1. Chapter 5, Analysis: Dynamic Modeling

2. Outline of the Lecture Dynamic modeling
Sequence diagrams
State diagrams
Using dynamic modeling for the design of user interfaces
Analysis example
Requirements analysis document template
Requirements analysis model validation

3. How do you find classes?
In previous lectures we have already established the following sources
Application domain analysis: Talk to client to identify abstractions
Application of general world knowledge and intuition
Scenarios
Natural language formulation of a concrete usage of the system
Use Cases
Natural language formulation of the functions of the system
Textual analysis of problem statement (Abbott)
Today we show how identify classes from dynamic models
Actions and activities in state chart diagrams are candidates for public operations in classes
Activity lines in sequence diagrams are also candidates for objects
Goal: Come up with
> a domain model (domain specific objects)
> a taxonomy (class hierarchy)

4. Dynamic Modeling with UML
Diagrams for dynamic modeling
Interaction diagrams describe the dynamic behavior between objects
Statecharts describe the dynamic behavior of a single object
Interaction diagrams
Sequence Diagram:
Dynamic behavior of a set of objects arranged in time sequence.
Good for real-time specifications and complex scenarios
Collaboration Diagram :
Shows the relationship among objects. Does not show time
State Chart Diagram:
A state machine that describes the response of an object of a given class to the receipt of outside stimuli (Events).
Activity Diagram: A special type of statechart diagram, where all states are action states (Moore Automaton)
Parts of the DAIMLER system has a rich dynamic behavior, so the dynamic model is important.
Note that the dynamic model is an abstraction like the other models as well. Depending on what properties are most characteristic in a system, one of the models will be richer or more significant than others. If you have a choice to decide on a single model, which one would you use?
Object Model: Database, Preventive maintenance,
Dynamic Model: Communication, Notification Server, AGTS Server, Change management (Information flow!)
Temporal relationships are difficult to understand. Rumbaugh thinks, that a system can be best understood by first examining its static structure, that is the structure of its objects and their relationships to each other at a single moment in time. Then we examine the changes to the objects and their relationships over time. (Note that not everybody thinks that way. Heraklit: Everything flows (requires dynamic modeling first), Jacobsen: Use cases (Functional description)
Control is that aspect of a system that describes the sequences of operations that occur in response to external stimuli, without consideration of what the operations do, what the operate on or how they are implemented.
The major concepts of dynamic modeling are events which represent external stimuli and states, which represent the values of objects.
We will use state diagrams, which are a standard computer science concept. Question: Who does not know what a finite state automaton is?
In general we assume, that FSA are covered in (Algorithms and Datastructures)
Similar to the problem in object modeling, we need to avoid to clutter our state diagrams (spaghetti diagrams). We will show that states and events can be organized into generalization hierarchies that share structure and behavior.

5. Dynamic Modeling Definition of dynamic model: Purpose:
A collection of multiple state chart diagrams, one state chart diagram for each class with important dynamic behavior.
Purpose:
Detect and supply methods for the object model
How do we do this?
Start with use case or scenario
Model interaction between objects => sequence diagram
Model dynamic behavior of a single object => statechart diagram

6. Start with Flow of Events from Use Case
Flow of events from “Dial a Number” Use case:
Caller lifts receiver
Dial tone begins
Caller dials
Phone rings
Callee answers phone
Ringing stops
....
A scenario is a sequence of events that occurs during one particular execution of a system. The scope of a scenario can vary: it may include all events in the system, or it may include only those events generated by certain objects of the system.
A scenario can be the historical record of executing a system or a thought experiment of exexuting a proposed system.
Each event transmits information from one object to another. For example, dial tone begins transmits a signal from the phone line to the caller.

7. What is an Event? Something that happens at a point in time
Relation of events to each other:
Causally related: Before, after,
Causally unrelated: concurrent
An event sends information from one object to another
Events can be grouped in event classes with a hierarchical structure. ‘Event’ is often used in two ways:
Instance of an event class: “New IETM issued on Thursday September 14 at 9:30 AM”.
Event class “New IETM”, Subclass “Figure Change”
Attribute of an event class
IETM Update (9:30 AM, 9/14/99)
Car starts at ( 4:45pm, Monroeville Mall, Parking Lot 23a)
Mouse button down(button#, tablet-location)

8. Sequence Diagram
From the flow of events in the use case or scenario proceed to the sequence diagram
A sequence diagram is a graphical description of objects participating in a use case or scenario using a DAG (direct acyclic graph) notation
Relation to object identification:
Objects/classes have already been identified during object modeling
Objects are identified as a result of dynamic modeling
Heuristic:
A event always has a sender and a receiver.
The representation of the event is sometimes called a message
Find them for each event => These are the objects participating in the use case

9. An Example Flow of events in a “Get SeatPosition” use case :
1. Establish connection between smart card and onboard computer
2. Establish connection between onboard computer and sensor for seat
3. Get current seat position and store on smart card
Which are the objects?
This is an example of a use case that describes the establishment of connection between a technician onboard the train trying to talk to a maintenance expert currently located at an unknown location. The location is known by Headquarters.
The Technician has to first reach headquarters to find the expert and is doing this via communication to the Wayside (Field Site) computer.

