This blog post is a collection of notes on some more common design patterns. Each design pattern is explained in simple terms and includes an example.
This is written mainly for beginners, but it’s also useful if you need to refreshen your understanding of a certain design pattern. I highly suggest you to further explore each pattern covered here, since this post’s goal is to simply give you an overview of them. I’ve included some additional sources at the end of the post.
I’ve written this mainly because it’s not always easy to understand what a pattern is about in its essence, since many of the explanations out there are either domain-specific or include examples that are somewhat complex. In my opinion, it’s a lot easier to to undestand them once you have a general idea of what the pattern does. This is why I’ve used somewhat “childish” examples here: those should help you concentrate on the pattern and not the domain details.
All of the code examples are in Java, but only the basic syntax is used, so if you’re coming from a language like C# or have basic Java knowledge, you should have not problems in undestanding this post. Even though a specific programming language is used here, the concepts are domain-independent and can be applied to other languages. The default access modifiers for member variables are used, to make the code interpertation more accessible to beginners.
The Singleton Pattern ensures that there is only one instance of the class and provides a global point to access that instance.
Java’s implementation of Singleton makes use of a private constructor (to make sure that no one in the application can call the new Singleton(), thus possibly creating more than one instance of the class) and a static method, combined with a static variable (to make sure that there is only one instance of the class in the application, instead of calling “new”, we ask the class itself to provide an instance and since we never directly instanciate it, the method has to be static). The class itself will be responsible for keeping track of its sole instance.
Below is an example implementation in Java. This version uses lazy initialization (the instance of the class isn’t created until it’s first requested via the getInstance() method).
The Command Pattern encapsulates a request as an object, thereby letting you parameterize other objects (the client) with different requests, queue or log requests, and support undoable operations.
- encapsulates a request by binding together a set of actions on a specific reciever. To achieve this, the actions and the reciever are packaged into one single object that only exposes one method: execute() (it can also have other like undo(), etc). An object from outside doesn’t know what actions will be performed, all it knows is that if they call execute() their request will be serviced.
- parameterize other objects with a command, in a sense that any command object can be passed to the client. The client doesn’t know (or care) what the command is, all it does is call the execute() method.
- the Command Pattern makes it easy to implement queue, log and undo operations
The key in this pattern is an abstract Command class (or an interface), which declares an interface for executing operations. In its most simple form, this inteface has an execute() method. A concrete commad then has an instance variable of the receiever, whose methods are called in the execute(). The command itself doesn’t perform any complex operations, those have to be done by the receiever, and that’s what happens in the execute() method, receiver’s methods get called.
Let’s say you have a remote control that controls the volume and the on/off state of your awesome sound system. For simplicity, let’s assume that your remote control is actually a single button, which you can reprogram to do different actions (like voulume up, volume donw, on and off).
The first thing to do is to create the Command interface, which in this version, will contain the minimum (i.e. only the execute() method).
Next step will be to create four command classes (TurnOnCommand, TurnOffCommand, VolumeUpCommand and VolumeDownCommand). Each of those classes will have a reference to the receiver (in this case, the sound system), the execute() method and a constructor.
Now, just to demonstrate how the Command Pattern will be used, there will also be a button, which will have a referece to the command, a method to activate the command (press()) and a setCommand() method, so that different commands can be assigned to the same button.
And just to show an example of usage, consider the main class below.
Which produces the following output:
ON! Sound is at 1 Sound is at 2 Sound is at 3 Sound is at 2 OFF!
Note that the command invoker doesn’t know what the command is doing or how it’s, doing it, all it knows is that it has an execute() method. Also note that the execute() is doing very little himself, the main work is done my the receiver through some method.
Implementing the “Undo” Operation
In its simplest form, the undo operation is pretty straight forward. For example, what would the undo be for the VolumeUpCommand? Well, if that command increases the volume by 1, then the undo would decrease it by one. So the VolumeUpCommand would now look something like this:
Besides that, the undo() method would also have to be added to the Command interface:
You could also store all of the commands in a list, for example, and that would allow you to go through that list, calling undo() on every command object in there (that would be like a history of actions and you can undo multiple times, just like pressing Ctrl + Z on your keyboard multiple times will do more than one undo).
