readable/cleaner code. Premature optimization is root of all evil.
This is really based on the logic of your program. But consider that increasing the number of classes unnecessarily, is not good practice.
In your example, if the User class only contains a Social field, and you will just delegate all the method calls to the Social class, then go with scenario one.
On the other hand, if the User class has many more fields, like name, date of joining ... then it would be even better to create a separate class for such fields such as UserInfo in order to better structure your program and enhance code readability.
Now the main concerns are not the memory or performance costs of class structure.
Way more important are readability and clean code, AND the possibility to persist domain classes in a DB in the most simple and efficient way.
The later include composition or aggregation concern which is specific for different DB's.
You should care about the design aspects becoz with this you will have maintainable,scalable and readable code.
Now going by your example , i find second scenario as good case as it follows the SRP(Single Responsibilty Principle)
Don't worry about memory here as it wont make iota of difference here.
So do you want to do something like:
for(Connection connection : userSocialConnections ){
sendMessageTo(connection);
}
If so, then the method sendMessageTo would need to accept a connection (friend or ignored, basically a user) and probably if the runtype connection is ignored (or has blocked the user) then the sendMessageTo will return without sending a message polymorphically. This would require that in java that the IgnoredPeople And Friends are subtypes of something called as Connection(or people or anything you like; in fact, a connection is also a user - current or potential, isn't it?). This approach seems (to me) more like thinking in problem domain. Storing as two list inside user or inside social inside user does not matter much as long as they both (ignored and friends) have a common interface.
I would ask, what all other scenarios can be there for user's friends or ignored list. Do they need to be loaded lazily or stored separately.
I'm designing a game, but I can't quite get my head around the inheritance structure. I'm normally fairly good at it, but this one just has too much overlap and I can't decide on it all.
I'm seeking to model sailing vessels - think the Age of Sail. Presumably therefore everything extends a Vessel class.
There are then several types of vessel style: rowed (galleys, canoes), square-rig, fore-and-aft rig, with different behaviour. Each of these is further subdivided into several other types. I can't decide whether this should be a series of interfaces or extensions of Vessel. Note also that there can be some cross over (a vessel can be both rowed and square rigged) which leads me to think interfaces?
Ships also have different behaviours: merchant vessels, men of war, privateer, pirates. I really can't work out whether this should be an interface or an extension of another class. There is no crossover of type in this case, however.
Finally there are several behaviours which individual ships can have. Merchants may be in a convoy (defend themselves) or independent (run away). Men of war almost always attack unless heavily outgunned... but may work in fleets, squadrons or independently. Privateers and pirates only attack if weaker - usually independently but occasionally in pairs. I'm assuming that this should be an interface too?
My big problem is that each style of ship (frigate, battleship etc) can fulfil almost any of these roles, so I can't build a simple solid inheritance structure. Frigate can't extend man-o-war because some are privateers. Sloop can't extend square rigged because some are fore and aft rigged. etcetc.
Any thoughts would be appreciated, I'm at a bit of a loose end.
Thanks
Make the "behavior" part as interfaces. That will help you assigning different behaviors to different ships without problem. Strategy Pattern is helpful here. In a nutshell, it states that the changeable and the constant properties should be separated.
For different means of movements, composition sounds like the most suitable answer at this moment.
Regarding the "but may work in fleets, squadrons or independently. Privateers and pirates only attack if weaker - usually independently but occasionally in pairs." part, I guess this has nothing to do with the inheritance tree. You can make "groups" of classes depending up on your need.
This might help you:
"There are then several types of vessel style:..." is specifying different possible behaviors. So "Movable" interface and its subclasses are for that. In the "Vessel" class, you can have a member of type "Movable". Since "Movable" is an interface, any class which implements it, is assignable to this member. So any subclass of Vessel can have any possible behavior, which we cna change at runtime as well. You can also make it an ArrayList. (Not sure if you actually want to do it or not). But if you need multiple different behaviors for a same vessel, you can do it.
When you say "Ships also have different behaviours:..." it feels like separate classes extending Vessel will satisfy this requirement. The sentence "There is no crossover of type in this case, however." makes life easier.
For the nest para "Finally there are several behaviours which individual ships can have...", you should add one more member for different possible behaviors. It will mostly be an ArrayList as one vessel will have multiple attack modes.
Fro the last para, if you can give some more details, I may be able to give some more ideas.
Ok, here are some ideas:
Vessels have one or more means of propulsion (oars, sails, etc.), which you can model by composition (e.g. have a list of propulsion methods).