10. Sequence Diagram for “Get SeatPosition”
Onboard Computer
Seat
Smart Card
1. Establish connection between smart card and onboard computer
2. Establish connection between onboard computer and sensor for seat
3. Get current seat position and store on smart card
Establish Connection
Establish Connection
The Technician informs the Wayside Computer that he/she would like to setup a connection to the Maintenance Expert (1). The Wayside Computer attempts to resolve the name of the Maintenance Expert in Pittsburgh, but that person cannot be located here (2). Thus, the Wayside Computer issues the setup connection event to the Headquarter Computer to find out where the Maintenance Expert is located (3). The Headquarter Computer is able to resolve the name of the Maintenance Expert since this person is located at AEG Headquarters (4). The setup connection is thus finally forwarded to the appropriate Maintenance Expert (5). The Maintenance Expert can reject or accept the connection. If he/she accepts the connection, then the scenario continues as follows (6). The Headquarter Computer establishes a connection to the Maintenance Expert (7). The Headquarter Computer then accepts a connection from the Way-side Computer to complete the second leg of the of the connection (8). In turn, the Wayside Com-puter establishes a connection to the Headquarter Computer (9). The Wayside Computer now accepts the connection from the Technician (10). Finally, the Technician establishes a connection to the Wayside Computer (11). The communication channel is now established from the Techni-cian to the Maintenance Expert and the two parties can exchange information (12).
Accept Connection
Accept Connection
Get SeatPosition
“500,575,300”
time

11. Heuristics for Sequence Diagrams
Layout:
1st column: Should correspond to the actor who initiated the use case
2nd column: Should be a boundary object
3rd column: Should be the control object that manages the rest of the use case
Creation:
Control objects are created at the initiation of a use case
Boundary objects are created by control objects
Access:
Entity objects are accessed by control and boundary objects,
Entity objects should never call boundary or control objects: This makes it easier to share entity objects across use cases and makes entity objects resilient against technology-induced changes in boundary objects.

12. Is this a good Sequence Diagram?
Smart Card
Onboard Computer
Seat
First column is not the actor
It is not clear where the boundary object is
It is not clear where the control object is
Establish Connection
Establish Connection
Layout problem: The actor is missing. Who is talking to the Smartcard?
(Create Driver actor who inserts Smart Card into Cardreader)
Creation problem: The control object is not created by the interface object (add dashed lifeline until create() call by interface object)
Access problem: The interface object accesses the control object (replace by dashed line)
Accept Connection
Accept Connection
Get SeatPosition
“500,575,300”

13. An ARENA Sequence Diagram : Create Tournament
League
Owner
:Tournament
Boundary
:Arena
:League
newTournament
(league)
:Announce
Tournament
Control
«new»
checkMax
Tournament()
setName(name)
setMaxPlayers
(maxp)
createTournament
(name, maxp)
commit()
create
Tournament
(name, maxp)
:Tournament
«new»

14. Impact on ARENA’s Object Model
Let’s assume, before we formulated the previous sequence diagram, ARENA’s object model contained the objects
League Owner, Arena, League, Tournament, Match and Player
The Sequence Diagram identified new Classes
Tournament Boundary, Announce_Tournament_Control

15. League
Owner
League 1 *
Attributes
Attributes
Operations
Operations
Tournament
Attributes
Operations
Player
Match * *
Attributes
Attributes
Operations
Operations

16. League
Owner
League 1 *
Attributes
Attributes
Operations
Operations
Attributes
Operations
Tournament_
Boundary
Tournament
Attributes
Operations
Announce_
Tournament_
Control
Attributes
Operations
Player
Match * *
Attributes
Attributes
Operations
Operations

17. Impact on ARENA’s Object Model (ctd)
The Sequence Diagram also supplied us with a lot of new events
newTournament(league)
setName(name)
setMaxPlayers(max)
Commit
checkMaxTournaments()
createTournament
Question: Who owns these events?
Answer: For each object that receives an event there is a public operation in the associated class.
The name of the operation is usually the name of the event.

18. Example from the Sequence Diagram
create
Tournament
(name, maxp)
League
Owner
:Tournament
Boundary
newTournament
(league)
:Announce
Control
«new»
setName(name)
setMaxPlayers
(maxp)
commit()
checkMax
Tournament()
:Arena
:League
createTournament is a (public)
operation owned by
Announce_Tournament_Control
createTournament
(name, maxp)

19. Tournament_ Boundary Announce_ Tournament_ Control
League
Owner
League 1 *
Attributes
Attributes
Operations
Operations
Attributes
Operations
Tournament_
Boundary
Tournament
Announce_
Tournament_
Control
Attributes
Operations
Attributes
createTournament
(name, maxp)
Player
Match * *
Attributes
Attributes
Operations
Operations

20. What else can we get out of sequence diagrams?
Sequence diagrams are derived from the use cases. We therefore see the structure of the use cases.
The structure of the sequence diagram helps us to determine how decentralized the system is.
We distinguish two structures for sequence diagrams: Fork and Stair Diagrams (Ivar Jacobsen)

21. Fork Diagram
Much of the dynamic behavior is placed in a single object, ususally the control object. It knows all the other objects and often uses them for direct questions and commands.

22. Stair Diagram
The dynamic behavior is distributed. Each object delegates some responsibility to other objects. Each object knows only a few of the other objects and knows which objects can hel with a specific behavior.

23. Fork or Stair? Which of these diagram types should be chosen?
Object-oriented fans claim that the stair structure is better
The more the responsibility is spread out, the better
However, this is not always true. Better heuristics:
Decentralized control structure
The operations have a strong connection
The operations will always be performed in the same order
Centralized control structure (better support of change)
The operations can change order
New operations can be inserted as a result of new requirements

24. UML Statechart Diagram Notation
Event trigger
With parameters
State1
State2
Event1(attr) [condition]/action
do/Activity
entry /action
Guard
condition
exit/action
Also: internal transition
and deferred events
A state is drawn as a rounded box containing a name. The state name appears in bold face
A transition is drawn as an arrow from the receiving state to the target state. Event names are written on the transition arrow, optionally followed by one or more attributes within parentheses.
A condition is a Bollean function of object values, such as emission is above legal value. Conditions can be guards on transitions. A guarded transition fires only when the event occurs and the condition is true. Conditions are shown as expressions in square brackets following the event name and its parameter list.
An action follows the event name and/or the condition by the slash (/) character.
Events that cause an action without causing a state change are written inside the state box. Internal action and self-transition are different. When an internal action occurs, neither the entry nor exit action is executed, when a self-transition occurs, these actions are executed.
Events sent to other objects are shown in dashed lines.
Notation based on work by Harel
Added are a few object-oriented modifications
A UML statechart diagram can be mapped into a finite state machine

25. Statechart Diagrams
Graph whose nodes are states and whose directed arcs are transitions labeled by event names.
We distinguish between two types of operations in statecharts:
Activity: Operation that takes time to complete
associated with states
Action: Instantaneous operation
associated with events
associated with states (reduces drawing complexity): Entry, Exit, Internal Action
A statechart diagram relates events and states for one class
An object model with a set of objects has a set of state diagrams
A state diagram relates events and states for one object.
A transition is drawn from
State diagrams would be quite useless if they only describe event patterns. A behavioral description of an object must specify what the object does in response to events. This is specified in operations attached to states and transition .
An activity is an operation that takes time to complete. Activities are always associated with a state. The notation do Activity1 within a state box indicates that activity A start on entry to the state and stops when complete.
Sometimes it is more advantageous to associate an action with a state. When? When all the state transitions cause the same action, it is a better notational convenience to list the action only once. This reduces clutter in the state diagrams.
An action is an instantaneous operation. This is of course a relative notion. What we mean by instantaneous is that the duration of the operation is insignificant when compared to the time resolution of the state diagram. It also means we do not care about the internal structure of this operation. An action is therefore associated with an event.

26. State An abstraction of the attributes of a class
State is the aggregation of several attributes a class
Basically an equivalence class of all those attribute values and links that do no need to be distinguished as far as the control structure of the system is concerned
Example: State of a bank
A bank is either solvent or insolvent
State has duration
An object model describes the possible patterns of objects, attribute values and links that can exist in a system.
A dynamic model describes the possible patterns of states, events and actions that can exist in a system.
Over time, the objects stimulate each other, resulting in a series of changes to their states.
An individual stimulus from one object to another is called an event.
Events can be organized into classes. Events can be error conditions as well as normal occurences. Event as an object : Some events are only signals (OK), others have attributes associated with them (Unix signals: Bus error, Segment violation).
Attributes are shown in parentheses after the class name. Events are important for UI, Communication and Notification. Other groups have to worry about events as well, because the interaction between the subsystems should be via the Request Broker.
The response to an event depends on the state of the object receiving it. It can include a state change or sending of another event to another object and sending a event back to the sender (return result, ok, ...)
The Events and states are duals of one another: An event separates 2 states and a state separates 2 events.

27. Example of a StateChart Diagram
coins_in(amount) / set balance
Collect Money
coins_in(amount) / add to balance
Idle
cancel / refund coins
[item empty]
[select(item)]
[change<0]
A good application for nested state diagrams in this example of a vending machine. Here we see the top level dynamic model of the object vending machine, which certainly has interesting dynamic behavior for the designer.
Question to the students: Is the dynamic model interesting for the end user? If the end user has the choice of one of the 3 models (Object model, dynamic model and functional model), which one would they use? Answer: The functional model! If I select this particular candy bar, then that is what I want to get.
Notice the activity do : dispense item. In the case, when we really have to engineer (develop) the dispenser, there is much more to it than just a activity dispense. Note, by the way, that this particular state does not have a name! The state in the upper left has a state, namely Collect Money. This is incomplete if we insist that all states have names, but it is completely natural while we do dynamic modeling. We do not necessarily immediately come up with good names for everything. We had encountered a similar problem during object modeling: In some cases, we found the class name first, in others we first find the method or methods or maybe we start with something where we only know the attribute.
There is no predefined roadmap to arrive at good models, be it object, dynamic or functional models. Creativity cannot be forced into a Procustres bed, or otherwise we will end up with an empty set of boxes!
do: test item and compute change
[change=0]
[change>0]
do: dispense item
do: make change

28. Nested State Diagram
Activities in states are composite items denoting other lower-level state diagrams
A lower-level state diagram corresponds to a sequence of lower-level states and events that are invisible in the higher-level diagram.
Sets of substates in a nested state diagram denote a superstate are enclosed by a large rounded box, also called contour.
Nested state diagrams are useful for two reasons:
As a solution to cope with the complexity in your design:
Abstraction: A state is actually more complex and leads to a finite state automaton itself. On the top level we don’t model all the complex states.
Modularization: Each state diagram has up to 7+-2 states.
Hierarchy: We apply the “ Consist of” association!

29. Example of a Nested Statechart Diagram
Idle
Collect Money
coins_in(amount) / add to balance
do: test item and compute change
do: make change
[change>0]
[item empty]
[select(item)]
[change<0]
coins_in(amount) / set balance
cancel / refund coins
[change=0]
A good application for nested state diagrams in this example of a vending machine. Here we see the top level dynamic model of the object vending machine, which certainly has interesting dynamic behavior for the designer.
Question to the students: Is the dynamic model interesting for the end user? If the end user has the choice of one of the 3 models (Object model, dynamic model and functional model), which one would they use? Answer: The functional model! If I select this particular candy bar, then that is what I want to get.
Notice the activity do : dispense item. In the case, when we really have to engineer (develop) the dispenser, there is much more to it than just a activity dispense. Note, by the way, that this particular state does not have a name! The state in the upper left has a state, namely Collect Money. This is incomplete if we insist that all states have names, but it is completely natural while we do dynamic modeling. We do not necessarily immediately come up with good names for everything. We had encountered a similar problem during object modeling: In some cases, we found the class name first, in others we first find the method or methods or maybe we start with something where we only know the attribute.
There is no predefined roadmap to arrive at good models, be it object, dynamic or functional models. Creativity cannot be forced into a Procustres bed, or otherwise we will end up with an empty set of boxes!
Superstate
do: dispense item

30. Example of a Nested Statechart Diagram
[change=0]
A good application for nested state diagrams in this example of a vending machine. Here we see the top level dynamic model of the object vending machine, which certainly has interesting dynamic behavior for the designer.
Question to the students: Is the dynamic model interesting for the end user? If the end user has the choice of one of the 3 models (Object model, dynamic model and functional model), which one would they use? Answer: The functional model! If I select this particular candy bar, then that is what I want to get.
Notice the activity do : dispense item. In the case, when we really have to engineer (develop) the dispenser, there is much more to it than just a activity dispense. Note, by the way, that this particular state does not have a name! The state in the upper left has a state, namely Collect Money. This is incomplete if we insist that all states have names, but it is completely natural while we do dynamic modeling. We do not necessarily immediately come up with good names for everything. We had encountered a similar problem during object modeling: In some cases, we found the class name first, in others we first find the method or methods or maybe we start with something where we only know the attribute.
There is no predefined roadmap to arrive at good models, be it object, dynamic or functional models. Creativity cannot be forced into a Procustres bed, or otherwise we will end up with an empty set of boxes!
Superstate
do: dispense item

31. Example of a Nested Statechart Diagram
‘Dispense item’ as
a composite activity:
‘Dispense item’ as
an atomic activity:
do: move arm
to row
arm ready
A good application for nested state diagrams in this example of a vending machine. Here we see the top level dynamic model of the object vending machine, which certainly has interesting dynamic behavior for the designer.
Question to the students: Is the dynamic model interesting for the end user? If the end user has the choice of one of the 3 models (Object model, dynamic model and functional model), which one would they use? Answer: The functional model! If I select this particular candy bar, then that is what I want to get.
Notice the activity do : dispense item. In the case, when we really have to engineer (develop) the dispenser, there is much more to it than just a activity dispense. Note, by the way, that this particular state does not have a name! The state in the upper left has a state, namely Collect Money. This is incomplete if we insist that all states have names, but it is completely natural while we do dynamic modeling. We do not necessarily immediately come up with good names for everything. We had encountered a similar problem during object modeling: In some cases, we found the class name first, in others we first find the method or methods or maybe we start with something where we only know the attribute.
There is no predefined roadmap to arrive at good models, be it object, dynamic or functional models. Creativity cannot be forced into a Procustres bed, or otherwise we will end up with an empty set of boxes!
do: move arm
to column
arm ready
do: dispense item
do: push item
off shelf

32. Expanding activity “do:dispense item”
‘Dispense item’ as
an atomic activity:
[change=0]
do: dispense item
Here is the nested state diagram for do : dispense item.
The activity dispense item is now decomposed (remember, decompostion is one of the 3 pillar stones to deal with complexity) into 3 distinct states, namely
Move arm to correct row
Move arm to correct column
Push item from the shelf
As a result of this refinement, we actually see two new objects emerging, namely row, column and shelf. And a new method, namely Push Item().
This is another example, of where the dynamic model helps with the identification of objects and methods.
Here are two good heuristics for dynamic modeling that I would like you to know as well:
1) Any event in a state diagram is a good candidate for a method of a class.
2) If the class does not yet exist to which this event should be attached as method, its existence should be postulated (and verified by the customer and/or analyst).
‘Dispense item’ as a composite activity:
do: move arm
to row
do: push item
off shelf
do: move arm
to column
arm
ready
arm
ready

33. Superstates Goal: Avoid spaghetti models
Reduce the number of lines in a state diagram
Transitions from other states to the superstate enter the first substate of the superstate.
Transitions to other states from a superstate are inherited by all the substates (state inheritance)

34. Modeling Concurrency Two types of concurrency 1. System concurrency
State of overall system as the aggregation of state diagrams, one for each object. Each state diagram is executing concurrently with the others.
2. Object concurrency
An object can be partitioned into subsets of states (attributes and links) such that each of them has its own subdiagram.
The state of the object consists of a set of states: one state from each subdiagram.
State diagrams are divided into subdiagrams by dotted lines.
Concurrency is an important aspect any application for two reasons:
Maintainability: If two classes are really concurrent, that means, they are not related to each other, and they can be designed and developed completely independently from each other.
Improved Performance: Concurrency also means in many cases a possible source for fast response time. If methods can be executed in parallel, and not in serial fashion, the systems response time will be better.
Concurrency within a state of a single object arises when the object can be paritioned into subsets of attributes, each of which has its own subdiagram.
In this case, the state of the object comprises one state from each subdiagram. Note that these subdiagrams don’t have to be independent. The same event can cause state transitions in more than one subdiagram.
Examples of concurrency within an object: A dispenser machine, that simultaneously dispenses cash and ejects the card.
Often concurrency within an object is discovered after the object has been identified. It might be the source for iterating on the object identification and question if there are not two separate objects in the problem that are worth modeling.

35. Example of Concurrency within an Object
Splitting control
Synchronization
Emitting
Do: Dispense
Cash taken
Cash
Concurrency might be detected within a single object. This sometimes means that our overall decomposition of the system or our initial object identification was too coarse grain.
If there is concurrency, the first question an analyst should ask: What two objects are hidden in the currently modeled single objects? In many cases, uncovering these two yet unknown classes will lead to new insights in the application or result in a better taxonomy or object model.
In some cases, the object is inherently not further decomposable. But even in this case, the discussion of concurrency is very important at analysis level, because it will have important ramifications during system design (mappability of the concurrent methods on two different processors due to data parallelism, for example) and implementation (choice of programming language that supports light level threads instead of heavy level processes).
Setting
Ready Up
to r
eset
Ready
Do: Eject
Card
Card taken

36. State Chart Diagram vs Sequence Diagram
State chart diagrams help to identify:
Changes to an individual object over time
Sequence diagrams help to identify
The temporal relationship of between objects over time
Sequence of operations as a response to one ore more events

37. Dynamic Modeling of User Interfaces
Statechart diagrams can be used for the design of user interfaces
Also called Navigation Path
States: Name of screens
Graphical layout of the screens associated with the states helps when presenting the dynamic model of a user interface
Activities/actions are shown as bullets under screen name
Often only the exit action is shown
State transitions: Result of exit action
Button click
Menu selection
Cursor movements
Good for web-based user interface design

38. Navigation Path Example
Diagnostics Menu
User moves cursor to Control Panel or Graph
Graph
User selects data group and type of graph
Control panel
User selects functionality of sensors
Selection
User selects data group
Field site
Car
Sensor group
Time range
User selects type of graph
time line
histogram
pie chart
Define
User defines a sensor event
from a list of events
Enable
User can enable a sensor event from a list of sensor events
Disable
User can disable a sensor event from a list of sensor events
List of events
User selects event(s)
Visualize
User views graph
User can add data groups for being viewed
List of sensor events
User selects sensor event(s)
Link
User makes a link (doclink)

39. Practical Tips for Dynamic Modeling
Construct dynamic models only for classes with significant dynamic behavior
Avoid “analysis paralysis”
Consider only relevant attributes
Use abstraction if necessary
Look at the granularity of the application when deciding on actions and activities
Reduce notational clutter
Try to put actions into state boxes (look for identical actions on events leading to the same state)
A few practical tips.
The most important one: If you have a couple of hundred objects, that does not necessarily mean that you have a couple of hundred dynamic models.
Only those objects with significant dynamic behavior need to be modeled dynamically.
Also, if two objects have the same dynamic behavior, it is often only necessary to model one of these objects with a dynamic model and then describe with a single sentence that the same dynamic model applies to both objects.
The Power and Headlight classes from our Mini People Mover are a good example. Only one of them needs to be modeled dynamically, because they both have the same identical dynamic behavior.
Because the states are permutations over different values of attribute settings, it is important to use abstraction if necessary to filter out irrelevant attributes. That is, if an attribute is not contributing to the dynamic behavior of an object, it does not have be modeled in the dynamic model. This often helps in reducing the number of states of the state diagram for the object.
What makes up an action and what makes up an activity is relative, as we have already discussed earlier. If in doubt, it is always better to model something first as an action, unless somebody points out that it has internal structure which is relevant for the application, or it has a duration which is significant with respect to the overall lifetime of the object. Example: An alarm “Propulsion System Failure” should be modeled as an activity, because the propulsion system will most probably be down for a while after such a failure.

40. Summary: Requirements Analysis
1. What are the transformations?
Create scenarios and use case diagrams
Talk to client, observe, get historical records, do thought experiments
Functional Modeling
2. What is the structure of the system?
Create class diagrams
Identify objects.
What are the associations between them? What is their multiplicity?
What are the attributes of the objects?
What operations are defined on the objects?
Object Modeling
The textbook has a relatively elaborate dynamic model for a programmable thermostat. To get the concept through to you and to show the relation to the object model, I would like to use an even simpler example: The movement of a toy train.
3. What is its behavior?
Create sequence diagrams
Identify senders and receivers
Show sequence of events exchanged between objects. Identify event dependencies and event concurrency.
Create state diagrams
Only for the dynamically interesting objects.
Dynamic Modeling

41. Let’s Do Analysis 1. Analyze the problem statement
Identify functional requirements
Identify nonfunctional requirements
Identify constraints (pseudo requirements)
2. Build the functional model:
Develop use cases to illustrate functionality requirements
3. Build the dynamic model:
Develop sequence diagrams to illustrate the interaction between objects
Develop state diagrams for objects with interesting behavior
4. Build the object model:
Develop class diagrams showing the structure of the system

42. Problem Statement: Direction Control for a Toy Car
Power is turned on
Car moves forward and car headlight shines
Power is turned off
Car stops and headlight goes out.
Headlight shines
Headlight goes out.
Car runs backward with its headlight shining.
Power is turned off
Car stops and headlight goes out.
Power is turned on
Headlight shines
Headlight goes out.
Car runs forward with its headlight shining.

43. Find the Functional Model: Do Use Case Modeling
Use case 1: System Initialization
Entry condition: Power is off, car is not moving
Flow of events:
Driver turns power on
Exit condition: Car moves forward, headlight is on
Use case 2: Turn headlight off
Entry condition: Car moves forward with headlights on
Driver turns power off, car stops and headlight goes out.
Driver turns power on, headlight shines and car does not move.
Driver turns power off, headlight goes out
Exit condition: Car does not move, headlight is out

44. Use Cases continued Use case 3: Move car backward
Entry condition: Car is stationary, headlights off
Flow of events:
Driver turns power on
Exit condition: Car moves backward, headlight on
Use case 4: Stop backward moving car
Entry condition: Car moves backward, headlights on
Driver turns power off, car stops, headlight goes out.
Power is turned on, headlight shines and car does not move.
Power is turned off, headlight goes out.
Exit condition: Car does not move, headlight is out.
Use case 5: Move car forward
Entry condition: Car does not move, headlight is out
Flow of events
Exit condition:
Car runs forward with its headlight shining.

45. Use Case Pruning Do we need use case 5?
Use case 1: System Initialization
Entry condition: Power is off, car is not moving
Flow of events:
Driver turns power on
Exit condition: Car moves forward, headlight is on
Use case 5: Move car forward
Entry condition: Car does not move, headlight is out
Flow of events
Exit condition:
Car runs forward with its headlight shining.

46. Find the Dynamic Model: Create sequence diagram
Name: Drive Car
Sequence of events:
Billy turns power on
Headlight goes on
Wheels starts moving forward
Wheels keeps moving forward
Billy turns power off
Headlight goes off
Wheels stops moving
. . .

