Behavioural software patterns


These patterns cover algorithms and the assignment of responsibilities between objects.

Iterator – this pattern is concerned with allowing iterative access to a composition without exposing the underlying implementation.

Typically an iterator is an abstract interface and the composite object creates a concrete iterator that implements it – but the iterator interface is what is used by other objects.

Mediator – the mediator pattern is concerned with ensuring loose coupling between objects by defining interfaces that allow them to interact without knowledge of each others’ internals.

Observer – this pattern defines a one-many dependency in which one object – called the subject – notifies subscribing objects of changes in its state.

UML to describe the Observer pattern
From Wikipedia

State – this pattern is about allowing an object to change its behaviour when its state changes. The object will appear to change its class.

The Gang of Four give the example of a TCP connection – which will respond differently to datagrams depending if it is in the LISTEN, ESTABLISHED or CLOSED state. In each of these cases a different subclass will handle behaviour, though a common interface is used by all three.

UML for state pattern
From Wikipedia

Strategy – this pattern allows objects with different algorithms to operate through a common interface. Typical examples in text books and lecture notes are including paper aeroplanes and rubber ducks in collections of aircraft and avians!

So, for instance, all aircraft may have a fly() method in their abstract interface, but the concrete implementation for the paper aeroplane implements behaviour appropriate for that type of craft.

UML for strategy pattern
From wikipedia

Template Method  – this pattern creates a skeleton for an algorithm but allows subclasses to over-ride specific parts.

Template method UML Class Diagram
Image via Wikipedia

Visitor – the visitor software pattern is concerned with allowing new algorithms to be applied to an object without changing the object itself. Instead a new visitor class implements the visitor interface. In this way many new methods can be dynamically added to a class.

Class Diagram for the Visitor Pattern
Image via Wikipedia

Structural software patterns


After a short digression into the prospect of an asteroid ending human civilisation and so making my Object Orientated Design and Programming exam on Tuesday something of a waste of effort, I am back to using the blog as a notepad and revision resource.

From the Gang of Four’s Design Patterns , I want to look at some of the structural software patterns: these are the patterns that cover issues such as composition.

The Adapter or Wrapper pattern allows classes that could not otherwise work together, work together, by converting the interface of one class into that which is expected.

The GoF use the example of integrating an “off the shelf” text editing component into a graphics package. The text editor may not have the interface that matches the graphics package, but an adapter can allow the two to work together.

A UML diagram of the Object Adapter design pat...
Image via Wikipedia

From wikipedia

Hopefully the bit of UML above makes it clearer – the client class calls the adapter in its doWork() method, and it in turn calls the correct method in the adaptee. This is the “object adapter” pattern, but there also exists a “class adapter” pattern where the adapter inherits the interfaces of both the “expected” interface and the adaptee.

Composite – this pattern allows us to build structures of objects in the form of trees that contain both the objects and compositions of objects. We can apply the same operations over both the composites and the individual objects – the example given in Head First Design Patterns is of a print() operation – we can apply this to any node in the tree to print it in whole or in part.

UML for composite pattern
From wikipedia

Decorator – this pattern allows new responsibilities to be added to an object dynamically – it is “decorated” with the additional responsibility. Decorator class methods must have the same signature of the decorated component but may have different behaviour.

UML for Decorator pattern
From wikipedia

Facade – this pattern provides a unified interface to a set of interfaces in a subsystem. The Gang of Four give the example of a compiler – which may have a simple compile() interface method as opposed to exposing the complex series of commands needed to compile some code.

Proxy – this pattern provides an object as a surrogate for another object to allow control over access to the proxied object. The example the gang of four use is for a graphical object in a document, but these days a webpage may be a better example – for instance I have set the browser on one of my machines not to automatically load Flash. Instead I get a component I can click on to load the Flash if I wish (this means I am not flooded with the advertisers’ web bloat) – sound pretty much like a proxy pattern to me!

UML for proxy pattern
From WikipediaFrom Wikipedia

Bridge  – this pattern places the implementation and the abstraction in two separate hierarchies and allows the two to vary independently.

Bridge UML
From Wikipedia


Creational patterns


The Gang of Four list the following creational patterns:

Abstract Factory – an interface for creating families of related or dependent objects without specifying their concrete classes – the example the Gang of Four give is of an abstract factory that produces “abstract widgets” (ie display artefacts) that can then be subclassed to the particular look and feel of the display (the book shows its age by referring to Motif and Presentation Manager as choices, but the principle is unaltered). This pattern can be used when a system should be independent of how its products are created and represented, or when a library of software objects is needed but one wishes only to expose the interface and not the implementation.

Abstract Factories have a certain inflexibility though – to add a new output object it is necessary to change the interface and that also means changing all the concrete subclasses, as they have to match the interface of their abstract parent.

Builder – this separates the construction of a complex object from its representation, so allowing one construction process to create many different representations.

Builder software pattern
From wikipedia

Factory – defines an interface for creating an object but lets subclasses decide which class to instantiate.

Factory pattern in UML
From wikipedia

Prototype creates a prototypical instance and new objects are instantiated by copying this prototype.

The Gang of Four give an example of a musical score editor, which might create graphic elements to add to the score – but instead of the editor needing to have a subclass for every type of musical object to be added, it calls a parameterised clone method on a graphic (abstract) class that creates a copy of the concrete element sought.

UML for prototype pattern
From wikipedia

Singleton – a pattern to ensure that there is only one instance of a given class and that there is a universal point of access to it.

Often implemented using static methods in the real world:

class Spooler {
private spooler = 0
static Spooler getSpooler() {
if (spooler == 0)
spooler = new Spooler()
return spooler
}
protected Spooler() {
…some initialisation code…..
}
}

GRASP software patterns


This is a UML diagram (created in IBM Rhapsody...
Image via Wikipedia

My final exam is on Tuesday and covers object-orientated design and programming – hence the Groovy Game of Life code.

Software patterns are a big part of the course and part of that is the simple task of remembering what each pattern does. I have found Head First Design Patterns useful but it covers few patterns and certainly not the basic GRASP ones, that I do have to know.

GRASP stands for General Responsibility Assignment Software Patterns (or Principles – I actually think this comes closer to expressing what this is all about anyway).

These are:

Information Expert, solving the problem of the general principle of assigning responsibilities to objects by assigning the responsibilities to the information expert – the class that has the information necessary to fulfil the responsibility.

Creator, solving the problem of what should be responsible for creating a new instance of some class by assigning the responsibility of creating a new instance of class A  to the class, B, that:

  • aggregates A objects; or
  • contains A objects; or
  • records instances of A objects; or
  • closely uses A objects; or
  • has the initialising data that will be passed to A when it is created.

(If more than one class displays these characteristics then aggregation or composition should be preferred.)

Low Coupling, solving the problem of creating classes that have low levels of dependency and face a low impact when other classes are changed and can be more easily reused by assigning responsibilities so that coupling remains low.

Controller, which addresses the question of what should be responsible for handling an input system event by assigning the responsibility to a class that represents one of the following choices:

  • It represents the overall system, device or subsystem (it is a facade controller); or
  • It represents a use case scenario within which the system event occurs, often named Handler or Coordinator, etc (it is a use-case or session controller).

High Cohesion, which addresses the problem of how to keep complexity manageable and how to keep objects focused by assigning responsibilities so that cohesion remains high.

%d bloggers like this: