The ultimate guide to the most useful design patterns
UPDATE NOTE: Updated the Proxy Pattern example to use ES6 Proxy and Reflect. Replaced images of source code snippets with GitHub gists.
This article is a long read as a whole because of the nature of the subject matter, so I have tried to keep the sections self-contained. So you as a reader can pick and choose specific parts (or, in this case, specific patterns) and ignore the ones you are not interested in or are well versed with. Now, let?s get started.
Note: Source code for the implementation of all the design patterns explained here is on GitHub.
We write code to solve problems. These problems usually have many similarities, and, when trying to solve them, we notice several common patterns. This is where design patterns come in.
A design pattern is a term used in software engineering for a general, reusable solution to a commonly occurring problem in software design.
The underlying concept of design patterns has been around in the software engineering industry since the very beginning, but they weren?t really so formalised. Design Patterns: Elements Of Reusable Object-Oriented Software written by Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides ? the famous Gang of Four (GoF)?was instrumental in pushing the formalised concept of design patterns in software engineering. Now, design patterns are an essential part of software development and have been so for a long time.
There were 23 design patterns introduced in the original book.
The Classic 23 Patterns introduced by GoF
Design patterns are beneficial for various reasons. They are proven solutions that industry veterans have tried and tested. They are solid approaches that solve issues in a widely accepted way and reflect the experience and insights of the industry-leading developers that helped define them. Patterns also make your code more reusable and readable while speeding up the development process vastly.
Design patterns are by no means finished solutions. They only provide us with approaches or schemes to solve a problem.
Categories of Design Patterns
Design patterns are usually categorized into three major groups.
Creational Design Patterns
As the name suggests, these patterns are for handling object creational mechanisms. A creational design pattern basically solves a problem by controlling the creation process of an object.
We will discuss the following patterns in detail: Constructor Pattern, Factory Pattern, Prototype Pattern, and Singleton Pattern.
Structural Design Patterns
These patterns are concerned with class and object composition. They help structure or restructure one or more parts without affecting the entire system. In other words, they help obtain new functionalities without tampering with the existing ones.
We will discuss the following patterns in detail: Adapter Pattern, Composite Pattern, Decorator Pattern, Faade Pattern, Flyweight Pattern, and Proxy Pattern.
Behavioral Design Patterns
These patterns are concerned with improving communication between dissimilar objects.
We will discuss the following patterns in detail: Chain of Responsibility Pattern, Command Pattern, Iterator Pattern, Mediator Pattern, Observer Pattern, State Pattern, Strategy Pattern, and Template Pattern.
This is a class-based creational design pattern. Constructors are special functions that can be used to instantiate new objects with methods and properties defined by that function.
In this example, we define a Hero class with attributes like name and specialAbility and methods like getDetails. Then, we instantiate an object IronMan by invoking the constructor method with the new keyword passing in the values for the respective attributes as arguments.
Factory pattern is another class-based creational pattern. In this, we provide a generic interface that delegates the responsibility of object instantiation to its subclasses.
This pattern is frequently used when we need to manage or manipulate collections of objects that are different yet have many similar characteristics.
In this example, we create a factory class named BallFactory that has a method that takes in parameters, and, depending on the parameters, it delegates the object instantiation responsibility to the respective class. If the type parameter is “football” or “soccer” object instantiation is handled by Football class, but if it is “basketball” object instantiation is handled by Basketball class.
This pattern is an object-based creational design pattern. In this, we use a sort of a ?skeleton? of an existing object to create or instantiate new objects.
Singleton is a special creational design pattern in which only one instance of a class can exist. It works like this ? if no instance of the singleton class exists then a new instance is created and returned, but if an instance already exists, then the reference to the existing instance is returned.
A perfect real-life example would be that of mongoose (the famous Node.js ODM library for MongoDB). It utilizes the singleton pattern.
In this example, we have a Database class that is a singleton. First, we create an object mongo by using the new operator to invoke the Database class constructor. This time an object is instantiated because none already exists. The second time, when we create the mysql object, no new object is instantiated but instead, the reference to the object that was instantiated earlier, i.e. the mongo object, is returned.
This is a structural pattern where the interface of one class is translated into another. This pattern lets classes work together that could not otherwise because of incompatible interfaces.
This pattern is often used to create wrappers for new refactored APIs so that other existing old APIs can still work with them. This is usually done when new implementations or code refactoring (done for reasons like performance gains) result in a different public API, while the other parts of the system are still using the old API and need to be adapted to work together.