47. Sequence Diagram for Drive Car Scenario
:Wheel
:Headlight
Billy:Driver
Power(on)
Power(on)
Power(off)
Power(on)

48. Toy Car: Dynamic Model Wheel Headlight Forward Off power power off on
Change arrows for Headlight states in opposite directions.
Then ask question: Anybody sees what is wrong with the slide? In many cases the students will identify the arrows pointing in the wrong direction.
Stationary
Stationary
power
power on
off
Backward

49. Toy Car: Object Model Car Power Headlight Wheel Train Headlight Wheel
Status: (On, Off)
TurnOn()
TurnOff()
Headlight
Wheel
Status: (On, Off)
Motion: (For
ward,
Backward,
Switch_On()
Stationary)
Let them find the key abstractions: Operator, Train. Then the 3 objects in the train: Power, Headlight, Wheel. Then show that the Power and Headlight objects operate the same way, so only one object is needed to for the event scenario.
Switch_Off()
Star
t_Moving()
Stop_Moving()
Train
Wheel
Motion: (For
ward,
Stationar y)
Backward,
Star
t_Moving()
Stop_Moving()
Power
Headlight
Status: (On, Off)
Status: (On, Of)f
TurnOn()
Switch_On()
TurnOff()
Switch_Of
f()

50. Additional constraints in ARENA Project
Interface Engineering
Provide ARENA players with access to an existing game: Bumpers
Complete Java Code for Bumpers posted on SE Discuss
Greenfield Engineering
Design a new game and provide ARENA players with access to the new game
Constraints:
Extensibility
Scalability
Additional Constraint:
The existing ARENA code does not have to be recompiled when the new game is introduced
ARENA does not have to be shut down (currently running games can continue) when the new game is introduced
Is the “NotShutDown” requirement realistic?

51. Impact on ARENA Object Model
New
System
Legacy
System

52. Clarification: Terminology in REQuest
User tasks describe application
domain Le v
el 2
el 1
el 3
el 4
Use Cases describe interactions
Services describe system
Participating Objects describe domain entities A B

53. ARENA user tasks (top level use cases)

54. AnnounceTournament (Part of OrganizeTournament)

55. When is a model dominant?
Object model: The system has objects with nontrivial state.
Dynamic model: The model has many different types of events: Input, output, exceptions, errors, etc.
Functional model: The model performs complicated transformations (e.g. computations consisting of many steps).
Which of these models is dominant in the following three cases?
Compiler
Database system
Spreadsheet program

56. Dominance of models Compiler: Database systems: Spreadsheet program:
The functional model most important. (Why?)
The dynamic model is trivial because there is only one type input and only a few outputs.
Database systems:
The object model most important.
The functional model is trivial, because the purpose of the functions is usually to store, organize and retrieve data.
Spreadsheet program:
The functional model most important.
The dynamic model is interesting if the program allows computations on a cell.
The object model is trivial, because the spreadsheet values are trivial and cannot be structured further. The only interesting object is the cell.

57. Collaborative Analysis
A system is a collection of subsystems providing services
Analysis of services is provided by a set of the teams who provide the models for their subsystems
Integration of subsystem models into the full system model by the architecture team
Analysis integration checklist:
Are all the classes mentioned in the data dictionary?
Are the names of the methods consistent with the names of actions, activities, events or processes?
Check for assumptions made by each of the services
Missing methods, classes
Unmatched associations

58. Analysis: UML Activity Diagram

59. Object Model Integration in a large Project
All Teams
Revised System
Model
Model
Changes
Analysis
Review
Team 1
User Interface
Team
Analysis
Analysis
Integrated
System
Model
User Interface
Module 1
Module
Integration
Architecture Team
Module 2
Module 3
Module 4
Module 5
Analysis
Analysis
Analysis
Analysis
Team 3
Team 4
Team 5
Team 2

60. Requirements Analysis Document Template
1. Introduction
2. Current system
3. Proposed system
3.1 Overview
3.2 Functional requirements
3.3 Nonfunctional requirements
3.4 Constraints (“Pseudo requirements”)
3.5 System models
Scenarios
Use case model
Object model
Data dictionary
Class diagrams
Dynamic models
User interface
4. Glossary

61. Section 3.5 System Model 3.5.1 Scenarios
- As-is scenarios, visionary scenarios
3.5.2 Use case model
- Actors and use cases
3.5.3 Object model
- Data dictionary
- Class diagrams (classes, associations, attributes and operations)
3.5.4 Dynamic model
- State diagrams for classes with significant dynamic behavior
- Sequence diagrams for collaborating objects (protocol)
3.5.5 User Interface
- Navigational Paths, Screen mockups

62. Section 3.3 Nonfunctional Requirements
User interface and human factors
Documentation
Hardware considerations
Performance characteristics
Error handling and extreme conditions
System interfacing
Quality issues
System modifications
Physical environment
Security issues
Resources and management issues

63. Nonfunctional Requirements: Trigger Questions
3.3.1 User interface and human factors
What type of user will be using the system?
Will more than one type of user be using the system?
What sort of training will be required for each type of user?
Is it particularly important that the system be easy to learn?
Is it particularly important that users be protected from making errors?
What sort of input/output devices for the human interface are available, and what are their characteristics?
3.3.2 Documentation
What kind of documentation is required?
What audience is to be addressed by each document?
3.3.3 Hardware considerations
What hardware is the proposed system to be used on?
What are the characteristics of the target hardware, including memory size and auxiliary storage space?

64. Nonfunctional Requirements, ctd
3.3.4 Performance characteristics
Are there any speed, throughput, or response time constraints on the system?
Are there size or capacity constraints on the data to be processed by the system?
3.3.5 Error handling and extreme conditions
How should the system respond to input errors?
How should the system respond to extreme conditions?
3.3.6 System interfacing
Is input coming from systems outside the proposed system?
Is output going to systems outside the proposed system?
Are there restrictions on the format or medium that must be used for input or output?