In a similar manner you could also extend the pattern to other functionality, such as queing or logging.
The Iterator Pattern provides a way to access the elements of an aggregate object (an object that contains an aggregate of something) sequentially without exposing its underlying representation.
So let’s say you have an object that contains a collection of items and you want to iterate through every item. But there is a problem here: let’s say you know what items you’re accessing, but you don’t know how they are stored, they might be in an ArrayList, HashMap, LinkedList, etc. Now how can you abstract the internal implementation (i.e. how the items are stored) away and have an interface that allows you to access those elements in a consistent manner, whichever the internal implementation is. The answer is (you’ve guessed it!) the Iterator Pattern.
So as mentioned before the Iterator Pattern allows you to step through the elements of an aggregate without knowing how the things are represented under the hood. It also allows to write polymorphic code that works with any of the aggregates (it doesn’t matter if it’s an ArrayList, a HashTable or even a LinkedList, it’s an aggregate, that’s all we care about).
The Iterator Pattern takes the responsibility of traversing the elements and gives that responsibility to the iterator object, not the aggregate object. This makes all the sense, since the aggregate object doesn’t have to know how to iterate through items, all it has to know is how to store and manage them. It has to know which objects have been traversed already and which ones haven’t.
In its simplest form, the Iterator Pattern consists of an Iterator interface, which contains two methods: * next(), which returns the next object in the aggregation that hasn’t been iterated through yet. * hasNext() which returns a boolean indicating whether there are more items to be iterated through.
Let’s say you asked two of your friends to provide a list of their favorite songs. Your idea is simple: iterate through both of the lists (lists of songs, no one is talking about Java Lists here) and display the info of each one of the songs (title and author). You then realized that you didn’t specify the format in which you want the songs to be sent (ArrayList, HashTable, LinkedList, etc) and you need to be able to iterate over the lists in a consistent way, without depending on the type of the aggregation.
You’ll have to create a concrete iterator for every type of the aggregation, so let’s say one of your friends send the songs in an HashTable and the other in a LinkedList.
Below is a possible solution using the Iterartor Pattern.
The application above produces the following output:
Name: Still Dre Artist: Dr.Dre Name: Let's Ride Artist: The Game Name: What's My Name? Artist: Snoop Dogg Name: Nas Is Coming Artist: NAS Name: What Up Gangsta Artist: 50 Cent Name: Warrior Artist: Lloyd Banks
Note how by using the Iterator pattern we decoupled our application (in this case just the Main class) from the underlying implementation of each aggregation.
All we had to do to our friends code is implement the Iterator interface (which we defined) and add the iterator() method to their classes. If we didn’t use the pattern, when iterating through the list of songs in the Main class, we would have to make a separate method for each aggregation we wanted to traverse.
The Composite pattern allows you treat objects and compositions of objects uniformly. It allows you to compose objects into tree structures to represent part-hole-hierarquies (components can be divided into smaller and smaller components).
The idea here is really simple: you have an object A that supports some operations, then you have another object B that is an aggregation of objects of type A. The composite pattern allows you treat both of them using the same interface. Let’s say object A is a sheep and supports the sheer() operation. So to sheer a single sheep you simply call the sheer method on a sheep object. Now, object B is a group of sheep. How do you sheer a group of sheep? Simple, the same way you would sheer an individual sheep: by calling the sheer method on the object B. Note that the object B can contain sheep or other objects of the same type(of the type of object B).
So basically, the composite pattern allows us to ignore the differences between compositions of objects and individual objects.
In this example, we will use the sheep example introduced above. So the idea is simple: each sheep has a name and the only operation it supports is sheer().
Now, you sheer a sheep by simply calling the sheer() method on it, but how can you sheer a group of sheep? The same way, by calling the sheer() method on it. As mentioned before, the goal is to inore the differences between an individual object and a group of objects, so we will define an abstract class, which is extended by both: an individual sheep and a group of sheep (here, it’ll be called SheepComoponent). This approach will also allow SheepComponents to contain other SheepComponents.
Below is a possible solution to the problem:
The output of the application above is:
Sheering Sheep 1... Group Name: Sheep Group 1 --- Sheering Sheep 2... Sheering Sheep 3... Sheering Sheep 4... Sheering Sheep 5... Group Name: SheepGroup 2 --- Sheering Sheep 1... Group Name: Sheep Group 1 --- Sheering Sheep 2... Sheering Sheep 3... Sheering Sheep 4... Sheering Sheep 5... Sheering Sheep 6...