Vessels use one of a variety of strategies (use the strategy pattern -- see http://en.wikipedia.org/wiki/Strategy_pattern -- for this)
Strategies which depend on the presence of other nearby ships will need some way of querying for those other ships -- so you'll need some sort of data structure that allows you to find which objects are near which other objects (look into the sort of data structures that are used for broad-phase collision detection)
As an alternative to the strategy pattern, you could switch to using a component-based design. In this setup, a vessel would be composed of one or more propulsion components, a strategy component, etc. You could then combine individual components as you see fit to make different vessels.
As an extra bonus, component-based designs are very helpful if you want your game to be data-driven, because you can just write a saver/loader for each different type of component, rather than for each possible type of vessel.
You might want to see here if you're interested in this sort of approach:
http://cowboyprogramming.com/2007/01/05/evolve-your-heirachy/
I want to provide a bit of advice based on the second paragraph of Bhusan's answer, which I quote here in full:
"Regarding the "but may work in fleets, squadrons or independently. Privateers and pirates only attack if weaker - usually independently but occasionally in pairs." part, I guess this has nothing to do with the inheritance tree. You can make "groups" of classes depending up on your need."
This leads me to believe that you additionally may want to consider the Composite pattern for certain groups of ships, at least those that are comprised of ships that all share the same behavior. See http://en.wikipedia.org/wiki/Composite_pattern where it is written that "the composite pattern describes that a group of objects are to be treated in the same way as a single instance of an object."
For instance you say "Merchants may be in a convoy (defend themselves)", but presumably they can defend themselves individually as well? This is all easier said than done of course, and my advice to you is to not over-think it and start with a very small subset of what you want to do as a protoype
You should separate the types and use the strategy pattern.
Immutable properties should be bound to the inheritance tree (e.g. a frigate won't turn into canoe, these are exact, non-behavioral types, inheriting from vessel) and everything that may change should be stored as references to behavioral types which are intrachangable. (Man-o-war is a behavioral type)
AI should be handled separately, for instance with states, but that is also have to be in a different module in your architecture.
Instead of thinking about it as strictly inheritance, I think you need to think about object Composition and how that can help make things easier.
For example: a Ship has a Behaviour. (Composition... the ship delegates to the behaviour to determine how to react to X situation)
A Pirate is a Behaviour (Inheritance from a Behaviour interface)...
You should not extend the Vessel class. Rather, a Vessel object should contain other objects that describe it. (This is known as Dependency Injection, to be annoyingly pedantic--forget I said that.) You have a Propulsion class, with instances for square-sails, fore-and-aft, and oars. You might want special instances of each for big and small hulls. You have a Behavior class to handle their attitude. Sometimes a simple integer works as well as a class. Your Armament class could be, instead, just a number of guns. (Or two numbers, one for poundage.) If you expect to ram, or have a ship loaded with rockets, or need to distinguish between long guns and carronades, you might need to go back to using the a class.
This lets you switch characteristics on the fly if you want--though you probably don't. Still, you can go from sail to using sweeps, or switch bold ship-of-the-line behavior with save-the-cargo behavior of a merchant to implement a captain losing his nerve. Anyway, it's there if you need it.
I have a domain object called VehicleRecord which is a hibernate entity. CRUD operations on these VehicleRecords are handled through an entity access object implemented as a stateless session bean interface
#Local
interface VehicleRecordEao {
void add(VehicleRecord record);
List<VehicleRecord> findAll();
...
}
#Stateless
class HibernateVehicleRecordEaoBean implements VehicleRecordEao { ... }
From a business layer perspective removing and adding these records is more than just a CRUD operation. There may be logging and security requirements for example. To provide these operations to clients sessions beans are created in a business layer.
#Local
interface VehicleRecordManager {
void createVehicleRecord(VehicleRecord record);
List<VehicleRecord> findAll(String make, String model);
...
}
#Stateless
class VehicleRecordManagerBean {
public void createVehicleRecordManager(VehicleRecord record) {
//business rules such as logging, security,
// perhaps a required web service interaction
//add the new record with the Eao bean
}
...
}
Controllers work between the presentation layer and the above business layer to get work done, converting between presentation objects (like forms) and entities as necessary.
Something here is not right (smells). I have a class called Manager which has got to be a red flag, but several examples in EJB books actually suggest this kind of high level class and tend to use the name manager themselves. I am new to EJB design but in OO design making high level classes called Manager or Handler or Utility screams of procedural and requires one to rethink their design.
Are these procedural-utility-class session beans the normal pattern, or is bad to organize your session by a bunch of methods related only to the entity they operate on? Are there any naming conventions for session beans? Should the Eao's and business session beans work at the same layer?
If there is a less smelly alternative to this pattern I would love to know, thanks.
