Object-oriented Design

Download Report

Transcript Object-oriented Design

Object-Oriented Design
©Ian Sommerville 2006
Objectives
• To explain how a software design may be represented
as a set of interacting objects that manage their own
state and operations
• To describe the activities in the object-oriented design
process
• To introduce various models that can be used to
describe an object-oriented design
• To show how the UML may be used to represent these
models
Topics covered
• 1- Objects and object classes
• 2- An object-oriented design process
• 3- Design evolution
Object-oriented development
• Object-oriented analysis, design and programming are
related but distinct.
• OOA is concerned with developing an object model of
the application domain.
• OOD is concerned with developing an object-oriented
system model to implement requirements.
• OOP is concerned with realising an OOD using an OO
programming language such as Java or C++.
Characteristics of OOD
• Objects are abstractions of real-world or system entities
and manage themselves.
• Objects are independent and encapsulate state and
representation information.
• System functionality is expressed in terms of object
services.
• Shared data areas are eliminated. Objects
communicate by message passing.
• Objects may be distributed and may execute
sequentially or in parallel.
Interacting objects
Advantages of OOD
• Easier maintenance. Objects may be
understood as stand-alone entities.
• Objects are potentially reusable components.
• For some systems, there may be an obvious
mapping from real world entities to system
objects.
Object Oriented Terminologies
•
•
Class
Defines the abstract characteristics of a thing (object), including the thing's characteristics (its
attributes (variables), fields or properties) and the thing's behaviors (the things it can do, or
methods, operations or features). One might say that a class is a blueprint or factory that describes
the nature of something. For example, the class Dog would consist of traits shared by all dogs,
such as breed and fur color (characteristics), and the ability to bark and sit (behaviors). Classes
provide modularity and structure in an object-oriented computer program. A class should typically
be recognizable to a non-programmer familiar with the problem domain, meaning that the
characteristics of the class should make sense in context. Also, the code for a class should be
relatively self-contained (generally using encapsulation). Collectively, the properties and methods
defined by a class are called members.
•
•
Object
A pattern (exemplar) of a class. The class of Dog defines all possible dogs by listing the
characteristics and behaviors they can have; the object Lassie is one particular dog, with particular
versions of the characteristics. A Dog has fur; Lassie has brown-and-white fur.
Object Oriented Terminologies
•
•
Instance
One can have an instance of a class or a particular object. The instance is the actual object
created at runtime. In programmer jargon, the Lassie object is an instance of the Dog class. The
set of values of the attributes of a particular object is called its state. The object consists of state
and the behavior that's defined in the object's class.
•
•
Method
An object's abilities. In language, methods (sometimes referred to as "functions") are verbs. Lassie,
being a Dog, has the ability to bark. So bark() is one of Lassie's methods. She may have other
methods as well, for example sit() or eat() or walk() or save_timmy(). Within the program, using a
method usually affects only one particular object; all Dogs can bark, but you need only one
particular dog to do the barking.
Object Oriented Terminologies
•
•
Message passing
"The process by which an object sends data to another object or asks the other object to invoke a
method." Also known to some programming languages as interfacing. For example, the object
called Breeder may tell the Lassie object to sit by passing a "sit" message which invokes Lassie's
"sit" method. The syntax varies between languages, for example: [Lassie sit] in Objective-C.
Object Oriented Terminologies
•
•
•
•
•
Inheritance
"Subclasses" are more specialized versions of a class, which inherit attributes and behaviors from
their parent classes, and can introduce their own.
For example, the class Dog might have sub-classes called Collie, Chihuahua, and
GoldenRetriever. In this case, Lassie would be an instance of the Collie subclass. Suppose the
Dog class defines a method called bark() and a property called furColor. Each of its sub-classes
(Collie, Chihuahua, and GoldenRetriever) will inherit these members, meaning that the
programmer only needs to write the code for them once.