65. Nonfunctional Requirements, ctd
3.3.7 Quality issues
What are the requirements for reliability?
Must the system trap faults?
What is the maximum time for restarting the system after a failure?
What is the acceptable system downtime per 24-hour period?
Is it important that the system be portable (able to move to different hardware or operating system environments)?
3.3.8 System Modifications
What parts of the system are likely candidates for later modification?
What sorts of modifications are expected?
3.3.9 Physical Environment
Where will the target equipment operate?
Will the target equipment be in one or several locations?
Will the environmental conditions in any way be out of the ordinary (for example, unusual temperatures, vibrations, magnetic fields, ...)?

66. Nonfunctional Requirements, ctd
Security Issues
Must access to any data or the system itself be controlled?
Is physical security an issue?
Resources and Management Issues
How often will the system be backed up?
Who will be responsible for the back up?
Who is responsible for system installation?
Who will be responsible for system maintenance?

67. Constraints (Pseudo Requirements)
Any client restriction on the solution domain
Examples:
The target platform must be an IBM/360
The implementation language must be COBOL
The documentation standard X must be used
A dataglove must be used
ActiveX must be used
The system must interface to a papertape reader

68. Outline of the Lecture Dynamic modeling
Sequence diagrams
State diagrams
Using dynamic modeling for the design of user interfaces
Analysis example
Requirements analysis document template
Requirements analysis model validation

69. Verification and Validation of modelsfR
fMA
MAnalysis
Analysis fM M I
fMS
MSystem
System
Design
fMD
MObject
Object
MImpl
fImpl
Implemen-
tation
Validation
Verification
Verification
Verification

70. Correctness, Completeness and Consistency
Verification is an equivalence check between the transformation of two models:
We have two models, is the transformation between them correct?
Validation is different. We don’t have two models, we need to compare one model with reality
“Reality” can also be an artificial system, like an legacy system
Validation is a critical step in the development process Requirements should be validated with the client and the user.
Techniques: Formal and informal reviews (Meetings, requirements review)
Requirements validation involves the following checks
Correctness
Completeness
Ambiguity
Realism

71. Modeling Checklist for the Review
Is the model correct?
A model is correct if it represents the client’s view of the the system: Everything is the model represents an aspect of reality
Is the model complete?
Every scenario through the system, including exceptions, is described.
Is the model consistent?
The model does not have components that contradict themselves (for example, deliver contradicting results)
Is the model unambiguous?
The model describes one system (one reality), not many
Is the model realistic?
The model can be implemented without problems

72. Diagram Checklist for the RAD
One problem with modeling:
We describe a system model with many different views (class diagram, use cases, sequence diagrams, )state charts)
We need to check the equivalence of these views as well
Syntactical check of the models
Check for consistent naming of classes, attributes, methods in different subsystems
Identify dangling associations (associations pointing to nowhere)
Identify double- defined classes
Identify missing classes (mentioned in one model but not defined anywhere)
Check for classes with the same name but different meanings
Don’t rely on CASE tools for these checks
Many of the existing tools don’t do all these checks for you.
Examples for syntactical problems with UML diagrams

73. Different spellings in different diagrams
(from)
UML Sequence Diagram
UML Class Diagram
Attributes
Operations
League
Tournament
Player
Match
Owner 1 *
Tournament_
Boundary
makeTournament
(name, maxp)
Announce_
Control
createTournament
(name, maxp)
Different spellings
In different models
for the same operation?

74. Omissions in some diagrams
Attributes
Operations
League
Tournament
Player
Match
Owner 1 *
Tournament_
Boundary
Class Diagram
Missing class
(associated control object
Announce_Tournament
is mentioned in
Sequence diagram)
Missing
Association
(Incomplete
Analysis?)

75. Project Agreement
The project agreement represents the acceptance of (parts of) the analysis model (as documented by the requirements analysis document) by the client.
The client and the developers converge on a single idea and agree about the functions and features that the system will have. In addition, they agree on:
a list of prioritized requirements
a revision process
a list of criteria that will be used to accept or reject the system
a schedule, and a budget

76. Prioritizing requirements
High priority (“Core requirements”)
Must be addressed during analysis, design, and implementation.
A high-priority feature must be demonstrated successfully during client acceptance.
Medium priority (“Optional requirements”)
Must be addressed during analysis and design.
Usually implemented and demonstrated in the second iteration of the system development.
Low priority (“Fancy requirements”)
Must be addressed during analysis (“very visionary scenarios”).
Illustrates how the system is going to be used in the future if not yet available technology enablers are available

77. Summary
In this lecture, we reviewed the construction of the dynamic model from use case and object models. In particular, we described: In particular, we described:
Sequence and statechart diagrams for identifying new classes and operations.
In addition, we described the requirements analysis document and its components