Pharo by Example 5

Pharo by Example 5
Stéphane Ducasse, Dimitris Chloupis, Nicolai Hess, Dmitri Zagidulin, Andrew
P. Black, Oscar Nierstrasz, Damien Pollet, Damien Cassou, Marcus Denker
To cite this version:
Stéphane Ducasse, Dimitris Chloupis, Nicolai Hess, Dmitri Zagidulin, Andrew P. Black, et al.. Pharo
by Example 5. Lulu.com & Square Bracket Associates, pp.1-358, 2017, 978-1-365-65459-6. <hal01659495>
HAL Id: hal-01659495
https://hal.inria.fr/hal-01659495
Submitted on 8 Dec 2017
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Distributed under a Creative Commons ttribution - ShareAlike�ttribution - ShareAlike 4.0
International License
Pharo by Example 5
Stéphane Ducasse
Dimitris Chloupis
Nicolai Hess
Dmitri Zagidulin
January 14, 2017
commit 7a4b6fc*
Copyright 2016 by Stéphane Ducasse, Dimitris Chloupis, Nicolai Hess and Dmitri
Zagidulin. Copyright 2009 by Andrew P. Black, Stéphane Ducasse, Oscar Nierstrasz,
Damien Pollet, Damien Cassou and Marcus Denker. The contents of this book are
protected under the Creative Commons Attribution-ShareAlike 3.0 Unported license.
You are free:
• to Share: to copy, distribute and transmit the work,
• to Remix: to adapt the work,
Under the following conditions:
Attribution. You must attribute the work in the manner specified by the author or
licensor (but not in any way that suggests that they endorse you or your use of
the work).
Share Alike. If you alter, transform, or build upon this work, you may distribute the
resulting work only under the same, similar or a compatible license.
For any reuse or distribution, you must make clear to others the license terms of this
work. The best way to do this is with a link to this web page:
http://creativecommons.org/licenses/by-sa/3.0/
Any of the above conditions can be waived if you get permission from the copyright
holder. Nothing in this license impairs or restricts the author’s moral rights.
Your fair dealing and other rights are in no way affected by the above. This is a humanreadable summary of the Legal Code (the full license):
http://creativecommons.org/licenses/by-sa/3.0/legalcode
[ ]
|
&
|
Published by Lulu.com & Square Bracket Associates.
http://squarebracketassociates.org
ISBN 978-1-365-65459-6
First Edition, January 2017.
Book layout and typography based on the sbabook LATEX class by Damien Pollet.
Contents
Contents
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
2.11
2.12
2.13
2.14
3
3.1
3.2
3.3
3.4
3.5
3.6
3.7
Preface
What is Pharo? . . . . . . . . . .
Who should read this book? . . .
A word of advice . . . . . . . . .
An open book . . . . . . . . . .
The Pharo community . . . . . .
Examples and exercises . . . . .
Acknowledgments . . . . . . . .
Hyper special acknowledgments
i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
2
3
3
3
4
4
4
A quick tour of Pharo
Installing Pharo . . . . . . . . . . . . . . . .
Pharo: File components . . . . . . . . . . . .
Launching Pharo . . . . . . . . . . . . . . . .
PharoLauncher . . . . . . . . . . . . . . . .
The World Menu . . . . . . . . . . . . . . . .
Sending messages . . . . . . . . . . . . . . .
Saving, quitting and restarting a Pharo session
Playgrounds and Transcripts . . . . . . . . .
Keyboard shortcuts . . . . . . . . . . . . . .
The System Browser . . . . . . . . . . . . . .
Finding classes . . . . . . . . . . . . . . . . .
Finding methods . . . . . . . . . . . . . . . .
Defining a new method . . . . . . . . . . . .
Chapter summary . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
7
8
9
10
11
13
15
15
16
19
22
25
27
31
A first application
The Lights Out game . . . . . . . .
Creating a new Package . . . . . .
Defining the class LOCell . . . . .
Adding methods to a class . . . . .
Inspecting an object . . . . . . . .
Defining the class LOGame . . . . .
Organizing methods into protocols
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
34
34
35
37
39
41
46
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
i
Contents
3.8 Finishing the game . . . . . .
3.9 Let’s try our code . . . . . . .
3.10 Saving and sharing Pharo code
3.11 Chapter summary . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
47
49
51
56
4
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
Syntax in a nutshell
Syntactic elements . . . . . . . . .
Pseudo-variables . . . . . . . . .
Message sends . . . . . . . . . .
Method syntax . . . . . . . . . . .
Block syntax . . . . . . . . . . . .
Conditionals and loops in a nutshell
Primitives and pragmas . . . . . .
Chapter summary . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
57
57
60
61
62
63
64
66
67
5
5.1
5.2
5.3
5.4
5.5
5.6
5.7
Understanding message syntax
Identifying messages . . . . . . . . .
Three types of messages . . . . . . .
Message composition . . . . . . . . .
Hints for identifying keyword messages
Expression sequences . . . . . . . . .
Cascaded messages . . . . . . . . . .
Chapter summary . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
69
69
71
74
79
81
81
82
The Pharo object model
The rules of the model . . . . . . . . . . .
Everything is an Object . . . . . . . . . . .