Your approach is more-or-less standard. Yes, at a fundamental level this is a procedural approach and goes hand-in-hand with what has been dubbed the Anemic Domain Model "anti-pattern". The alternative is to incorporate your business logic into your domain model so as to create a more OO design where your operations are coupled with your DO's. If you were to go down this route you should be aware of the inherent pros and cons. If you feel that your approach makes the most sense, is easy to understand, test, scale, etc... Then role with it. I have worked on several n-tier projects where this exact "procedural" style is used -- rest assured it is quite standard in EE applications to do things this way.
It's an age old discussion. Should a bread bake itself, or does an oven bake it? Does a message sends itself, or does the postoffice do it? Do I send a message to my friend Pete by asking him to send a message to himself?
According to those who came up with the term ADM, it's more "natural" and more "OO" to let each object do all those things itself. Taken to the extreme (which no one advocates of course, but just as an example) class User would contain all logic for the entire application, since ultimately the user does everything.
If you use slim entities (entities containing only data) and service or DAO objects to work with them, then you're not necessarily not doing OO. There is still a lot of OO going around. Services implement interfaces, inherit from base classes, have encapsulated state (e.g. the EntityManager in JPA), have transparent interceptors via dynamic proxies (e.g. to check security and start/commit a transaction), etc etc.
The term "Anemic" has a definite negative association, but calling the same thing "Slim" actually sounds good. This isn't just a case of using different names to hide code smells, but represents a different thinking among different groups of people.
I think it's up to your taste, and depending on the complexity of your application - these things matter less / more. I think in design, you need to simply thing around the line of:
"If someone new came in without any prior knowledge of the system in question - is it intuitive enough for that person to follow and trace the codes, and straight forward to find where things are"
In your case - naming the EJB related to the Entity Object makes it straight forward and simple - what I dont get is why did you separate it to 2 classes EAO and Manager. Why dont you just combine them into one, so if the EJB/Bean class deals with VehicleRecord Entity, then it will be the "VehicleRecordEAO" or "VehicleRecordManager" or "VehicleRecordAccess" or anything really.
I think EAO / DAO / Access sounds more like getter / setter - or any other simple operations. I dont see anything wrong with "Manager" and make it consistent across the board that all business layer will be called "Manager".
Or if you feel better, think of it as the Facade Pattern - so you can call your business layer (the Manager) as VehicleRecordFacade and VehicleRecordFacadeBean.
That way you basically follow the name and concept of Facade pattern, where it becomes intermediary between application layer and the data layer.
Something here is not right (smells).
I have a class called Manager which
has got to be a red flag
My answer here is towards this concern of yours.
Yes. It is a red flag. Naming a class like "VehicleRecordManager" would be a code smell suggesting that Single responsibility principle would be violated sooner or later.
To elaborate, let me take a few use cases to deal with VehicleRecord
Buy a vehicle
Rent a vehicle
Search for a vehicle
Sell a vehicle
Find Dealers
In most Java applications when we write up a "VehicleService" ( or "VehicleManager") all the above operations would be placed in this class ! Well, this one is easy to do, but hard to maintain. And certainly this class has a lot of responsibilities, hence many reasons to change. (violating Single responsibility principle )
Would calling it VehicleDao eliminate some of the smelliness? A simple change, but indicates clearly it's concerned with data access concerns.
Whilst there are many good examples on this forum that contain examples of coupling and cohesion, I am struggling to apply it to my code fully. I can identify parts in my code that may need changing. Would any Java experts be able to take a look at my code and explain to me what aspects are good and bad. I don't mind changing it myself at all. It's just that many people seem to disagree with each other and I'm finding it hard to actually understand what principles to follow...
First, I'd like to say that the primary reason you get such varying answers is that this really does become an art over time. Many of the opinions you get don't boil down to a hard fast rule or fact, more it comes down to general experience. After 10-20 years doing this, you start to remember what things you did that caused pain, and how you avoided doing them again. Many answers work for some problems, but it's the individual's experience that determines their opinion.
There is really only 1 really big thing I would change in your code. I would consider looking into what's called the Command Pattern. Information on this shouldn't be difficult to find either on the web or in the GoF book.
The primary idea is that each of your commands "add child", "add parent" become a separate class. The logic for a single command is enclosed in a single small class that is easy to test and modify. That class should then be "executed" to do the work from your main class. In this way, your main class only has to deal with command line parsing, and can lose most of it's knowledge of a FamilyTree. It just has to know what command line maps into which Command classes and kick them off.
That's my 2 cents.