Each subclass can alter its inherited traits. For example, the Collie class might specify that the
default furColor for a collie is brown-and-white. The Chihuahua subclass might specify that the
bark() method produces a high pitch by default. Subclasses can also add new members. The
Chihuahua subclass could add a method called tremble(). So an individual chihuahua instance
would use a high-pitched bark() from the Chihuahua subclass, which in turn inherited the usual
bark() from Dog. The chihuahua object would also have the tremble() method, but Lassie would
not, because she is a Collie, not a Chihuahua. In fact, inheritance is an "a... is a" relationship
between classes, while instantiation is an "is a" relationship between an object and a class: a Collie
is a Dog ("a... is a"), but Lassie is a Collie ("is a"). Thus, the object named Lassie has the methods
from both classes Collie and Dog.
Multiple inheritance is inheritance from more than one ancestor class, neither of these ancestors
being an ancestor of the other. For example, independent classes could define Dogs and Cats, and
a Chimera object could be created from these two which inherits all the (multiple) behavior of cats
and dogs. This is not always supported, as it can be hard both to implement and to use well.
Object Oriented Terminologies
•
•
•
•
Abstraction
Abstraction is simplifying complex reality by modelling classes appropriate to the problem, and
working at the most appropriate level of inheritance for a given aspect of the problem.
For example, Lassie the Dog may be treated as a Dog much of the time, a Collie when necessary
to access Collie-specific attributes or behaviors, and as an Animal (perhaps the parent class of
Dog) when counting Timmy's pets.
Abstraction is also achieved through Composition. For example, a class Car would be made up of
an Engine, Gearbox, Steering objects, and many more components. To build the Car class, one
does not need to know how the different components work internally, but only how to interface with
them, i.e., send messages to them, receive messages from them, and perhaps make the different
objects composing the class interact with each other.
Object Oriented Terminologies
•
•
•
Encapsulation
Encapsulation conceals the functional details of a class from objects that send messages to it.
For example, the Dog class has a bark() method. The code for the bark() method defines exactly
how a bark happens (e.g., by inhale() and then exhale(), at a particular pitch and volume). Timmy,
Lassie's friend, however, does not need to know exactly how she barks. Encapsulation is achieved
by specifying which classes may use the members of an object. The result is that each object
exposes to any class a certain interface — those members accessible to that class. The reason for
encapsulation is to prevent clients of an interface from depending on those parts of the
implementation that are likely to change in future, thereby allowing those changes to be made
more easily, that is, without changes to clients. For example, an interface can ensure that puppies
can only be added to an object of the class Dog by code in that class. Members are often specified
as public, protected or private, determining whether they are available to all classes, sub-classes
or only the defining class. Some languages go further: Java uses the default access modifier to
restrict access also to classes in the same package, C# and VB.NET reserve some members to
classes in the same assembly using keywords internal (C#) or Friend (VB.NET), and Eiffel and
C++ allow one to specify which classes may access any member.
Object Oriented Terminologies
•
•
Polymorphism
Polymorphism allows the programmer to treat derived class members just like their parent class'
members. More precisely, Polymorphism in object-oriented programming is the ability of objects
belonging to different data types to respond to method calls of methods of the same name, each
one according to an appropriate type-specific behavior. One method, or an operator such as +, -,
or *, can be abstractly applied in many different situations. If a Dog is commanded to speak(), this
may elicit a bark(). However, if a Pig is commanded to speak(), this may elicit an oink(). They both
inherit speak() from Animal, but their derived class methods override the methods of the parent
class; this is Overriding Polymorphism. Overloading Polymorphism is the use of one method
signature, or one operator such as "+", to perform several different functions depending on the
implementation. The "+" operator, for example, may be used to perform integer addition, float
addition, list concatenation, or string concatenation. Any two subclasses of Number, such as
Integer and Double, are expected to add together properly in an OOP language. The language
must therefore overload the addition operator, "+", to work this way. This helps improve code
readability. How this is implemented varies from language to language, but most OOP languages
support at least some level of overloading polymorphism. Many OOP languages also support
Parametric Polymorphism, where code is written without mention of any specific type and thus
can be used transparently with any number of new types. Pointers are an example of a simple
polymorphic routine that can be used with many different types of objects.
1- Objects and object classes
• Objects are entities in a software system which
represent instances of real-world and system
entities.
• Object classes are templates for objects. They
may be used to create objects.
• Object classes may inherit attributes and
services from other object classes.
Objects and object classes
An object is an entity that has a state and a defined set of
operations which operate on that state. The state is represented as a
set of object attributes. The operations associated with the object
provide services to other objects (clients) which request these
services when some computation is required.