Every object is an instance of a class . . . .
Instance structure and behavior . . . . . . .
The instance side and the class side . . . . .
Every class has a superclass . . . . . . . . .
Everything happens by sending messages .
Method lookup follows the inheritance chain
Shared variables . . . . . . . . . . . . . . .
Internal object implementation note . . . .
Chapter summary . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
85
85
85
86
87
88
94
97
99
105
109
110
Some of the key tools of the Pharo environment
Pharo environment overview . . . . . . . . . . .
The main code browser . . . . . . . . . . . . . .
The inspector . . . . . . . . . . . . . . . . . . .
The debugger . . . . . . . . . . . . . . . . . . .
The process browser . . . . . . . . . . . . . . .
Finding methods . . . . . . . . . . . . . . . . . .
Chapter summary . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
113
113
115
128
130
139
140
140
6
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
6.10
6.11
7
7.1
7.2
7.3
7.4
7.5
7.6
7.7
ii
.
.
.
.
Contents
8
8.1
8.2
8.3
8.4
8.5
Sharing code and source control
Packages: groups of classes and methods
Source control . . . . . . . . . . . . . . .
The File List Browser . . . . . . . . . . . .
In Pharo, you can’t lose code . . . . . . .
Chapter summary . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
141
141
148
151
152
153
9
9.1
9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9
9.10
9.11
SUnit
Introduction . . . . . . . . .
Why testing is important . .
What makes a good test? . .
SUnit by example . . . . . .
The SUnit cookbook . . . . .
The SUnit framework . . . .
Advanced features of SUnit .
Continuing after a failure . .
SUnit implementation . . . .
A piece of advices on testing
Chapter summary . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
155
155
156
157
158
162
163
166
168
168
171
172
10
10.1
10.2
10.3
10.4
10.5
10.6
Basic classes
Object . . . . . .
Numbers . . . . .
Characters . . . .
Strings . . . . . .
Booleans . . . . .
Chapter summary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
175
175
184
188
189
190
192
11
11.1
11.2
11.3
11.4
11.5
11.6
11.7
Collections
Introduction . . . . . . . . . .
The varieties of collections . .
Collection implementations . .
Examples of key classes . . . .
Collection iterators . . . . . .
Some hints for using collections
Chapter summary . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 193
. 193
. 196
. 197
. 209
. 213
. 214
12
12.1
12.2
12.3
12.4
12.5
Streams
Two sequences of elements .
Streams vs. collections . . .
Streaming over collections .
Using streams for file access
Chapter summary . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 217
. 218
. 219
. 227
. 229
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
193
217
.
.
.
.
.
13 Morphic
13.1 The history of Morphic . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13.2 Manipulating morphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
231
231
233
iii
Contents
Composing morphs . . . . . . . . . . .
Creating and drawing your own morphs
Interaction and animation . . . . . . . .
Interactors . . . . . . . . . . . . . . . .
Drag-and-drop . . . . . . . . . . . . . .
A complete example . . . . . . . . . . .
More about the canvas . . . . . . . . .
Chapter summary . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
234
236
239
242
243
245
250
251
14
14.1
14.2
14.3
14.4
14.5
14.6
14.7
14.8
14.9
Seaside by example
Why do we need Seaside? . .
Getting started . . . . . . . .
Seaside components . . . .
Rendering HTML . . . . . . .
CSS: Cascading style sheets .
Managing control flow . . . .
A complete tutorial example
A quick look at AJAX . . . . .
Chapter summary . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
253
253
254
260
264
270
272
278
285
288
15
15.1
15.2
15.3
15.4
15.5
15.6
15.7
15.8
Classes and metaclasses
Rules for classes and metaclasses . . . . . . . . . . . . .
Revisiting the Pharo object model . . . . . . . . . . . . .
Every class is an instance of a metaclass . . . . . . . . .
The metaclass hierarchy parallels the class hierarchy . . .
Every metaclass inherits from Class and Behavior . . .
Every metaclass is an instance of Metaclass . . . . . . .
The metaclass of Metaclass is an instance of Metaclass
Chapter summary . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
291
291
292
293
295
297
300
300
302
16
16.1
16.2
16.3
16.4
16.5
16.6
16.7
16.8
16.9
Reflection
Introspection . . . . . . . . . . . . . . .
Browsing code . . . . . . . . . . . . . . .
Classes, method dictionaries and methods
Browsing environments . . . . . . . . . .
Accessing the run-time context . . . . . .
Intercepting messages not understood . .
Objects as method wrappers . . . . . . .
Pragmas . . . . . . . . . . . . . . . . . .
Chapter summary . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
303
304
309
311
313
315
318
322
325
326
17
17.1
17.2
17.3
17.4
17.5
Regular expressions in Pharo
Tutorial example — generating a site map
Regex syntax . . . . . . . . . . . . . . . .
Regex API . . . . . . . . . . . . . . . . .
Implementation Notes by Vassili Bykov . .
Chapter Summary . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 330
. 338
. 345
. 351
. 351
13.3
13.4
13.5
13.6
13.7
13.8
13.9
13.10
iv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
329
CHAPTER
1
Preface
1.1
What is Pharo?