The Visitor pattern allows you to add new methods to the classes without changing them too much. You can add operations to a Composite structure without changing the structure itself.
Visitor is very useful when you have some unrelated operations that need to be performed on an object in an object structure and you don’t want to “pollute” their classes by adding new methods to them to perform those operations. This pattern allows you to keep related operations together defined in a separate class. This is very useful when your object structure is shared by many applications, but only some of those applications actually use those extra operations, since visitor allows you to put those operations only in the applications that need them.
So let’s say you have the class code written and now you’re wondering what changes you’ll have to make to the class for it to support the visitor pattern. Okay, make sure you have a piece of paper and pen to write them down. Not really, actually all you have to do is add an accept() method to your class (it’s a convention to call the method accept(), you can name it whatever you want). Yes, that’s it! All you have to do for your classes to support the visitor pattern is add a single method. That’s what we’re basically going to do, except we’re going to put this method in an Interface, we’ll call it “Visistable” (this name makes sense, since the class will be visted by a visitor).
Okay, now how do we create those “visitors”? Simple, first we create another interface called Visitor (again, this name is just a convention). And what will that interface contain? It will contain the new operations we want to add. Now, since we’re putting them in an interface, those operations have to be related (i.e. be somehow related, belong to the same “group”).
Now you might be wondering what will you do if you had more than just one group of operations? Well, in that case you would have to add another accept() method to your Visitable interface (use method overloading).
Let’s say you have a store and your store sells three products: drinks, food and gadgets. Each of those three objects has a price. Here is how your code looks now:
(Okay, a better choice would be to add an abstract class from which those methods derive, but I want to enforce the idea that the classes don’t have to be related in any way).
Now you are asked to be able to calculate taxes for each product(the tax is 21%). That’s where the visitor comes in. First you’ll create a Visitor interface and add three visit() methods there. Then you’ll create a new class: a concrete visitor, which implements the Visitor interface. Then, we’ll create a Visitable interface and add the accept() method to it, which takes an object of type Visitor as an argument. The Food, Drink and Gadgets classes will implement the Visitable interface. Now your code will look something like this:
Now you are asked to add yet another type of tax: a holiday tax. It’s basically the same as the previous one, except the tax value is now 18% and you always subtract two cents (0.02) from the value after tax (here we’ll ignore that you can get negative or zero values). Well, that’s simple, just create a new class HolidayTaxVisitor, implement the Visitor interface and add override the methods with the requested functionality. Your new class will look something like this:
And now an example application:
The output of the Main class above is the following:
Drink price after normal tax: 1.815 Drink price after holiday tax: 1.75 Food price after normal tax: 3.3275 Food price after holiday tax: 3.225 Gadget price after normal tax: 8.7725 Gadget price after holiday tax: 8.535
Now what if you wanted a different group of operations, not related in any way to tax calculation? Let’s say you wanted to be able to print out the name of the class of which an object is instance of. Well, in that case you’d have to create a new visitor interface, for example NameVisitor, create a concrete instance, which implements that interface, for example NormalNameVisitor and add a new accept() method to the Visitable interface.
The idea behind the Factory Method pattern is to be able to decide which class you want to instantiate dynamically (i.e. at runtime). Basically you will have a method that will return one of several possible classes that share a common superclass. The factory method has this name, because it’s responsible for “manufacturing” an object. This pattern allows you to encapsulate object creation in a method: the factory method.
The typical implementation uses a single class with a single method (the factory method) and this method returns an object based on the input passed as an argument. Which object is it? Well, that depends on the passed parameter.
Let’s say that you have a Fruit superclass that has two subclasses: Apple and Orange. You want your application to be able to instantiate each one of the subclasses dynamically, because you don’t know which fruit you’ll need (it might depend on the user input, for example).
All you’ll have to do is create a FruitFactory class and add a makeFruit() method to it, which accepts, for example a string and returns a Fruit (note it returns the generic Fruit object, not a concrete Apple or Orange).
To simplify things, let’s say that to create an apple you pass the “Apple” string as an argument and to create an orange you use the “Orange” string.