Objects are created according to some object class definition.
An object class definition serves as a template for objects. It
includes declarations of all the attributes and services which should
be associated with an object of that class.
The Unified Modeling Language
• Several different notations for describing object-oriented
designs were proposed in the 1980s and 1990s.
• The Unified Modeling Language is an integration of
these notations.
• It describes notations for a number of different models
that may be produced during OO analysis and design.
• It is now a de facto standard for OO modelling.
Employee object class (UML)
Emp lo yee
n ame: s tring
ad dress : s tring
d ateOfBir th : Date
emp loy eeNo : integer
so cialSecurityNo : strin g
d epar tmen t: Dept
man ager: Employ ee
salar y : in teg er
statu s: {curren t, left, retired }
taxCod e: in teg er
. ..
jo in ()
leav e ()
retire ()
ch an geDetails ()
Object communication
• Conceptually, objects communicate by
message passing.
• Messages
– The name of the service requested by the calling object;
– Copies of the information required to execute the service
and the name of a holder for the result of the service.
• In practice, messages are often implemented
by procedure calls
– Name = procedure name;
– Information = parameter list.
Message examples
// Call a method (function) associated with a
buffer
// object that returns the next value
// in the buffer
v = circularBuffer.Get () ;
// Call the method associated with a
//
thermostat
object
that
sets
the
// temperature to be maintained
thermostat.setTemp (20) ;
Generalization and inheritance
• Objects are members of classes that define
attribute types and operations.
• Classes may be arranged in a class hierarchy
where one class (a super-class) is a generalization of
one or more other classes (sub-classes).
• A sub-class inherits the attributes and
operations from its super class and may add
new methods or attributes of its own.
• Generalization in the UML is implemented as
inheritance in OO programming languages.
A Generalization Hierarchy
Emp lo yee
Man a ger
Pro grammer
b ud getsCon tro lled
p roject
p rog Lang uages
d ateAp po in ted
Pro ject
Man a ger
Dep t.
Man a ger
p rojects
d ept
Strateg ic
Man a ger
resp on sib ilities
Advantages of inheritance
• It is an abstraction mechanism which may be
used to classify entities.
• It is a reuse mechanism at both the design and
the programming level.
• The inheritance graph is a source of
organizational knowledge about domains and
systems.
Problems with inheritance
• Object classes are not self-contained. they cannot be
understood without reference to their super-classes.
• Designers have a tendency to reuse the inheritance
graph created during analysis. Can lead to significant
inefficiency.
• The inheritance graphs of analysis, design and
implementation have different functions and should be
separately maintained.
UML associations
• Objects and object classes participate in relationships
with other objects and object classes.
• In the UML, a generalized relationship is indicated by an
association.
• Associations may be annotated with information that
describes the association.
• Associations are general but may indicate that an
attribute of an object is an associated object or that a
method relies on an associated object.
An association model
Emp lo yee
is -member-o f
Dep ar tmen t
is -man aged-by
man ag es
Man ag er
1.1 Concurrent objects
• The nature of objects as self-contained entities
make them suitable for concurrent
implementation.
• The message-passing model of object
communication can be implemented directly if
objects are running on separate processors in a
distributed system.
Servers and active objects
• Servers.
– The object is implemented as a parallel process (server)
with entry points corresponding to object operations. If no
calls are made to it, the object suspends itself and waits for
further requests for service.
• Active objects
– Objects are implemented as parallel processes and the
internal object state may be changed by the object itself and
not simply by external calls.
Active transponder object
• Active objects may have their attributes modified
by operations but may also update them
autonomously using internal operations.
• A Transponder object broadcasts an aircraft’s
position. The position may be updated using a
satellite positioning system. The object
periodically update the position by triangulation
from satellites.
An active transponder object
class Transponder extends Thread {
Position currentPosition ;
Coords c1, c2 ;
Satellite sat1, sat2 ;
Navigator theNavigator ;
public Position givePosition ()
{
return currentPosition ;
}
public void run ()
{
while (true)
{
c1 = sat1.position () ;
c2 = sat2.position () ;
currentPosition = theNavigator.compute (c1, c2) ;
}
}
} //Transponder
Java threads
• Threads in Java are a simple construct for
implementing concurrent objects.