Pharo is a modern, open-source, dynamically typed language supporting live
coding inspired by Smalltalk. Pharo and its ecosystems is composed of the six
fundamental elements:
• A dynamically-typed language with a syntax so simple it can fit on a
postcard and yet is readable even for someone not familiar with it.
• A live coding environment that allows the coder to modify its code while
the code executes, without any need to slow down his or her workflow.
• A powerful IDE providing all the tools to help manage complex code and
promote good code design.
• A rich library that creates an environment so powerful that it can be
viewed even as a virtual OS, including a very fast JITing VM and full
access to OS libraries and features via its FFI.
• A culture where changes and improvements are encouraged and highly
valued.
• A community that welcomes coders from any corner of the world with
any skill and any programming languages.
Pharo strives to offer a lean, open platform for professional software development, as well as a robust and stable platform for research and development
into dynamic languages and environments. Pharo serves as the reference
implementation for the Seaside web development framework available at
http://www.seaside.st.
1
Preface
Pharo core contains only code that has been contributed under the MIT
license. The Pharo project started in March 2008 as a fork of Squeak (a modern
implementation of Smalltalk-80), and the first 1.0 beta version was released on
July 31, 2009. The current version is Pharo 5.0, released in May 2016. Pharo 6.0
is in alpha development and planned for release in April 2017.
Pharo is highly portable. Pharo can run on OS X, Windows, Linux, Android,
iOS, and Raspberry Pi. Its virtual machine is written entirely in a subset
of Pharo itself, making it easy to simulate, debug, analyze, and change from
within Pharo itself. Pharo is the vehicle for a wide range of innovative projects,
from multimedia applications and educational platforms to commercial web
development environments.
There is an important principle behind Pharo: Pharo doesn’t just copy the
past, but it reinvents the essence behind Smalltalk. However we realize that
Big Bang style approaches rarely succeed. Pharo instead favors evolutionary
and incremental changes. Rather than leaping for the final perfect solution
in one big step, a multitude of small changes keeps even the bleeding edge
relatively stable while experimenting with important new features and libraries. This facilitates rapid feedback and contributions from the community,
on which Pharo depends for its success. Finally Pharo is not read-only, Pharo
integrates changes made by the community, daily. Pharo got 100 contributors
from all over the world. You can have an impact on Pharo, too!
1.2
Who should read this book?
The previous revision of this book was based on Pharo 1.4. This revision has
been liberally updated to align with Pharo 5.0. Various aspects of Pharo are
presented, starting with the basics then proceeding to intermediate topics.
Advanced topics are presented in Deep into Pharo, a book on the internals
of Pharo that is freely available at http://books.pharo.org. In addition for
readers interested in web development, a new book Enterprise Pharo: a Web
Perspective is freely available at http://books.pharo.org.
This book will not teach you how to program. The reader should have some
familiarity with programming languages. Some background with objectoriented programming would also be helpful.
This book will introduce the Pharo programming environment, the language
and the associated tools. You will be exposed to common idioms and practices,
but the focus is on the technology, not on object-oriented design. Wherever
possible, we will show you lots of examples.
There are numerous other books on Smalltalk freely available on the web at
http://stephane.ducasse.free.fr/FreeBooks.html.
2
1.3
1.3
A word of advice
A word of advice
Do not be frustrated by parts of Pharo that you do not immediately understand. You do not have to know everything! Alan Knight expresses this as
follows:
Try not to care. Beginning Smalltalk programmers often have
trouble because they think they need to understand all the details
of how a thing works before they can use it. This means it takes
quite a while before they can master Transcript show: ’Hello
World’. One of the great leaps in OO is to be able to answer the
question “How does this work?” with ”I don’t care”.
When you don’t understand something, simple or complex, do not hesitate
for a second to ask us at our mailing lists (pharo-users@lists.pharo.org or
pharo-dev@lists.pharo.org), irc and Slack. We love questions and we welcome
people of any skill.
1.4
An open book
This book is an open book in the following senses:
• The content of this book is released under the Creative Commons
Attribution-ShareAlike (by-sa) license. In short, you are allowed to
freely share and adapt this book, as long as you respect the conditions
of the license available at the following URL http://creativecommons.org/
licenses/by-sa/3.0/.
• This book just describes the core of Pharo. We encourage others to
contribute chapters on the parts of Pharo that we have not described. If
you would like to participate in this effort, please contact us. We would
like to see more books around Pharo!
• It is also possible to contribute directly to this book via Github. Just follow the instructions there and ask any question on the mailing list. You
can find the Github repo at https://github.com/SquareBracketAssociates/
UpdatedPharoByExample
1.5
The Pharo community
The Pharo community is friendly and active. Here is a short list of resources
that you may find useful:
• http://www.pharo.org is the main web site of Pharo.
• On IRC, you can find us on the freenode.net server, channel ”pharo”.
3
Preface
• SmalltalkHub (http://www.smalltalkhub.com/) is the equivalent of SourceForge/Github for Pharo projects. Many extra packages and projects for
Pharo live there.
• Pharo is also active on Slack - a platform for chat based on IRC (http:
//pharoproject.slack.com), just ask for an invitation at http://slackinvites.
pharo.org. Everybody is welcomed.
