Smalltalk and Object Orientation: an Introduction - Free

More documents

Recommendations

Info

23. Method and Class Testing 23.1 Introduction In the last chapter we discussed the problems facing the tester of an object oriented system (and in particular a Smalltalk system). We also briefly considered some approaches which overcome these problems. In this chapter we will consider the current best practice in testing object oriented systems. 23.1.1 Class and instance sides As was stated in the last chapter, the basic unit of test in Smalltalk is the class. However, there are two sides to a class, one is referred to as the class side and the other the instance side. The class side can be tested directly by send ing messages to the class. However, the instance side can only be tested by creating instances of the class. That is, although you define the instances’ methods in the class, you must test them using an instance. An important point to remember is that, it may be necessary to use both the class and instances of the class together, to adequately test both the class side and the instance side. For example, let us assume we have a class such as that illustrated in Figure 23.1. Clas side Methods: currentLinkType ^XLinkType Instance side Instance variables myLink Methods: createLink myLink := self class currentLinkType new. Figure 23.1: Instance and Class method interaction In this situation, it is possible to test the class side method currentLinkType directly by sending a message to the class an d in isolation. However, it is not possible to test the instance method createLink without considering the range of values which may be returned by currentLinkType. In this case, it is the class XLinkType which is returned, however, it could easily be one of a range of values. In addition, the class to which the currentLinkType message is sent, depends on the class of the object in which the createLink method is executing. As the createLink method may be inherited by other classes, we cannot guarantee which class will receive the currentLinkType message. 23.1.2 Testing methods in a class Each method in a class should first be tested in isolation. However, once all the individual methods in the class have been tested, threads through the meth ods in the class should be identified. This can be done by postulating scenarios of normal and exceptional usage (which may have been produced when the class was being designed). By tracing the result of these scenarios through the class it is possible to identify threads of execution amongst the methods in the class. An example of a thread of execution is illustrated in Figure 23.2. This thread was obtained by considering a scenario in which a person object has a birthd ay. This leads to the person object being sent the message birthday. The associated method birthday is presented below: birthday Transcript show: 'A happy birthday to ' , self name. self incrementAge. Transcript show: 'I am now ' , self age. 190
From the scenario that a person might have a birthday and therefore receive the message birthday, and from examining the source code of birthday (and incrementAge) we can obtain the thread in Figure 23.2. birthday incrementAge age age: age instance variable Figure 23.2: A thread of execution through an object 23.1.3 Object state Another point to bear in mind when testing a class, is that both instances and classes have state. For example, a class may have a class in stance variable which is used to keep a record of the number of instances which have been created of that class. This information may be used purely for book keeping or may be used to limit the number of instances created. In the latter case, a test needs to be performed which sets the counter to its limit and then a new instance should be requested. This could be done, either by creating the required number of instances, or by manually setting the counter to the appropriate value (e.g. by using the debugger or inspector). It is also important to remember that although a method has been tested in the context of a state instance, it is possible that some unintended sequence of message sends could result in the method being executed with the object in an unin tended state. It is therefore not possible to test individual methods in isolation thoroughly, nor is it sensible to ignore the state of the object when performing these tests. It is of course quite possible to ignore the state, however the tester does so at their peril. 