• Threads must include a method called run() and
this is started up by the Java run-time system.
• Active objects typically include an infinite loop so
that they are always carrying out the
computation.
2- Object-oriented Design Process
• Structured design processes involve developing
a number of different system models.
• They require a lot of effort for development and
maintenance of these models and, for small
systems, this may not be cost-effective.
• However, for large systems developed by
different groups design models are an essential
communication mechanism.
Object-oriented Design Process Stages
• Highlights key activities without being tied to any
proprietary process such as the RUP & OPEN.
Third generation OO methods have two examples: the rational unified process
(RUP) and object-oriented process, environment and notation (OPEN) that have
acceptable standards in process support, project management guidelines and full
lifecycle description for OO software development.
–
–
–
–
–
2.1 Define the context and modes of use of the system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
Weather system description
A weather mapping system is required to generate weather maps on a
regular basis using data collected from remote, unattended weather stations
and other data sources such as weather observers, balloons and satellites.
Weather stations transmit their data to the area computer in response to a
request from that machine.
The area computer system validates the collected data and integrates it with
the data from different sources. The integrated data is archived and, using
data from this archive and a digitised map database a set of local weather
maps is created. Maps may be printed for distribution on a special-purpose
map printer or may be displayed in a number of different formats.
2.1 System context and models of use
•
•
•
2.1 Define the context and modes of use of the
system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
Develop an understanding of the relationships between the
software being designed and its external environment
System context
– A static model that describes other systems in the
environment. Can use An association model. The create
UMLs (next slides) for sub-systems. Use a subsystem model
to show other systems. Following slide shows the systems
around the weather station system.
Model of system use
– When your system is dynamic model that describes how the
system interacts with its environment. The RUP recommend
to use use-cases to show interactions
Layered architecture
« sub sy stem»
Data d isp lay
Data display layerwhereobjectsare
co ncerned with p reparin g and
p res en tin g the d ata in a hu manread ab le form
« sub sy stem»
Data arch iving
Data arch iving lay er wh ere o bjects
are concernedwiths toring the data
for future p ro ces sin g
« sub sy stem»
Data p roces sin g
Data process ing layerwhere objects
are con cerned with ch eckin g an d
in teg ratin g the co llected data
« sub sy stem»
Data co llection
Data co llection layer wh ere o bjects
are con cerned with acqu irin g data
from remo te so urces
Subsystems in the weather mapping system
« sub sy stem»
Data co llection
« sub sy stem»
Data d isp lay
Obs erv er
Satellite
Comms
Weath er
station
Balloo n
Data
Data
storage
storage
Data
in teg ratio n
Map store
Map
d isp lay
Map
p rin ter
Map
« sub sy stem»
Data arch iving
« sub sy stem»
Data p roces sin g
Data
ch ecking
User
User
ininter
terface
face
Data store
Use-case models
• Use-case models are used to represent each interaction
with the system.
• A use-case model shows the system features as
ellipses and the interacting entity as a stick figure.
• Next slides show weather station interact with external
entities
• Use case describes in natural language to help identify
system objects …
Use-cases for the weather station
Star tu p
Shu tdo wn
Rep ort
Calibrate
Tes t
Use-case description
System
Use -case
Actors
Data
Stimulus
Response
C omments
Weather station
Report
Weather data collection system, Weather station
The weather station sends a summary of the weather data that has been
collected from the instruments in the collection period to the weather data
collection system. T he data sent are the maximum minimum and average
ground and air temperatures, the maximum, minimum and average air
pressures, the maximum, minimum and average wind speeds, the total
rainfall and the wind direction as sampled at 5 minute intervals.
The weather data collection system establishes a modem link with the
weather station and reque sts transmission of the data.
The summarised data is sent to the weather data collection system
Weather stations are usua lly asked to report once per hour but this
frequency may differ from one station to the other and may be modified in
future.
2.2 Architectural design
2.1 Define the context and modes of use of the system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
• Once interactions between the system and its
environment have been understood, you use this
information for designing the system architecture.