1.6
Examples and exercises
We have tried to provide as many examples as possible. In particular, there
are many examples that show a fragment of code which can be evaluated.
We use a long arrow to indicate the result you obtain when you select an
expression and from its context menu choose print it:
3 + 4
>>> 7 "if you select 3+4 and 'print it', you will see 7"
In case you want to play with these code snippets in Pharo, you can download
a plain text file with all the example code from the Resources sidebar of the
book’s web site: http://books.pharo.org.
1.7
Acknowledgments
We would like to thank Alan Kay, Dan Ingalls and their team for making
Squeak, an amazing Smalltalk development environment, that became the
open-source project from which Pharo took roots. Pharo also would not be
possible without the incredible work of the Squeak developers.
We would also like to thank Hilaire Fernandes and Serge Stinckwich who
allowed us to translate parts of their columns on Smalltalk, and Damien Cassou for contributing the chapter on Streams. We especially thank Alexandre
Bergel, Orla Greevy, Fabrizio Perin, Lukas Renggli, Jorge Ressia and Erwann
Wernli for their detailed reviews.
We thank the University of Bern, Switzerland, for graciously supporting this
open-source project and for hosting the web site of this book.
We also thank the Pharo community for their enthusiastic support of this
book project, as well as for all the translations of the first edition of Pharo by
Example.
1.8
Hyper special acknowledgments
We want to thank the original authors of this book! Without this initial version it would have been difficult to make this one. Pharo by Example is a
central book to welcome newcomers and it has a great value.
4
1.8
Hyper special acknowledgments
Thanks to Manfred Kröhnert, Markus Schlager, Werner Kassens, Michael
OKeefe, Aryeh Hoffman, Paul MacIntosh, Gaurav Singh, Jigyasa Grover, Craig
Allen, Serge Stinckwich, avh-on1, Yuriy Timchuk, zio-pietro for the typos
and feedback. Special thanks to Damien Cassou and Cyril Ferlicot for their
great help in the book update. Finally we want to thank Inria for its steady and
important financial support, and the RMoD team members for the constant
energy pushing Pharo forward.
Super special thanks to Damien Pollet for this great book template.
5
CHAPTER
2
A quick tour of Pharo
This chapter will take you on a high level tour of Pharo, to help you get comfortable with the environment. There will be plenty of opportunities to try
things out, so it would be a good idea if you have a computer handy when you
read this chapter.
In particular, you will fire up Pharo, learn about the different ways of interacting with the system, and discover some of the basic tools. You will also learn
how to define a new method, create an object and send it messages.
Note: Most of the introductory material in this book will work with any Pharo
version, so if you already have one installed, you may as well continue to use
it. However, since this book is written for Pharo 5.0, if you notice differences
between the appearance or behaviour of your system and what is described
here, do not be surprised.
2.1
Installing Pharo
Downloading Pharo
Pharo is available as a free download from http://pharo.org/download. Click
the button for your operating system to download the appropriate .zip
file. For example, the full Pharo 5.0 distribution for OS X will be available at
http://files.pharo.org/platform/Pharo5.0-mac.zip.
Once that file is unzipped, it will contain everything you need to run Pharo
(this includes the VM, the image, and the sources, as explained below).
Using handy scripts.
http://files.pharo.org/get/ offers a collection of scripts
7
A quick tour of Pharo
to download specific versions of Pharo. This is really handy to automate the
process.
To download the latest 50 full system, use the following snippet.
wget -O- get.pharo.org/50+vm | bash
Installing Pharo
Pharo does not need to install anything in your system, as it’s perfectly capable of running standalone. Depending on your platform, download the
appropriate zip file, uncompress it in a directory of your choice and now you
are ready to launch Pharo. In case of Ubuntu Linux, there is also the extra
option of installing Pharo via the Pharo PPA. Pharo can be also installed via
the command line.
2.2 Pharo: File components
Pharo consists of four main component files. Although you do not need to deal
with them directly for the purposes of this book, it is useful to understand the
roles they play.
1. The virtual machine (VM) is the only component that is different for each
operating system. The VM is the execution engine (similar to a JVM). It takes
Pharo bytcode that is generated each time user compiles a piece of code,
converts it to machine code and executes it. Pharo comes with the Cog VM a
very fast JITing VM. The VM executable is named:
• Pharo.exe for Windows;
• pharo for Linux ; and
• Pharo for OSX (inside a package also named Pharo.app).
The other components below are portable across operating systems, and can
be copied and run on any appropriate virtual machine.
2. The sources file contains source code for parts of Pharo that don’t change
frequently. Sources file is important because the image file format stores
only objects including compiled methods and their bytecode and not their
source code. Typically a new sources file is generated once per major release
of Pharo. For Pharo 5.0, this file is named PharoV50.sources.
3. The changes file logs of all source code modifications (especially all the
changes you did while programming) since the .sources file was generated.
Each release provides a near empty file named for the release, for example
Pharo5.0.changes. This facilitates a per method history for diffs or reverting.
It means that even if you did not manage to save the image file on a crash or
you just forgot, you can recover your changes from this file. A changes file is
always coupled with a image file. They work in pair.