Below is a possible solution:
The Main class above produces the following output:
The fruit is an Apple. The fruit is an Orange.
The Strategy pattern lets you define a family of algorithms, encapsulates each one, and makes them interchangable. It lets the algorithms vary independently from clients that use it.
This pattern is very useful in cases when you have many hierarchically related classes that differ only in their behaviour. Strategy allows you to confugure every induvidual class in the hierarchy with one of the many behaviours. Another use case for it is when you have many variants of an algorithm.
Let’s say you have the class structure below:
Now you are asked to create a Bird class, which is also a subclass of the Animal class. The thing about the Bird class, is that it needs to be able to fly. You must also keep a common interface for all of the three subclasses, that means you can’t just add a fly() method to the Bird class, from now on every animal must print “Can’t fly!” or “Flying!” when the fly() method is invoked on an Animal object.
So how can we solve this? Well, we surely can just add a “fly()” method to the abstract class (or implement an interface) and override its behaviour in every subclass. But this creates several problems:
- code duplication.
- the change in super class will break the code in subclasses.
- implementing an inteface wich only has one method in it is usually a bad approach.
So what can we do to avoid those problems? Well, since all of our classes only differ in one behaviour (some fly and some don’t) it’s the right time to use the Strategy pattern!
What we’ll do instead is instead of inheriting the ability to fly, we will compose the class with objects which have the correct ability built-in (in out case we will only add one object).
This approach will give us many benefits, for example:
- it’s possible to create many types of “flying” without affecting the Animal class or any of its subclasses (since the “types of flying” are separate classes).
- we are decoupling, that means we are encapsulating the behavior that varies(in our case the behaviour is the ability to fly).
- it’s possible to change the ability at runtime. That means we can, for example, create an object that couldn’t fly before and make it flyable!
What we’ll do is create an interface called Fly, it will only have one method: fly() (which is the behaviour we want to encapsulate) and then for every “type of flying” we will create a subclass, each one implementing the Fly interface (in our case it will be two classes: CanFly and CantFly). We’ll also have to add a new varibale to the Animal class of type Fly.
Anyways, here is a possible soluition using the Strategy pattern, with an example of a client application that uses it:
The main class produces the following output:
Mittens: Can't fly! Rufus: Can't fly! Tweetie: Flying! Rufus: Flying!
The State pattern allows to modify its behavior when its internal state changes. The object will also appear to change its class. It actually mimics the finite state machine: you have different states and certain actions make you transition from one state to another.
This pattern encapsulates every state into a separate class (all those states derive from a common class). If you have an object and you completely change its behavior, it appears that the object changed it class (in reality you will just be using composition to give the appearance of class change by referencing different state objects).
The State pattern is a great way to get rid of if statements, but it usually requires a large amount of extra classes to be created.
Each state has a reference to the object whose state it represents and is able to change it dynamically (usually through a setter method).
Too keep it simple, consider the following problem: You have an automatic door with two buttons: open and close. You have four states:
The transitions between the states are intuitive: for example, if a door is closed, by pressing the open button, the door will go to the “open” state. To simplify, you can’t interrupt an opening/closing action(there are no intermediate DoorOpeningState and DoorClosingState classes).
What you’ll do is create an abstract DoorState class, with two operations:
Then, you will create two subclasses: DoorOpenState, DoorClosedState. Each one of the states will have a reference to the Door object (passed as an argument in the constructor) whose state it represents and it will be able to change its state through the setStateMethod() which will be a part of the Doors interface.
In this example, the Door class will contain references to every possible state which will be instantiated when a new Door object is created.
Below is a possible solution to the problem.
The output of the application above is:
Opening door... New state: DoorOpen. Closing door... New state: DoorClosedState. Door is already closed!
The Abstract Factory pattern provides an interface for creating families of related or depended objects without specifying their concrete classes.
The idea if an abstract factory is to define a common interface for a group of factories, then use those factories to produce objects. It provides an abstract type for creating a family of products and subclasses of that abstract type define how those products are produced. To use a specific factory, you’ll have to instantiate it and pass into some code that is written against the abstract type.