23.1.4 Private methods Some methods are defined such that the user of the class is never expected to see them. This may be because they do some housekeeping which is local to the object (for example the initialization method) or because they provide facilities which are relevant to the internals of the object and not to the external interface to the object In Smalltalk, this method hiding is not enforced, it is usually left implicit (unlike C++ or Java in which the equivalent of privat e methods really are private and are not available as external messages). For example, in VisualWorks Smalltalk this can be done by placing the appropriate set of methods in a protocol entitled private or private-*. The convention is that a user of this class is not expected to use or to need to know about these methods. This is useful from a testing point of view as the tester can test the individual execution of these methods in a black box fashion. They can also use the approaches described above to perf orm white box testing. However, it also means that a user of the class, if they so wish, can call these methods. As this is not what is intended it is arguably reasonable to ignore this scenario. However, it may be wise in some situations at least to know what the effects will be and whether the object’s methods will respond in a meaningful or at least well defined manner. For example, it may be useful to know what will happen if another object sends the message initialize to the current object at some point after the object has started to be used. 23.1.5 Pool variables An additional problem with testing classes (as opposed to unit testing in other languages) is the presence of pool variables. These are rather like limited scope global variables. They are common to a set of classes and may affect the operation of these classes. As any one of the classes associated with the pool can change their value it is possible for one class (or its instances) to have a profound effect on the operation of a completely different class or its instances elsewhere in the system. This is something which requires careful thought. 191
Page 1 and 2:
Smalltalk and Object Orientation: A
Page 3 and 4:
Chapter 7: Smalltalk Constructs Thi
Page 5 and 6:
The user interface construction fac
Page 7 and 8:
Preface 1. INTRODUCTION TO OBJECT O
Page 9 and 10:
12.5 A CLASS SIDE METHOD...........
Page 11 and 12:
27.3 VISUALWORKS WINDOW PAINTING TO
Page 13:
Part One Introduction to Object Ori
Page 16 and 17:
interfaces between different parts
Page 18 and 19:
een to ease the transition from one
Page 20 and 21:
encapsulation. Attempting to introd
Page 22 and 23:
is a controversial subject which is
Page 24 and 25:
will become second nature to many o
Page 26 and 27:
or return something. It has since b
Page 28 and 29:
defining the behavior of different
Page 30 and 31:
In time most developers start to ac
Page 32 and 33:
2.4.3.4 Smalltalk environments are
Page 34 and 35:
3. Constructing an Object Oriented
Page 36 and 37:
Wash wipe switch Wiper motor Relay
Page 38 and 39:
Point 2 above is more complicated.
Page 40 and 41:
services working? working? working?
Page 42:
Part Two The Smalltalk Language 42
Page 45 and 46:
other systems. As the browsers, ins
Page 47 and 48:
VisualAge (a version of Smalltalk w
Page 49 and 50:
4.5.2 The VisualWorks Launcher and
Page 51 and 52:
Class categories. These are group s
Page 53 and 54:
5. A Little Smalltalk 5.1 Introduct
Page 55 and 56:
Notifiers can be displayed under a
Page 57 and 58:
These concepts of messages, receive
Page 59 and 60:
Select the “Debug” option. This
Page 61 and 62:
6. Smalltalk Building Blocks 6.1 In
Page 63 and 64:
6.3.1 Class definitions In Smalltal
Page 65 and 66:
Object Class1 Class 2 method search
Page 67 and 68:
“Now calculate y “ y := x * 23.
Page 69 and 70:
7. Smalltalk Constructs 7.1 Introdu
Page 71 and 72:
They are more efficient for storage
Page 73 and 74:
7.5.3 true, false and nil variables
Page 75 and 76:
anArray at: 1 max: 10. This will tr
Page 77 and 78:
Figure 8.1: Defining a new class 8.
Page 79 and 80:
this convention at the moment, howe
Page 81 and 82:
9. Control and Iteration 9.1 Introd
Page 83 and 84:
9.3.3 Block temporary variables Blo
Page 85 and 86:
The brackets round the 1 to: 10 are
Page 87 and 88:
OrderedCollection class (respective
Page 89 and 90:
holding other objects. Typically, c
Page 91 and 92:
| temp | temp := Set new. temp add:
Page 93 and 94:
Next we will define a message proto
Page 95 and 96:
11. Further Collection Classes 11.1
Page 97 and 98:
for (i = 0; i
Page 99 and 100:
x inspect. This has some similariti
Page 101 and 102:
11.9.3 The reject: message The reje
Page 103 and 104:
• asSortedCollection. This create
Page 105 and 106:
Figure 12.2: The Organizer class de
Page 107 and 108:
12.6 The private-updating protocol
Page 109 and 110:
do: [:item | Transcript show: item
Page 111 and 112:
13. Streams and Files 13.1 Introduc
Page 113 and 114:
Notice that in this example we used
Page 115 and 116:
to reconstruct the data into object
Page 117 and 118:
1. The ability to save the current
Page 119 and 120:
14.3 The class Date Class Date, a s
Page 121 and 122:
aCharacter isAlphabetic. aCharacter
Page 123 and 124:
Breakpoints may be placed in a me t
Page 125 and 126:
If the file does not exist you will
Page 127 and 128:
Figure 15.4: Creating a project To
Page 129 and 130:
Part Three Object Oriented Design 1
Page 131 and 132:
16. Object Oriented Analysis and De
Page 133 and 134:
devolved amongst the objects). This
Page 135 and 136:
driving force behind the whole of t
Page 137 and 138:
17. The Unified Modeling Language 1
Page 139 and 140: Deployment diagrams . Deployment di
Page 141 and 142: Window abstract +size : Area = (10,
Page 143 and 144: Micro Computer 1.. * Monitor System
Page 145 and 146: connectedTo * Lamp * Weight Sensor
Page 147 and 148: 18. UML: Dynamic Modeling And Deplo
Page 149 and 150: edisplay() Controller window A Wind
Page 151 and 152: 18.2.3.1 A start point A start poin
Page 153 and 154: etween nodes and their stereotype.
Page 155 and 156: OMT is a prescriptive method, it pr
Page 157 and 158: confirming the users’ needs, thei
Page 159 and 160: 19.3.2.2 Prepare a data dictionary
Page 161 and 162: car, truck and bus under a supercla
Page 163 and 164: 20.2.1.3 Prepare sequence and colla
Page 165 and 166: 20.2.2.5 Specify optimization crite
Page 167 and 168: are therefore concurrent and can be
Page 169 and 170: 20.3.2.3 Design optimization The an
Page 171 and 172: 21. Frameworks and Patterns for Obj
Page 173 and 174: As the framework takes shape you sh
Page 175 and 176: about a set of classes, than to att
Page 177 and 178: instanceVariableNames: '' classVari
Page 179 and 180: 21.5.5 Running the editor You have
Page 181 and 182: Part Four Testing and Style 181
Page 183 and 184: made worse, because the symbol held
Page 185 and 186: 22.3.2 Incremental testing Incremen
Page 187 and 188: control the execution of the method
Page 189: 22.7.2 The Perform: message The per
Page 193 and 194: 23.4 Self testing An approach, sugg
Page 195 and 196: 24. Smalltalk Style Guidelines 24.1
Page 197 and 198: 24.2.4 Adding variables to system c
Page 199 and 200: should use a dictionary to hold the
Page 201 and 202: • A method should only send messa
Page 203 and 204: Table 24.2: Class message categorie
Page 205 and 206: 25. The Perform and Dependency Mech
Page 207 and 208: • Inheritance (class to class rel
Page 209 and 210: You can find the methods which impl
Page 211 and 212: These messages are sent automatical
Page 213 and 214: 26. The Model-View-Controller Archi
Page 215 and 216: and wit hdrawals and requests for t
Page 217 and 218: Figure 26.5: Part of the Controller
Page 219 and 220: model 4. update 1. balance (view) 0
Page 221 and 222: displayed. The second component add
Page 223 and 224: 27. Graphical User Interface Constr
Page 225 and 226: 27.3.1 Palette The Palette tool pro
Page 227 and 228: uild a running user interface from
Page 229 and 230: names) and the actions for buttons
Page 231 and 232: Edit button, then the user interfac
Page 233 and 234: 27.11 Visual reuse The aim of visua
Page 235 and 236: Figure 27.17: Defining the parent c
Page 237 and 238: important, as these facilities hand
Page 239 and 240: 28.3 The AddressBook class 28.3.1 T
Page 241 and 242:
The aspect methods can then be simp
Page 243 and 244:
the name is in the address book, it
Page 245 and 246:
29.2 The custom view widget 29.2.1
Page 247 and 248:
DrawingView model SmallDraw drawing
Page 249 and 250:
The action of adding a new graphic
Page 251 and 252:
The next method we shall consider i
Page 253 and 254:
In the updating protocol we need to
Page 255 and 256:
The displaying protocol contains th
Page 257 and 258:
30. Memory Management and Garbage C
Page 259 and 260:
30.3.3 PermSpace This is used to ho
Page 261 and 262:
31. Concurrency in Smalltalk 31.1 I
Page 263 and 264:
The Semaphore class provides facili
Page 265 and 266:
Once the process wakes up it can de
Page 267 and 268:
31.4.3 Acknowledgments Acknowledgme
Page 269 and 270:
“There are two kinds of objects i
Page 271 and 272:
Object Object class Collection Coll
Page 273 and 274:
ehavior, but from the point of view
Page 275 and 276:
32.5.5 The whole enchilada Behaviou
Page 278 and 279:
33. The Future for Object Technolog
Page 280 and 281:
solutions are needed which suit the
Page 282 and 283:
scripting and query language to use
Page 284 and 285:
34. Appendix: The Smalltalk Languag
Page 286 and 287:
A capitalization convention is used
Page 288 and 289:
• marries: is the message selecto
Page 290 and 291:
Array OrderedCollection SortedColle
Page 292 and 293:
[Coleman et al 1994] D. Coleman, P.
Page 294 and 295:
OOPSLA/ECOOP’90, Joint Conference
Page 296 and 297:
detect.............................
Page 298:
True...............................
show all

Smalltalk and Object Orientation: an Introduction - Free

Create successful ePaper yourself

Delete template?

Save as template?