• A layered architecture as discussed earlier is
appropriate for the weather station
– Interface layer for handling communications;
– Data collection layer for managing instruments;
– Instruments layer for collecting data.
• There should normally be no more than 7 entities in an
architectural model.
Weather station architecture
2.3 Object identification
2.1 Define the context and modes of use of the system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
• Identifying objects (or object classes) is the most
difficult part of object oriented design.
• There is no 'magic formula' for object
identification. It relies on the skill, experience
and domain knowledge of system designers.
• Object identification is an iterative process. You
are unlikely to get it right first time.
Approaches to identification
• Use a grammatical approach based on a natural language
description of the system (used in Hood OOD method, in European
aerospace industry). Objectives and Attributes are nouns;
operations and services are verbs.
• Base the identification on tangible things in the application
domain. i.e domain:aircraft, role:manager, event: request, interaction:meeting
• Understand overall behavioural of the system, break it into
parts, assign parts, participants play important roles
• Use a scenario-based analysis. The objects, attributes
and methods in each scenario are identified. Teams
responsible define object, attributes and methods
Weather station Object Description
A weather station is a package of software controlled instruments
which collects data, performs some data processing and transmits
this data for further processing. The instruments include air and
ground thermometers, an anemometer, a wind vane, a barometer
and a rain gauge. Data is collected periodically.
When a command is issued to transmit the weather data, the
weather station processes and summarises the collected data. The
summarised data is transmitted to the mapping computer when a
request is received.
Weather station object classes
• Weather station
– The basic interface of the weather station to its environment.
It therefore reflects the interactions identified in the use-case
model.
• Weather data
– Encapsulates the summarised data from the instruments.
• Ground thermometer, Anemometer, Barometer
– Application domain objects that are ‘hardware’ objects
related to the instruments in the system.
Weather station object classes
Weath erData
Weath erStatio n
id en tifier
airTemp er atures
gro un dTemp er atures
wind Speeds
wind Direction s
p res su res
rainf all
repo r tWeath er ()
calib rate (in strumen ts)
test ()
star tu p (in strumen ts)
sh utdo wn (ins tru men ts)
co llect ()
su mmaris e ()
Gro un d
th ermo met er
Anemo met er
Baro met er
temp er ature
wind Speed
wind Direction
p res su re
h eig ht
tes t ()
calib rate ()
tes t ()
test ()
calib rate ()
•
•
•
Weather station
–
The basic interface of the weather station to its
environment. It therefore reflects the interactions
identified in the use-case model.
Weather data
–
Encapsulates the summarised data from the
instruments.
Ground thermometer, Anemometer, Barometer
–
Application domain objects that are ‘hardware’
objects related to the instruments in the system.
Further objects and object refinement
• Use domain knowledge to identify more objects and
operations
– Weather stations should have a unique identifier;
– Weather stations are remotely situated so instrument failures
have to be reported automatically. Therefore attributes and
operations for self-checking are required.
• Active or passive objects
– In this case, objects are passive and collect data on request
rather than autonomously. This introduces flexibility at the
expense of controller processing time.
2.4 Design models
2.1 Define the context and modes of use of the system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
• Design models show the objects and object
classes and relationships between these entities.
• Static models describe the static structure of the
system in terms of object classes and
relationships.
• Dynamic models describe the dynamic
interactions between objects.
Design models Examples
• A- Sub-system models that show logical groupings of
objects into coherent subsystems.
• B- Sequence models that show the sequence of object
interactions.
• C- State machine models that show how individual
objects change their state in response to events.
• D- Other models include use-case models, aggregation
models, generalisation models, etc.
A) Subsystem Models
• Shows how the design is organized into logically
related groups of objects.
• In the UML, these are shown using packages an encapsulation construct. This is a logical
model. The actual organization of objects in the
system may be different.
Weather station subsystems
« sub sy stem»
In ter face
« sub sy stem»
Data co llection
CommsCon tro ller
Weath erData
In strumen t
Statu s
Weath erStation
« sub sy stem»
In strumen ts
Air
t hermo meter
Grou n d
t hermo meter
Rain Gaug e
Baro meter
Anemo meter
Win dVan e
B) Sequence Models
• Sequence models show the sequence of object
interactions that take place
– Objects are arranged horizontally across the top;
– Time is represented vertically so models are read top to
bottom;
– Interactions are represented by labelled arrows, Different
styles of arrow represent different types of interaction;
– A thin rectangle in an object lifeline represents the time when
the object is the controlling object in the system.