In this example, we have an old API, i.e. OldCalculator class, and a new API, i.e. NewCalculator class. The OldCalculator class provides an operationmethod for both addition and subtraction, while the NewCalculator provides separate methods for addition and subtraction. The Adapter class CalcAdapterwraps the NewCalculator to add the operation method to the public-facing API while using its own addition and subtraction implementation under the hood.
This is a structural design pattern that composes objects into tree-like structures to represent whole-part hierarchies. In this pattern, each node in the tree-like structure can be either an individual object or a composed collection of objects. Regardless, each node is treated uniformly.
A Multi-level Menu Structure
It is a bit complex to visualize this pattern. The easiest way to think about this is with the example of a multi-level menu. Each node can be a distinct option, or it can be a menu itself, which has multiple options as its child. A node component with children is a composite component, while a node component without any child is a leaf component.
In this example, we create a base class of Component that implements the common functionalities needed and abstracts the other methods needed. The base class also has a static method that utilises recursion to traverse a composite tree structure made with its subclasses. Then we create two subclasses extending the base class ? Leaf that does not have any children and Composite that can have children?and hence have methods handling adding, searching, and removing child functionalities. The two subclasses are used to create a composite structure?a tree, in this case.
This is also a structural design pattern that focuses on the ability to add behaviour or functionalities to existing classes dynamically. It is another viable alternative to sub-classing.
If you want to read about the proposal itself, feel free.
In this example, we create a Book class. We further create two decorator functions that accept a book object and return a ?decorated? book object ? giftWrap that adds one new attribute and one new function and hardbindBook that adds one new attribute and edits the value of one existing attribute.
The use of this pattern is very common in libraries like jQuery.
In this example, we create a public facing API with the class ComplaintRegistry. It exposes only one method to be used by the client, i.e. registerComplaint. It internally handles instantiating required objects of either ProductComplaint or ServiceComplaint based on the type argument. It also handles all the other complex functionalities like generating a unique ID, storing the complaint in memory, etc. But, all these complexities are hidden away using the faade pattern.
This is a structural design pattern focused on efficient data sharing through fine-grained objects. It is used for efficiency and memory conservation purposes.
This pattern can be used for any kind of caching purposes. In fact, modern browsers use a variant of a flyweight pattern to prevent loading the same images twice.
In this example, we create a fine-grained flyweight class Icecream for sharing data regarding ice-cream flavours and a factory class IcecreamFactory to create those flyweight objects. For memory conservation, the objects are recycled if the same object is instantiated twice. This is a simple example of flyweight implementation.
This is a structural design pattern that behaves exactly as its name suggests. It acts as a surrogate or placeholder for another object to control access to it.
It is usually used in situations in which a target object is under constraints and may not be able to handle all its responsibilities efficiently. A proxy, in this case, usually provides the same interface to the client and adds a level of indirection to support controlled access to the target object to avoid undue pressure on it.
The proxy pattern can be very useful when working with network request-heavy applications to avoid unnecessary or redundant network requests.
Now, we create a function that can be thought of as a network request. We named it as networkFetch. It accepts a URL and responds accordingly. We want to implement a proxy where we only get the response from the network if it is not available in our cache. Otherwise, we just return a response from the cache.
The cache global variable will store our cached responses. We create a proxy named proxiedNetworkFetch with our original networkFetch as the targetand use apply method in our handler object to proxy the function invocation. The apply method gets passed on the target object itself. This value as thisArg and the arguments are passed to it in an array-like structure args.
We check if the passed url argument is in the cache. If it exists in the cache, we return the response from there, never invoking the original target function. If it does not, then we use the Reflect.apply method to invoke the targetfunction with thisArg (although it?s not of any significance in our case here) and the arguments it passed.
Chain of Responsibility Pattern
This is a behavioural design pattern that provides a chain of loosely coupled objects. Each of these objects can choose to act on or handle the request of the client.
A good example of the chain of responsibility pattern is the event bubbling in DOM in which an event propagates through a series of nested DOM elements, one of which may have an ?event listener? attached to listen to and act on the event.
In this example, we create a class CumulativeSum, which can be instantiated with an optional initialValue. It has a method add that adds the passed value to the sum attribute of the object and returns the object itself to allow chaining of add method calls.