Factory Method vs Abstract Factory
It’s important to understand the difference between the Abstract Factory and the Factory Method patterns. While both are really good at decoupling applications from specific implementations, they do it in different ways. Both of the patterns create objects, but they do it in a different way:
- Factory Method uses classes - the objects are created through inheritance.
- Abstract Factory uses objects - the objects are created through object composition.
Often the Abstract Factory uses Factory Methods to implement its concrete factories. The concrete factories use a factory method to create their products (they are used purely to create products).
Consider the example from the Factory Method section. Well, here we also have fruit, except that one of the farmers has discovered a new sort of fruit: the Advanced Fruit. What distinguishes a “normal” Fruit from an AdvancedFruit? Well, unlike the “normal” Fruit, all AdvancedFruit have guns and engines!
The farmer in question is only producing AdvancedApple and AdvancedOrange types of AdvancedFruit, so those are the ones that will be covered in this example.
What distinguished an AdvancedApple from an AdvancedOrange? Well, while AdvancedApple has a BlueGun and a V16Engine, the AdvancedOrange has a RedGun and a V8Engine.
Hmmm… Isn’t that a good place to use the Abstact Factory pattern to abstract out the specific gun and engine creation? Note, that both, advanced apples and oranges have guns and engines, except that they are of different types. We sure can come up with a common abstract interface to produce those pieces of equipment and then have two specific implementations of those: one for the advanced apples and one for the advanced oranges.
When defining an abstract interface for out factories we have to question what makes an AdvancedFruit an AdvancedFruit? Well, it’s the fact that every AdvancedFruit has an Engine and a Gun.
Out abstract interface only will have to methods: makeEngine() and makeGun() (we’ll call it the AdvancedFruitFactory). Then in AdvancedAppleFactory and AdvancedOrangeFactory we’ll make those methods return the correct Engine and Gun objects.
Now every advanced fruit will receive a concrete factory in its constructor, because it’s through a method call on the AdvancedFuit object, that the specific engine and gun will be “given” to it.
Below is a possible solution to the problem:
The output of the Main class is as follows:
Name: Generic AdvancedApple Gun: Blue Gun Engine: V16 Engine Name: Generic AdvancedOrange Gun: Red Gun Engine: V8 Engine
The Template Method defines the skeleton of an algorithm in a method, deferring some steps to the subclasses. This pattern allows the subclasses to change certain parts of the algorithm without changing the algorithm’s structure.
So let’s say you have a generic algorithm, for example an algorithm for making tea. The general(abstract) idea is simple:
- Boil water.
- Put tea bag in cup.
- Pour in the boiled water(still hot).
- Add condiments.
The general algorithm is simple, however it can vary. Some prefer tea bags, while others prefer to use loose leafs. Some like to pour in water at 100ªC, while others prefer it a bit cooler. Some like to add condiments and some don’t.
So there are variations to the algorithm, but in essence, the algorithm doesn’t change.
That’s where the Template Method comes in. Basically you define an abstract template method, and then let the subclasses customize parts of it(or even the whole algorithm). This customization is obtained using method overriding and the so called hooks (mehods that return True/False).
Consider two types of cars: SportsCar and CityCar. The generic algorithm for car creation is the following:
- Add chassis.
- Add body.
- Add wheels.
- Add windows.
- Add air conditioning.
- Add radio.
That’s the general algorithm that all cars follow, the SportsCar however, doesn’t have air conditioning or radio and it uses a different type of wheels (sports wheels), while the CityCar uses regular wheels has A/C and radio. All of the cars must have a chassis, a body and windows.
A possible solution using the Template Method can be found right below:
The output of the client application is:
Sports Car Production Begin! --- Adding chassis... Adding body... Adding sports wheels... Adding windows... Sports Car Production End! --- Sports Car Production Begin! --- Adding chassis... Adding body... Adding regular wheels... Adding windows... Adding A/C... Adding radio... City Car Production End! ---
- The template method is declared final because we don’t want the subclasses to be changing the algorithm.
- Have some “hooks” that return booleans, to decide whether we should run a certain method or not.
- Make a method abstract when you want to force the user to override the method.
- Create a hook when you want to make a part of your algorithm totally optional.
The Observer pattern defines a one-to-many dependency between objects so that when one object changes state all of its dependents get notified and updated automatically.