I can recommend Alan's and James's book Design Patterns explained -- A new perspective on object-oriented design (ISBN-13: 978-0321247148):
It's a great book about has-a and is-a decissions, including cohesion and coupling in object-oriented design.
In short:
Cohesion in software engineering, as in real life, is how much the elements consisting a whole(in our case let's say a class) can be said that they actually belong together. Thus, it is a measure of how strongly related each piece of functionality expressed by the source code of a software module is.
One way of looking at cohesion in terms of OO is if the methods in the class are using any of the private attributes.
Now the discussion is bigger than this but High Cohesion (or the cohesion's best type - the functional cohesion) is when parts of a module are grouped because they all contribute to a single well-defined task of the module.
Coupling in simple words, is how much one component (again, imagine a class, although not necessarily) knows about the inner workings or inner elements of another one, i.e. how much knowledge it has of the other component.
Loose coupling is a method of interconnecting the components in a system or network so that those components, depend on each other to the least extent practically possible…
In long:
I wrote a blog post about this. It discusses all this in much detail, with examples etc. It also explains the benefits of why you should follow these principles. I think it could help...
Coupling defines the degree to which each component depends on other components in the system. Given two components A and B ,how much code in B must change if A changes.
Cohesion defines the measure of how coherent or strongly related the various functions of a single software component are.It refers to what the class does.
Low cohesion would mean that the class does a great variety of actions and is not focused on what it should do. High cohesion would then mean that the class is focused on what it should be doing, i.e. only methods relating to the intention of the class.
Note: Good APIs exhibit loose coupling and high cohesion.
One particularly abhorrent form of tight coupling that should always be avoided is having two components that depend on each other directly or indirectly, that is, a dependency cycle or circular dependency.
Detailed info in below link
http://softwarematerial.blogspot.sg/2015/12/coupling-and-cohesion.html
Not sure if the title captures what I'm trying to say here.
When designing in OO should I be splitting my objects up into their most specific areas - so if I have a factory object that deals with creating objects but later on i come across a way of creating objects for another purpose even though they may be the same objects is it worth creating a seperate fcatory or just add to the exsiting.
My biggest worry is bulking up classes with tons of stuff, or splitting objects and diluting my projects into a sea of classes.
Any help?
EDIT:
I guess on a side note/sub topic part of me wants to find out the level of granularity you should use in a program. Kind of, how low should you go?
My biggest worry is bulking up classes with tons of stuff, or
splitting objects and diluting my
projects into a sea of classes
This is a very valid point and in any even reasonably sized project, extremely difficult to get right up front especially because realistically, requirements themselves evolve over time in most cases.
This is where "Refactoring" come in. You design based on what you know at any given point and try not too make too many leaps of faith as to what you think the system MAY evolve to.
Given that you know what you are building right now, you design your classes trying to make the best possible use of OO concepts - eg encapsulation / polymorphism. This is itself, like others have noted as well, can be notoriously difficult to achieve and thats where experience, both in designing OO systems as well as knowledge of the domain can really come in handy.
Design based on what you know --> Build It --> Review it --> Refactor it --> Re-design --> and it goes on and on..
Finding the correct level of detail and responsibility is what makes OOP design so difficult. We can help you with a specific case but not with anything this general. If there were algorithms or strict methodologies of how to solve this, everyone could be an OOP designer.
A rule of thumb I like for deciding "is this getting too big now?" is "can I explain the purpose of it concisely?" If you start having to introduce caveats and lots of weasel words to explain the functions of a component of your design (be it class, member variable, method or whatever) it might be a good indicator that it's getting too complex and should be split up.
In your specific case, if you already have a factory object then the DRY Principle (Don't Repeat Yourself) would say that it's a bad idea to create another factory that does the same thing.
Is this an actual problem that you face? Or merely a fear about how your code might grow in the future?
If you are using the same type of object to solve drastically different problems then you may need to redesign the class to focus on seperation of concerns. If you need a more specific answer, you will need to provide an example of a type of class that would need this functionality.
I might have worded things badly in
the Q. I guess I wouldnt be repeating
myself its just more of a case of
where to put the code, it could be
added to an exsiting factory that
creates design objects for exporing
data to excel spreadsheets. On the
other hand I could see it could also
have its own factory for importing
excel data. Both factories would
produce the same objects but the inner
workings are completely different. –
If you aren't doing or plan on doing any class abstraction (subclassing or using interfaces) you may not need to use the factory pattern at all. The factory pattern is generally best suited for supplying objects of a base class type or that implement a specific interface.
Both
factories would produce the same
objects but the inner workings are
completely different.
Not sure if I've understood you correctly, but this sounds like a candidate for the AbstractFactory pattern.