This is a common pattern that can be seen in jQuery as well, where almost any method call on a jQuery object returns a jQuery object so that method calls can be chained together.
Chain of Responsibility Pattern
This is a behavioural design pattern that aims to encapsulate actions or operations as objects. This pattern allows loose coupling of systems and classes by separating the objects that request an operation or invoke a method from the ones that execute or process the actual implementation.
The clipboard interaction API somewhat resembles the command pattern. If you are a Redux user, you have already come across the command pattern. The actions that allow the awesome time-travel debugging feature are nothing but encapsulated operations that can be tracked to redo or undo operations. Hence, time-travelling made possible.
In this example, we have a class called SpecialMath that has multiple methods and a Command class that encapsulates commands that are to be executed on its subject, i.e. an object of the SpecialMath class. The Command class also keeps track of all the commands executed, which can be used to extend its functionality to include undo and redo type operations.
It is a behavioural design pattern that provides a way to access the elements of an aggregate object sequentially without exposing its underlying representation.
Iterators have a special kind of behaviour where we step through an ordered set of values one at a time by calling next() until we reach the end. The introduction of Iterator and Generators in ES6 made the implementation of the iterator pattern extremely straightforward.
We have two examples below. First, one IteratorClass uses iterator spec, while the other one iteratorUsingGenerator uses generator functions.
The Symbol.iterator ( Symbol?a new kind of primitive data type) is used to specify the default iterator for an object. It must be defined for a collection to be able to use the for…of looping construct. In the first example, we define the constructor to store some collection of data and then define Symbol.iterator, which returns an object with next method for iteration.
For the second case, we define a generator function passing it an array of data and returning its elements iteratively using next and yield. A generator function is a special type of function that works as a factory for iterators and can explicitly maintain its own internal state and yield values iteratively. It can pause and resume its own execution cycle.
It is a behavioural design pattern that encapsulates how a set of objects interact with each other. It provides the central authority over a group of objects by promoting loose coupling, keeping objects from referring to each other explicitly.
In this example, we have TrafficTower as Mediator that controls the way Airplane objects interact with each other. All the Airplane objects register themselves with a TrafficTower object, and it is the mediator class object that handles how an Airplane object receives coordinates data of all the other Airplane objects.
It is a crucial behavioural design pattern that defines one-to-many dependencies between objects so that when one object (publisher) changes its state, all the other dependent objects (subscribers) are notified and updated automatically. This is also called PubSub (publisher/subscribers) or event dispatcher/listeners pattern. The publisher is sometimes called the subject, and the subscribers are sometimes called observers.
Chances are, you?re already somewhat familiar with this pattern if you have used addEventListener or jQuery?s .on to write even-handling code. It has its influences in Reactive Programming (think RxJS) as well.
In the example, we create a simple Subject class that has methods to add and remove objects of Observer class from subscriber collection. Also, a firemethod to propagate any changes in the Subject class object to the subscribed Observers. The Observer class, on the other hand, has its internal state and a method to update its internal state based on the change propagated from the Subject it has subscribed to.
It is a behavioural design pattern that allows an object to alter its behaviour based on changes to its internal state. The object returned by a state pattern class seems to change its class. It provides state-specific logic to a limited set of objects in which each object type represents a particular state.
We will take a simple example of a traffic light to understand this pattern. The TrafficLight class changes the object it returns based on its internal state, which is an object of Red, Yellow, or Green class.
It is a behavioural design pattern that allows encapsulation of alternative algorithms for a particular task. It defines a family of algorithms and encapsulates them in such a way that they are interchangeable at runtime without client interference or knowledge.
In the example below, we create a class Commute for encapsulating all the possible strategies for commuting to work. Then, we define three strategies namely Bus, PersonalCar, and Taxi. Using this pattern we can swap the implementation to use for the travel method of the Commute object at runtime.
This is a behavioural design pattern based on defining the skeleton of the algorithm or implementation of an operation, but deferring some steps to subclasses. It lets subclasses redefine certain steps of an algorithm without changing the algorithm?s outward structure.
In this example, we have a Template class Employee that implements workmethod partially. It is for the subclasses to implement responsibilities method to make it work as a whole. We then create two subclasses Developer and Tester that extend the template class and implement the required method to fill the implementation gap.
- Design Patterns: Elements Of Reusable Object-Oriented Software by Erich Gamma, Richard Helm, Ralph Johnson and John Vlissides (Gang of Four)