A good analogy is a newspaper. A newspaper (the Subject) can have zero or more subscribers (Observers) and every time a new edition comes out (the newspaper, i.e. the Subject changes), the subscribers receive the new edition in the mail (i.e. the Observers get notified and updated about that change.
Consider you have a Subject: a Facebook user. Anyone can subscribe to it and they receive updates every time the Subject updates its status or “likes” something.
Of course you could put each “observer” in an infinite loop testing whether there has been a new status update or a new like, but that would be extremely uneficient! The Template pattern provides a much more elegant and resourse-friendly solution.
Below is a possible solution to the problem:
The output of the application is as follows:
Status: Second status update!|**| Last Liked: FistLike Status: Second status update!|**| Last Liked: SecondLike Status: Third status update!|**| Last Liked: SecondLike Status: Third status update!|**| Last Liked: SecondLike Status: Third status update!|**| Last Liked: SecondLike Status: Third status update!|**| Last Liked: ThirdLike Status: Third status update!|**| Last Liked: ThirdLike Status: Third status update!|**| Last Liked: ThirdLike Status: Forth status update!|**| Last Liked: ThirdLike Status: Forth status update!|**| Last Liked: ThirdLike Status: Forth status update!|**| Last Liked: ForthLike Status: Forth status update!|**| Last Liked: ForthLike
The Decorator design pattern allows you to attach additional responsibilities to an object dynamically. It provides a flexible alternative to subclassing for extending functionality.
This is usually a great choice of a pattern whenever you want to be able to add responsibilities to individual objects dynamically (at runtime) without affecting other objects. The responsibilities you add can be withdrawn later on, also dynamically. Sometimes extension by subclassing is simply impractical: you you might have a large amount of independent extentions and that would produce an explosion of subclasses and every time you added a new extentsion, you would have to create another subclass, which only aggravates the problem. The pattern gives you the capabilities of inheritance, but with functionality added at runtime.
Let’s say a car dealership is selling a certain model of a car. It’s base price (for model with no extras) is $100.000, but you can also add extras like AirConditioning(extra $5.000), Spoiler(extra $3.000) and a custom BodyKit (extra $15.000).
One way to solve this would to create a subclass for every possible combination, but that would give you 7 subclasses just for the cars(3! + 1 = 7)! And what if you wanted to add another extra? You can understand the way this is heading…
This seems like the right place to use the Decorator pattern. We want to add responsibilities dynamically.
First, we’ll create the Car interface, which is the common interface for every car and every extra. Then, we’ll create a BasiCar model, which represents the most basic car, which then will be decorated by extras. All of the extras derive from the CarDecorator, which, just as the BasicCar, implements the Car interface. The decorator will store a reference to the object it decorates. In the decorator’s superclass we want to implement the interface of the object that that group of decorators will be decorating.
Here is a possible way to solve this exercise:
The application outputs:
Basic Car Model |**| Price : 100000 Basic Car Model + Body Kit |**| Price : 115000 Basic Car Model + Body Kit + A/C |**| Price : 120000 Basic Car Model + Body Kit + A/C + Spoiler |**| Price : 123000
Note, that on second and third “decorations” you are actually passing a CarDecorator and not a BasicCar object, so when you call getDescription() on one, it calls the getDescription() on another, until a getDescription() reaches a plain return of a string (here it happens in the BasicCar class). This is why it’s both, the decorator and the decorated (the car) must implement a common interface.
The Adapter design pattern converts the interface of a class into another interface that the client expects. Adapter lets classes work together that couldn’t otherwise because they have incompatible interfaces.
This is usually the right pattern to go with when you want to use an existing class without modifying it, but its interface doesn’t match the one that you need.
Adapters in OO are just like adapters in real life. Have you ever needed to use an electronic device with an US AC plug in Europe? Well, since the European wall outlet (the client) expects a different “shape” of adapter what do you do? Change the wall outlet? Of course not! You use an adapter to adapt the US AC Plug into a Europe AC plug, as shown in the image below (image from www.safaribooksonline.com).
(This is kind of off-topic, but not all European wall outlets looks like that, in fact I believe I’ve never encountered one that’s of the same shape as in the image, but they do exist.)
Let’s say that you have a duck simulator application. What does it do? Well, it simulates ducks by invoking quack() and fly() methods on various ducks.
