Software Engineering for Self-Directed Learners »

UML

Class Diagrams

Introduction

What

UML class diagrams describe the structure (but not the behavior) of an OOP solution. These are possibly the most often used diagrams in the industry and are an indispensable tool for an OO programmer.

An example class diagram:

Classes

What

The basic UML notations used to represent a class:

A Table class shown in UML notation:

The equivalent code


The 'Operations' compartment and/or the 'Attributes' compartment may be omitted if such details are not important for the task at hand. Similarly, some attributes/operations can be omitted if not relevant. 'Attributes' always appear above the 'Operations' compartment. All operations should be in one compartment rather than each operation in a separate compartment. Same goes for attributes.

The visibility of attributes and operations is used to indicate the level of access allowed for each attribute or operation. The types of visibility and their exact meanings depend on the programming language used. Here are some common visibilities and how they are indicated in a class diagram:

  • + : public
  • - : private
  • # : protected
  • ~ : package private

How visibilities map to programming language features


Table class with visibilities shown:

The equivalent code


Generic classes can be shown as given below. The notation format is shown on the left, followed by two examples.

Associations

What

You should use a solid line to show an association between two classes.

This example shows an association between the Admin class and the Student class:

Use arrowheads to indicate the navigability of an association.

In this example, the navigability is unidirectional, and is from the Logic class to the Minefield class. That means if a Logic object L is associated with a Minefield object M, L has a reference to M but M doesn't have a reference to L.

class Logic {
    Minefield minefield;
    // ...
}

class Minefield {
    //...
}
class Logic:
  
  def __init__(self):
    self.minefield = None
    
  # ...


class Minefield:
  # ...

Here is an example of a bidirectional navigability; i.e., if a Dog object d is associated with a Man object m, d has a reference to m and m has a reference to d.

class Dog {
    Man man;
    // ...
}

class Man {
    Dog dog;
    // ...
}
class Dog:
  
  def __init__(self):
    self.man = None
    
  # ...


class Man:
  def __init__(self):
    self.dog = None

  # ...

Navigability can be shown in class diagrams as well as object diagrams.

According to this object diagram, the given Logic object is associated with and aware of two MineField objects.

Roles

Association Role are used to indicate the role played by the classes in the association.

This association represents a marriage between a Man object and a Woman object. The respective roles played by objects of these two classes are husband and wife.

Note how the variable names match closely with the association roles.

class Man {
    Woman wife;
}

class Woman {
    Man husband;
}
class Man:
  def __init__(self):
    self.wife = None # a Woman object

class Woman:
   def __init__(self):
     self.husband = None # a Man object

The role of Student objects in this association is charges (i.e. Admin is in charge of students)

class Admin {
    List<Student> charges;
}
class Admin:
  def __init__(self):
    self.charges = [] # list of Student objects

Association roles are optional to show. They are particularly useful for differentiating among multiple associations between the same two classes.

In each the three associations between the Flight class and the Airport class given below, the Airport class plays a different role.

Labels

Association labels describe the meaning of the association. The arrow head indicates the direction in which the label is to be read.

In this example, the same association is described using two different labels.

  • Diagram on the left: Admin class is associated with Student class because an Admin object uses a Student object.
  • Diagram on the right: Admin class is associated with Student class because a Student object is used by an Admin object.

Multiplicity

Commonly used multiplicities:

  • 0..1 : optional, can be linked to 0 or 1 objects.
  • 1 : compulsory, must be linked to one object at all times.
  • * : can be linked to 0 or more objects.
  • n..m : the number of linked objects must be within n to m inclusive.

In the diagram below, an Admin object administers (is in charge of) any number of students but a Student object must always be under the charge of exactly one Admin object.

In the diagram below,

  • Each student must be supervised by exactly one professor. i.e. There cannot be a student who doesn't have a supervisor or has multiple supervisors.
  • A professor cannot supervise more than 5 students but can have no students to supervise.
  • An admin can handle any number of professors and any number of students, including none.
  • A professor/student can be handled by any number of admins, including none.

Associations as Attributes

Associations as attributes

An association can be shown as an attribute instead of a line.

Association multiplicities and the default value can be shown as part of the attribute using the following notation. Both are optional.

name: type [multiplicity] = default value

The diagram below depicts a multi-player Square Game being played on a board comprising of 100 squares. Each of the squares may be occupied with any number of pieces, each belonging to a certain player.

A Piece may or may not be on a Square. Note how that association can be replaced by an isOn attribute of the Piece class. The isOn attribute can either be null or hold a reference to a Square object, matching the 0..1 multiplicity of the association it replaces. The default value is null.

The association that a Board has 100 Squares can be shown in either of these two ways:

Show each association as either an attribute or a line but not both. A line is preferred as it is easier to spot.

Diagram (a) given below shows the 'author' association between the Book class and the Person class as a line while (b) shows the same association as an attribute in the Book class. Both are correct and the two are equivalent. But (c) is not correct as it uses both a line and an attribute to show the same association.

(a)
(b)
(c)

Enumerations

What

Notation:

In the class diagram below, there are two enumerations in use:

Class-Level Members

Class-level members

In UML class diagrams, underlines denote class-level attributes and methods.

In the class diagram below, the totalStudents attribute and the getTotalStudents method are class-level.

Association Classes

Association classes

Association classes are denoted as a connection to an association link using a dashed line as shown below.

In this example Loan is an association class because it stores information about the borrows association between the User and the Book.

Composition

Composition

UML uses a solid diamond symbol to denote composition.

Notation:

A Book consists of Chapter objects. When the Book object is destroyed, its Chapter objects are destroyed too.

Aggregation

Aggregation

UML uses a hollow diamond to indicate an aggregation.

Notation:

Example:

Aggregation vs Composition

The distinction between composition (◆) and aggregation (◇) is rather blurred. Martin Fowler’s famous book UML Distilled advocates omitting the aggregation symbol altogether because using it adds more confusion than clarity.

Class Inheritance

Inheritance

You can use a triangle and a solid line (not to be confused with an arrow) to indicate class inheritance.

Notation:

Examples: The Car class inherits from the Vehicle class. The Cat and Dog classes inherit from the Pet class.

Interfaces

Interfaces

An interface is shown similar to a class with an additional keyword <<interface>>. When a class implements an interface, it is shown similar to class inheritance except a dashed line is used instead of a solid line.

The AcademicStaff and the AdminStaff classes implement the SalariedStaff interface.

Abstract Classes

What

You can use italics or {abstract} (preferred) keyword to denote abstract classes/methods.

Example:

Object Diagrams

Introduction

An object diagram shows an object structure at a given point of time.

An example object diagram:

Objects

Notation:

Notes:

  • The class name and object name e.g. car1:Car are underlined.
  • objectName:ClassName is meant to say 'an instance of ClassName identified as objectName'.
  • Unlike classes, there is no compartment for methods.
  • Attributes compartment can be omitted if it is not relevant to the task at hand.
  • Object name can be omitted too e.g. :Car which is meant to say 'an unnamed instance of a Car object'.

Some example objects:

Associations

A solid line indicates an association between two objects.

An example object diagram showing two associations:

Sequence Diagrams

Introduction

A UML sequence diagram captures the interactions between multiple objects for a given scenario.

Consider the code below.

class Machine {

    Unit producePrototype() {
        Unit prototype = new Unit();
        for (int i = 0; i < 5; i++) {
            prototype.stressTest();
        }
        return prototype;
    }
}

class Unit {

    public void stressTest() {

    }
}

Here is the sequence diagram to model the interactions for the method call producePrototype() on a Machine object.

Basic

Notation:

This sequence diagram shows some interactions between a human user and the Text UI of a Minesweeper game.

The player runs the newgame action on the TextUi object which results in the TextUi showing the minefield to the player. Then, the player runs the clear x y command; in response, the TextUi object shows the updated minefield.

The :TextUi in the above example denotes an unnamed instance of the class TextUi. If there were two instances of TextUi in the diagram, they can be distinguished by naming them e.g. TextUi1:TextUi and TextUi2:TextUi.

Arrows representing method calls should be solid arrows while those representing method returns should be dashed arrows.

Note that unlike in object diagrams, the class/object name is not underlined in sequence diagrams.

[Common notation error] Activation bar too long: The activation bar of a method cannot start before the method call arrives and a method cannot remain active after the method has returned. In the two sequence diagrams below, the one on the left commits this error because the activation bar starts before the method Foo#xyz() is called and remains active after the method returns.

[Common notation error] Broken activation bar: The activation bar should remain unbroken from the point the method is called until the method returns. In the two sequence diagrams below, the one on the left commits this error because the activation bar for the method Foo#abc() is not contiguous, but appears as two pieces instead.

Object creation

Notation:

  • The arrow that represents the constructor arrives at the side of the box representing the instance.
  • The activation bar represents the period the constructor is active.

The Logic object creates a Minefield object.

Object deletion

UML uses an X at the end of the lifeline of an object to show its deletion.

Although object deletion is not that important in languages such as Java that support automatic memory management, you can still show object deletion in UML diagrams to indicate the point at which the object ceases to be used.

Notation:

Note how the below diagram shows the deletion of the Minefield object.

Loops

Notation:

The Player calls the mark x,y command or clear x y command repeatedly until the game is won or lost.

Self invocation

UML can show a method of an object calling another of its own methods.

Notation:

The markCellAt(...) method of a Logic object is calling its own updateState(...) method.

In this variation, the Book#write() method is calling the Chapter#getText() method which in turn does a call back by calling the getAuthor() method of the calling object.

Alternative paths

UML uses alt frames to indicate alternative paths.

Notation:

Minefield calls the Cell#setMine method if the cell is supposed to be a mined cell, and calls the Cell:setMineCount(...) method otherwise.

No more than one alternative partitions be executed in an alt frame. That is, it is acceptable for none of the alternative partitions to be executed but it is not acceptable for multiple partitions to be executed.

Optional paths

UML uses opt frames to indicate optional paths.