8
2.3 Launching Pharo
4. The image file provides a frozen in time snapshot of a running Pharo system. This is the file where all objects are stored and as such it’s a cross platform format. An image file contains the live state of all objects of the system
(including classes and compiled methods, since they are objects too) at a given
point. An image is a virtual object container. The file is named for the release
(like Pharo5.0.image) and it is synched with the Pharo5.0.changes file.
Image/Change pair
The .image and .changes files provided by a Pharo release are the starting point for a live environment that you adapt to your needs. As you work
in Pharo, these files are modified, so you need to make sure that they are
writable. Pay attention to remove the changes and image files from the list
of files to be checked by anti-viruses. The .image and .changes files are intimately linked and should always be kept together, with matching base
filenames. Never edit them directly with a text editor, as .images holds your
live object runtime memory, which indexes into the .changes files for the
source. It is a good idea to keep a backup copy of the downloaded .image and
.changes files so you can always start from a fresh image and reload your
code. However, the most efficient way for backing up code is to use a version
control system that will provide an easier and powerful way to back up and
track your changes.
Common setup
The four main component files above can be placed in the same directory,
but it’s a common practice to put the Virtual Machine and sources file in a
separate directory where everyone has read-only access to them.
Do whatever works best for your style of working and your operating system.
2.3 Launching Pharo
To start Pharo, do whatever your operating system expects: drag the .image
file onto the icon of the virtual machine, or double-click the .image file, or at
the command line type the name of the virtual machine followed by the path
to the .image file.
• On OS X, double click the Pharo5.0.app bundle in the unzipped download.
• On Linux, double click (or invoke from the command line) the pharo
executable bash script from the unzipped Pharo folder.
• On Windows, enter the unzipped Pharo folder and double click Pharo.exe.
9
A quick tour of Pharo
In general, Pharo tries to ”do the right thing”. If you double click on the VM, it
looks for an image file in the default location. If you double click on an .image
file, it tries to find the nearest VM to launch it with.
If you have multiple VMs installed on your machine, the operating system
may no longer be able to guess the right one. In this case, it is safer to specify
exactly which ones you meant to launch, either by dragging and dropping the
image file onto the VM, or specifying the image on the command line (see the
next section).
Launching Pharo via the command line
The general pattern for launching Pharo from a terminal is:
<Pharo executable> <path to Pharo image>
Linux command line.
For Linux, assuming that you’re in the unzipped
pharo5.0 folder:
./pharo shared/Pharo5.0.image
For OS X, assuming that you’re in the directory with the
unzipped Pharo5.0.app bundle:
OS X command line.
Pharo5.0.app/Contents/MacOS/Pharo
Pharo5.0.app/Contents/Resources/Pharo5.0.image
When using a Pharo bundle, you need to right-click on Pharo5.0.app and
select ’Show Package Contents’ to get access to the image. If you need this
often, just download a separated image/changes pair and drop that image into
the Pharo5.0.app.
Windows command line.
For Windows, assuming that you’re in the unzipped
Pharo5.0 folder:
Pharo.exe Pharo5.0.image
2.4
PharoLauncher
PharoLauncher is a tool that helps you download and manage Pharo images.
It is very useful for getting new versions of Pharo (as well as updates to the
existing versions that contain important bug fixes). It also gives you access to
images preloaded with specific libraries that make it very easy to use those
tools without having to manually install and configure them.
PharoLauncher can be found on SmalltalkHub at http://smalltalkhub.com/#!
/~Pharo/PharoLauncher together with installation instructions and download
10
2.5 The World Menu
Figure 2.1: PharoLauncher - GUI
links depending on your platform. PharoLauncher is basically composed of
two columns.
After installing PharoLauncher and opening it (like you would do for any
Pharo image), you should get a GUI similar to Figure 2.1.
• The left column lists images that live locally on your machine (usually in
a shared system folder). You can launch any local image directly (either
by double-clicking, or by selecting it and pressing the Launch button). A
right-click context menu provides several useful functions like copying
and renaming your images, as well as locating them on the file system.
• The right column lists Templates, which are remote images available for
download. To download a remote image, select it and click the Create
image button (located on the top right, next to the Refresh template
list button).
You can use your own local images with PharoLauncher, in addition to working with the images you donwloaded. To do so, simply ensure that your .image
and its associated .changes files are placed in a folder (with the same name as
your image) in your default image location. You can find the location in the
PharoLauncher settings.
2.5 The World Menu
Once Pharo is running, you should see a single large window, possibly containing some open playground windows (see Figure 2.2). You might notice a menu
bar, but Pharo mainly makes use of context-dependent pop-up menus.
11
A quick tour of Pharo
Figure 2.2: Clicking anywhere on the Pharo window background activates the
World Menu
Clicking anywhere on the background of the Pharo window will display the
World Menu, which contains many of the Pharo tools, utilities and settings.
At the top of the World Menu, you will see a list of several core tools in Pharo,
including the System Browser, the Playground, the Monticello package manager, and others. We will discuss them in more detail in the coming chapters.