Data collection sequence
–
–
–
–
Objects are arranged horizontally across the top;
Time is represented vertically so models are read top to
bottom;
Interactions are represented by labelled arrows, Different
styles of arrow represent different types of interaction;
A thin rectangle in an object lifeline represents the time
when the object is the controlling object in the system.
C) State Machines Models
• Show how objects respond to different service requests
and the state transitions triggered by these requests
– If object state is Shutdown then it responds to a Startup()
message;
– In the waiting state the object is waiting for further messages;
– If reportWeather () then system moves to summarising state;
– If calibrate () the system moves to a calibrating state;
– A collecting state is entered when a clock signal is received.
Weather station state diagram
If object state is Shutdown then it responds to a Startup() message;
In the waiting state the object is waiting for further messages;
If reportWeather () then system moves to summarising state;
If calibrate () the system moves to a calibrating state;
A collecting state is entered when a clock signal is received.
2.5 Object interface specification
2.1 Define the context and modes of use of the system;
2.2 Design the system architecture;
2.3 Identify the principal system objects;
2.4 Develop design models;
2.5 Specify object interfaces.
• Object interfaces have to be specified so that the
objects and other components can be designed in
parallel.
• Designers should avoid designing the interface
representation and should hide this in the object itself.
• Objects may have several interfaces which are
viewpoints on the methods provided.
• The UML uses class diagrams for interface specification
Weather station interface
interface We atherStation {
public void WeatherStation () ;
public void startup () ;
public void startup (Instrument i) ;
public void shutdown () ;
public void shutdown (Instrument i) ;
public void reportWeather ( ) ;
public void test () ;
public void test ( Instrument i ) ;
public void calibrate ( Instrument i) ;
public int getID () ;
} //WeatherStation
3- Design evolution
• Hiding information inside objects means that
changes made to an object do not affect other
objects in an unpredictable way.
• Assume pollution monitoring facilities are to be
added to weather stations. These sample the
air and compute the amount of different
pollutants in the atmosphere.
• Pollution readings are transmitted with weather
data.
Changes required
• Add an object class called Air quality as part of
WeatherStation.
• Add an operation reportAirQuality to
WeatherStation. Modify the control software to
collect pollution readings.
• Add objects representing pollution monitoring
instruments.
Pollution monitoring
Weath erStation
id en tifier
repo r tWeath er ()
repo r tAirQu ality ()
calib rate (ins tru ments)
test ()
star tu p (in strumen ts)
sh utdo wn (ins tru ments)
Air q uality
NOData
smo k eData
b enz en eData
co llect ()
su mmaris e ()
Po llu tio n mo nitoring in strumen ts
NOmeter
Smo keMeter
Ben zen eMeter
Key points
• OOD is an approach to design so that design
components have their own private state and
operations.
• Objects should have constructor and inspection
operations. They provide services to other objects.
• Objects may be implemented sequentially or
concurrently.
• The Unified Modeling Language provides different
notations for defining different object models.
Key points
• A range of different models may be produced
during an object-oriented design process. These
include static and dynamic system models.
• Object interfaces should be defined precisely
using e.g. a programming language like Java.
• Object-oriented design potentially simplifies
system evolution.
Fortune 2008: Top 20 Most Admired Companies
1 Apple
2 Berkshire Hathaway
3 General Electric
4 Google
5 Toyota Motor
6 Starbucks
7 FedEx
8 Procter & Gamble
9 Johnson & Johnson
10 Goldman Sachs Group
11 Target
12 Southwest Airlines
13 American Express
14* BMW
14* Costco Wholesale
16 Microsoft
17 United Parcel Service
18 Cisco Systems
19 3M
20 Nordstrom
Keys:
Innovation
People management
Use of corporate assets
Social responsibility
Quality of management
Financial soundness
Long-term investment
Quality of products/services
Best Cities for Jobs
5. Fort Worth, Texas
4. Atlanta, Ga.,
3. Austin, Texas
2. Witchita, Kansas
1. Salt Lake City, Utah