Notation:

Logic#markCellAt(...) calls Timer#start() only if it is the first move of the player.

Parallel paths

UML uses par frames to indicate parallel paths.

Notation:

Logic is calling methods CloudServer#poll() and LocalData#poll() in parallel.

If you show parallel paths in a sequence diagram, the corresponding Java implementation is likely to be multi-threaded because a normal Java program cannot do multiple things at the same time.

Reference frames

UML uses ref frame to allow a segment of the interaction to be omitted and shown as a separate sequence diagram. Reference frames help you to break complicated sequence diagrams into multiple parts or simply to omit details you are not interested in showing.

Notation:

The details of the get minefield appearance interactions have been omitted from the diagram.

Those details are shown in a separate sequence diagram given below.

Minimal notation

To reduce clutter, optional elements (e.g, activation bars, return arrows) may be omitted if the omission does not result in ambiguities or loss of . Informal operation descriptions such as those given in the example below can be used, if more precise details are not required for the task at hand.

A minimal sequence diagram

Activity Diagrams

Introduction

Introduction

UML  activity diagrams (AD) can model workflows.  Flow charts are another type of diagram that can model workflows. Activity diagrams are the UML equivalent of flow charts.

An example activity diagram:

[source:wikipeida]

Basic Notations

Linear paths

An activity diagram (AD) captures an activity through the actions and control flows that make up the activity.

  • An action is a single step in an activity. It is shown as a rectangle with rounded corners.
  • A control flow shows the flow of control from one action to the next. It is shown by drawing a line with an arrow-head to show the direction of the flow.

Note the slight difference between the start node and the end node which represent the start and the end of the activity, respectively.

This activity diagram shows the action sequence of the activity a passenger rides the bus:

Alternate paths

A branch node shows the start of alternate paths. Each control flow exiting a branch node has a guard condition: a boolean condition that should be true for execution to take that path. Exactly one of the guard conditions should be true at any given branch node.

A merge node shows the end of alternate paths.

Both branch nodes and merge nodes are diamond shapes. Guard conditions must be in square brackets.

The AD below shows alternate paths involved in the workflow of the activity shop for product:

Some acceptable simplifications (by convention):

  • Omitting the merge node if it doesn't cause any ambiguities.
  • Multiple arrows can starting from the same corner of a branch node.
  • Omitting the [Else] condition.

The AD below illustrates the simplifications mentioned above:

Parallel paths

Fork nodes indicate the start of flows of control.

Join nodes indicate the end of parallel paths.

Both have the same notation: a bar.

In a , execution along all parallel paths should be complete before the execution can start on the outgoing control flow of the join.

In this activity diagram (from an online shop website) the actions User browses products and System records browsing data happen in parallel. Both of them need to finish before the log out action can take place.

Rakes

The rake notation is used to indicate that a part of the activity is given as a separate diagram.

Here is the AD for a game of ‘Snakes and Ladders’.

The rake symbol (in the Move piece action above) is used to show that the action is described in another subsidiary activity diagram elsewhere. That diagram is given below.

Swimlanes

It is possible to partition an activity diagram to show who is doing which action. Such partitioned activity diagrams are sometime called swimlane diagrams.

A simple example of a swimlane diagram:

Notes

Notes

UML notes can augment UML diagrams with additional information. These notes can be shown connected to a particular element in the diagram or can be shown without a connection. The diagram below shows examples of both.

Example:

Constraints

A constraint can be given inside a note, within curly braces. Natural language or a formal notation such as OCL (Object Constraint Language) may be used to specify constraints.

Example:

Miscellaneous

Object versus class diagrams

Compared to the notation for class diagrams, object diagrams differ in the following ways:

  • Show objects instead of classes:
    • Instance name may be shown
    • There is a : before the class name
    • Instance and class names are underlined
  • Methods are omitted
  • Multiplicities are omitted. Reason: an association line in an object diagram represents a connection to exactly one object (i.e., the multiplicity is always 1).

Furthermore, multiple object diagrams can correspond to a single class diagram.

Both object diagrams are derived from the same class diagram shown earlier. In other words, each of these object diagrams shows ‘an instance of’ the same class diagram.

When the class diagram has an inheritance relationship, the object diagram should show either an object of the parent class or the child class, but not both.

Suppose Employee is a child class of the Person class. The class diagram will be as follows:

Now, how do you show an Employee object named jake?

  • This is not correct, as there should be only one object.

  • This is OK.

  • This is OK, as jake is a Person too. That is, we can show the parent class instead of the child class if the child class doesn't matter to the purpose of the diagram (i.e., the reader of this diagram will not need to know that jake is in fact an Employee).

Association labels/roles can be omitted unless they add value (e.g., showing them is useful if there are multiple associations between the two classes in concern -- otherwise you wouldn't know which association the object diagram is showing)

Consider this class diagram and the object diagram:

We can clearly see that both Adam and Eve lives in hall h1 (i.e., OK to omit the association label lives in) but we can't see if History is Adam's major or his minor (i.e., the diagram should have included either an association label or a role there). In contrast, we can see Eve is an English major.