Interacting with Pharo
Pharo offers three ways to interact with the system using a mouse or other
pointing device.
click (or left-click): this is the most often used mouse button, and is normally
equivalent to left-clicking (or clicking a single-mouse button without any
modifier key). For example, click on the background of the Pharo window to
bring up the World menu (Figure 2.2).
action-click (or right-click): this is the next most used button. It is used to
bring up a contextual menu that offers different sets of actions depending
on where the mouse is pointing (see Figure 2.3). If you do not have a multibutton mouse, then normally you will configure the control modifier key to
action-click with the mouse button.
meta-click: Finally, you may meta-click on any object displayed in the image
to activate the ”morphic halo”, an array of handles that are used to perform
operations on the on-screen objects themselves, such as inspecting or resizing
them (see Figure 2.4). If you let the mouse linger over a handle, a help balloon
12
2.6 Sending messages
Figure 2.3: Action Click (right click) brings the contextual menu.
Figure 2.4: Meta-Clicking on a window opens the Halos
will explain its function. In Pharo, how you meta-click depends on your operating system: either you must hold Shift-Ctrl or Shift-Alt (on Windows or
Linux) or Shift-Option (on OS X) while clicking.
2.6 Sending messages
In the Pharo window, click on an open space to open the World Menu, and
then select the Playground menu option. The Playground tool will open (you
may recognize it as the Workspace tool, from previous versions of Pharo). We
can use Playground to quickly execute Pharo code. Enter the following code
in it, then right click and select Do it:
ProfStef go.
This expression will trigger the Pharo tutorial (as shown in Figure 2.6). It is a
simple and interactive tutorial that will teach you the basics of Pharo.
Congratulations, you have just sent your first message! Pharo is based on
13
A quick tour of Pharo
Figure 2.5: Executing an expression is simple with the Do it menu item.
Figure 2.6: PharoTutorial is a simple interactive tutorial to learn about Pharo
the concept of sending messages to objects. The Pharo objects are like your
soldiers ready to obey once you send them a message they can understand. We
will see how an object can understand a message, later on.
If you talk to Pharoers for a while, you will notice that they generally do not
use expressions like call an operation or invoke a method, as developers do in
other programming languages. Instead they will say send a message. This
reflects the idea that objects are responsible for their own actions and that the
method associated with the message is looked up dynamically. When sending
a message to an object, the object, and not the sender, selects the appropriate
method for responding to your message. In most cases, the method with the
same name as the message is executed.
As a user you don’t need to understand how each message works, the only
thing you need to know is what the available messages are for the objects that
interest you. This way an object can hide its complexity, and coding can be
kept as simple as possible without losing flexibility.
14
2.7 Saving, quitting and restarting a Pharo session
How to find the available messages for each object is something we will explore later on.
2.7 Saving, quitting and restarting a Pharo session
You can exit Pharo at any point, by closing the Pharo window as you do any
other application window. Additionally you can use the World Menu and
select either Save and quit or Quit.
In any case, Pharo will display a prompt to ask you about saving your image.
If you do save your image and reopen it, you will see that things are exactly
as you left them. This happens because the image file stores all the objects
(edited text, window positions, added methods... of course since they are all
objects) that Pharo has loaded into your memory so that nothing is lost on
exit.
When you start Pharo for the first time, the Pharo virtual machine loads the
image file that you specified. This file contains a snapshot of a large number of
objects, including a vast amount of pre-existing code and programming tools
(all of which are objects). As you work with Pharo, you will send messages to
these objects, you will create new objects, and some of these objects will die
and their memory will be reclaimed (garbage-collected).
When you quit Pharo, you will normally save a snapshot that contains all of
your objects. If you save normally, you will overwrite your old image file with
the new snapshot. Alternatively, you may save the image under a new name.
As mentioned earlier, in addition to the .image file, there is also a .changes
file. This file contains a log of all the changes to the source code that you have
made using the standard tools. Most of the time you do not need to worry
about this file at all. As we shall see, however, the .changes file can be very
useful for recovering from errors, or replaying lost changes. More about this
later!
It may seem like the image is the key mechanism for storing and managing
software projects, but that is not the case. As we shall see soon, there are
much better tools for managing code and sharing software developed by
teams. Images are very useful, but you should learn to be very cavalier about
creating and throwing away images, since versioning tools like Monticello offer much better ways to manage versions and share code amongst developers.
In addition, if you need to persist objects, you can use several systems such
as Fuel (a fast object binary serializer), STON (a textual object serializer) or a
database.
2.8 Playgrounds and Transcripts
Let us start with some exercises:
15
A quick tour of Pharo
1. Close all open windows within Pharo.
2. Open a Transcript and a Playground/workspace. (The Transcript can be
opened from the World > Tools > ... submenu.)
3. Position and resize the transcript and playground windows so that the
playground just overlaps the transcript (see Figure 2.7).
You can resize windows by dragging one of the corners. At any time only one
window is active; it is in front and has its border highlighted.
The Transcript is an object that is often used for logging
system messages. It is a kind of system console.
About Transcript.
Playgrounds are useful for typing snippets of code that
you would like to experiment with. You can also use playgrounds simply
for typing any text that you would like to remember, such as to-do lists or
instructions for anyone who will use your image.
About Playground.
Type the following text into the playground:
Transcript show: 'hello world'; cr.
Try double-clicking at various points on the text you have just typed. Notice
how an entire word, entire string, or all of the text is selected, depending
on whether you click within a word, at the end of the string, or at the end of
the entire expression. In particular, if you place the cursor before the first
character or after the last character and double-click, you select the complete
paragraph.