Everything has been going good, until your users demanded that you include turkeys from another popular simulator “Turkey Simulator” straight into your game. The developers of “Turckey Simulator” are willing to share they turkey classes from they gigantic turkey database with you. That’s great, except for one thing: all of the turkey classes have a different interface: instead of the quack() method, they have gobble() and despite the fly() method having the same name, they don’t really fly the same way as the ducks do. You are certainly not going to change every class provided by the “Turkey Simulator” team, since that would take too much time and you would have to do that to every new turkey added, besides that, to ensure the quality of the turkeys used, they gave you the permission to use their code, but not modify it.
To solve our problem, we will create a new TurkeyAdapter class, which will store a reference to the turkey that it is adapting into a duck as well as implement the same interface as the Duck class is using.
Below is a possible solution using the pattern:
The application outputs the following:
Duck --- Flying! Flying! --- Turkey --- Flying a short distance. Flying a short distance. Flying a short distance. Flying a short distance. Flying a short distance. Gobble! ---
There are two kind of adapters: class adapter and object adapter. This text only covered the object adapters. Class adapters aren’t possible in Java, since they require multiple inheritance.
Object Adapter vs Class Adapter
The only difference between a class and an object adapter is that in the in class adapters, the Target and the Adaptee are subclassed to create the Adapter. In object adapters object composition is used to pass the requests to an adaptee. Below is the class adapter UML.
The Facade pattern provides a unified interface to a set of interfaces in a subsystem. Facade defines a higher-level interface that makes the subsystem easier to use.
This pattern also allows you to avoid tight coupling between clients and subsystems, as well as provides a simple default view of the subsystem that is good enough for most clients. Only the clients that need more customizability will look beyond the pattern.
So in other words, the pattern allows you to take a complex subsystem and make it easier to use by implementing a Facade class, that provides one, more simple interface. It not only simplifies an interface, as well as decouples a client from a subsystem of components.
As an example consider that you have a home theater system installed at your place. Suppose it consists of a DVD player, a projector, a screen, a sound system and lighs. Each one of the components is defined in it’s own class:
Okay, now let’s say you want to watch a movie. In order to do that, you have to perform a few tasks:
- Turn the screen on
- Turn the projector on
- Turn the DVD player on
- Set the sound to
- Set the light’s intensity to
- Start the DVD playback
The code to do that without using the pattern would look something like this:
Here are some issues with that approach:
- It’s complex
- Let’s say you want to reset everything after the movie is over, wouldn’t you have to do everything again, but in reverse order?
- If you decide to change your system, and let’s say put your phone on silence when the movie starts, as well as
set the volume to
5, you’re going to have to change your code in every place you’ve used it
- Wouldn’t it be as complex to play a music CD or turn on the radio?
So what can we do here? The complexity of the current approach is evident. Facade pattern to the rescue!
What we’re going to do is create a Facade for our home theater system. To do this, we’ll create a
class which exposes a few methods such as
What are going be the contents of those methods? Well, it’s going to be all that complex code that was mentioned above. Here is a possible implementation:
Now to watch a movie all we have to do is call the
watchMovie() method on the
The code is now a lot simpler and if we at some point decide to change the steps in any of the methods, the
only place where we need to modify the code is in the
HomeTheaterFacade class, since you decoupled your
client implementation from any one subsystem.
Here is how the application that uses
HomeTheaterFacade would look like:
It produces the following output:
Get ready for the movie... Turning Screen On... Turning Projector On... Turning DVD On... Volume set to: 5 Intensity set to: 1 Starting movie playback... Ending movie playback... Stopping movie playback... Intensity set to: 5 Volume set to: 0 Turning DVD Off... Turning Projector On... Turning Screen Off...
Facade vs Adapter
Facades and adapters may wrap multiple classes, but a facade’s intent is to simplify, while an adapter’s intent is to convert one interface to another.
This post was heavily based on the Head First Design Patterns book. It provides an easy to read and undestand introduction to the topic. I would suggest you reading it if you’re new to design patterns. It’s also a good book if you need to quickly refreshen some of the concepts.
Another great book, probably the most well-known one is Design patterns : elements of reusable object-oriented software. It’s more complex, but it explores real world examples.