Select the text you have typed, right click and select Do it. Notice how the
text ”hello world” appears in the transcript window (See Figure 2.7). Do it
again.
2.9 Keyboard shortcuts
If you want to evaluate an expression, you do not always have to right click.
Instead, you can use keyboard shortcuts shown in menu items. Even though
Pharo may seem like a mouse driven enviroment it contains over 200 shortcuts that allow you operate a variaty of tools, as well as the facility to assign a
keyboard shortcut to any of the 80000 methods contained in the Pharo image.
To have a look at the available shortcuts go to World Menu > System > Keymap
Browser.
Depending on your platform, you may have to press one of the modifier
keys which are Control, Alt, and Command. We will use CMD in the rest of
the book: so each time you see something like CMD-d, just replace it with the
appropriate modifier key depending on your OS. The corresponding modifier
16
2.9 Keyboard shortcuts
Figure 2.7: Executing an expresssion: displaying a string in the Transcript.
key in Windows is CTRL, and in Linux is either ALT or CTRL, so each time you
see something like CMD-d, just replace it with the appropriate modifier key
depending on your OS.
In addition to Do it, you might have noticed Do it and go, Print it, Inspect it and several other options in the context menu. Let’s have a quick
look at each of these.
Doing vs. printing
Type the expression 3 + 4 into the playground. Now Do it with the keyboard
shortcut.
Do not be surprised if you saw nothing happen! What you just did is send
the message + with argument 4 to the number 3. Normally the resulting 7
would have been computed and returned to you, but since the playground did
not know what to do with this answer, it simply did not show the answer. If
you want to see the result, you should Print it instead. Print it actually
compiles the expression, executes it, sends the message printString to the
result, and displays the resulting string.
Select 3+4 and Print it (CMD-p). This time we see the result we expect.
3 + 4
>>> 7
We use the notation >>> as a convention in this book to indicate that a particular Pharo expression yields a given result when you Print it.
17
A quick tour of Pharo
Figure 2.8: Inspecting a simple number using Inspect
Inspect
Select or place the cursor on the line of 3+4, and this time Inspect it (CMD-i).
Now you should see a new window titled ”Inspector on a SmallInteger(7)” as
shown in Figure 2.8. The inspector is an extremely useful tool that allows you
to browse and interact with any object in the system. The title tells us that 7 is
an instance of the class SmallInteger. The top panel allows us to browse the
instance variables of an object and their values. The bottom panel can be used
to write expressions to send messages to the object. Type self squared in the
bottom panel of the inspector, and Print it.
The inspector presents specific tabs that will show different information
and views on the object depending on the kind of object you are inspecting.
Inspect Morph new openInWorld you should get a situation similar to the one
of Figure 2.9.
Other operations
Other right-click options that may be used are the following:
• Do it and go additionally opens a navigable inspector on the side of
the playground. It allows us to navigate the object structure. Try with
the previous expression Morph new openInWorld and navigate the
structure.
• Basic Inspect it opens the classic inspector that offers a more minimal GUI and live updates of changes to the object.
• Debug it opens the debugger on the code.
18
2.10
The System Browser
Figure 2.9: Inspecting a Morph using Inspect
• Profile it profiles the code with the Pharo profile tool which shows
how much time is spent for each message sent.
• Code search offers several options provided by System Browser, such
as browsing the source code of an expression, searching for senders and
implementors, and so on.
2.10
The System Browser
The System Browser, also known as ”Class Browser”, is one of the key tools
used for programming. As we shall see, there are several interesting browsers
available for Pharo, but this is the basic one you will find in any image. The
current implementation of the System Browser is called Nautilus (this is the
name of the Jules Verne’s submarine).
Opening the System Browser on a given method
This is not the usual way that we open a browser on a method: we use more
advanced tools! Now for the sake of this presentation, execute the following
snippet:
Nautilus openOnMethod: Integer>>#factorial
It will open a system browser on the method factorial. We should get a System Browser like in Figure 2.10. The title bar indicates that we are browsing
the class Integer and its method factorial. Figure 2.10 shows the different
entities displayed by the browser: packages, classes, protocols, methods and
method definition.
19
A quick tour of Pharo
Figure 2.10: The System Browser showing the factorial method of class Inte-
ger
In Pharo, the default System Browser is Nautilus. However, it is possible
to have other System Browsers installed in the Pharo enviroment such as
AltBrowser. Each System Browser may have its own GUI that may be very
different from the Nautilus GUI. From now on, we will use the terms Browser,
System Browser and (occasionally) Nautilus interchangeably.
Navigating using the System Browser
Pharo has Spotter (see below) to navigate the system. Now we just want to
show you the working flow of the System Browser. Usually with Spotter we go
directly to the class or the method.
Let us look how to find the printString method defined in class Object. At
the end of the navigation, we will get the situation depicted in 2.11.
Open the Browser by selecting World > System Browser.
When a new System Browser window first opens, all panes but the leftmost are empty. This
first pane lists all known packages, which contain groups of related classes.
Type part of the name of the package in the left most filter.
It filters the list of packages to be shown in the list under it. Type ’Kern’ for
Filter packages.
20
2.10
The System Browser
Figure 2.11: The System Browser showing the printString method of class
Object
example.
Expand the Kernel package and select the Objects element.
When we select
a package, it causes the second pane to show a list of all of the classes in the
selected package. You should see the hierarchy of ProtoObject
Select the Object class. When a class is selected, the remaining two panes
will be populated. The third pane displays the protocols of the currently
selected class. These are convenient groupings of related methods. If no
protocol is selected you should see all methods in the fourth pane.
Select the printing protocol.
You may have to scroll down to find it. You
can also click on the third pane and type pr, to typeahead-find the printing
protocol. Now select it, and you will see in the fourth pane only methods
related to printing.
Select the printString method.
Now we see in the bottom pane the source
code of the printString method, shared by all objects in the system (except
those that override it).
There are much better way to find a method and we will look at them now.
21
A quick tour of Pharo
2.11
Finding classes
There are several ways to find a class in Pharo. The first, as we have just seen
above, is to know (or guess) what package it is in, and to navigate to it using
the browser.
A second way is to send the browse message to the class, asking it to open a
browser on itself. Suppose we want to browse the class Point.
Using the message browse
Type Point browse into a playground and Do it. A browser will open on the
Point class.
Using CMD-b to browse
There is also a keyboard shortcut CMD-b (browse) that you can use in any text
pane; select the word and press CMD-b. Use this keyboard shortcut to browse
the class Point.
Notice that when the Point class is selected but no protocol or method is
selected, instead of the source code of a method, we see a class definition. This
is nothing more than an ordinary message that is sent to the parent class,
asking it to create a subclass. Here we see that the class Object is being asked
to create a subclass named Point with two instance variables, class variables,
and to put the class Point in the Kernel-BasicObjects package. If you click
on the Comments button at the bottom of the class pane, you can see the class
comment in a dedicated pane.
In addition the system supports the following mouse shortcuts
• CMD-Click on a word: open the definition of a class when the word is
a class name. You get also the implementors of the message when you
click on a selector that is in the body of a method.
• CMD-Shift-Click on a word: open a list browser with all the refs of the
class when the word is a class name. You get also the senders of the
message when you click on a selector that is in the body of a method.
Using Spotter
The fastest (and probably the coolest) way to find a class is to use Spotter.
Pressing Shift+Enter opens Spotter, a very powerful tool for finding classes,
methods, and many other related actions. Figure 2.12 shows that we look for
Point.
Spotter offers several possibilities as shown in Figure 2.12. You can specify to
Spotter the kind of categories you are interested in. For example, using #class
22
2.11
Finding classes
Figure 2.12: Opening Spotter
followed by the word you look for, indicates that you are interested in classes.
This is the default so you do not need to type #class.
Figure 2.13 shows how we can ask Spotter to show all the implementors of
a given messages. We do not have to type the full category name. Other Categories are menu, packages, method (#implementor), examples (#example),
pragma (#pragma), senders (#sender), class references (#reference) but also
playground code snippets (using #playground).You can just type the beginning of the category to identify it i.e., #ref Point will give all the reference to
the class Point.
Spotter can be used even to browse through the OS file system, and has a
history category where previous searches are stored for quickly going back to
popular searches.
Navigating results
In addition we can use Spotter to navigate to our search results similarly
to how we use System Browser. Spotter categorizes its search results: for
example, classes are under Classes category, methods under the Implementors
category, help topics under Help Topics category, etc.
Clicking on the right arrow will take us to our selection and create a tab on top
23
A quick tour of Pharo
Figure 2.13: Looking for implementors matching printString
that we can click to go back to where we were. Depending on what we click on,
we step into our selection and are exposed to more categories.
For example, if our selection is the Point class, we will dive inside a group of
categories made for instance methods, class methods, super instance methods
etc.
The interface is fully controllable through the keyboard. The user can move
with Up/Down arrows between items or Cmd-Shift-Up/Cmd-Shift-Down arrows (note that on Windows and Linux Cmd key is the Alt key) through categories. At the same time, the search field has the focus, so the user can switch
seamlessly between selecting items and refining the search. Pressing Enter on
a selection opens the System Browser on that specific selected search result.
Using ’Find class’ in System Browser
In the SystemBrowser you can also search for a class via its name. For example, suppose that you are looking for some unknown class that represents
dates and times.
In the System Browser, click anywhere in the package pane or the class pane,
24
2.12 Finding methods
and launch the Class Search window by typing CMD-f CMD-c, or selecting Find
class (f,c) from the right-click context menu. Type time in the dialog box
and click OK (or press Enter).
A list of classes is displayed, whose names contain the substring time. Choose
one (say, Time), and the browser will show it, along with a class comment that
suggests other classes that might be useful. If you want to browse one of the
others, select its name (in any text pane), and type CMD-b.
Note that if you type the complete (and correctly capitalized) name of a class
in the find dialog, the browser will go directly to that class without showing
you the list of options.
Using the Finder
You can also open the Finder that is available from the World > Tools...
menu, and type part of the name of the class and change the Selectors to
Classes in the right combo box. It is less efficient than using Spotter or the
SystemBrowser as explained above. The Finder is more useful for other types
of code searches such as find methods based on examples, as we will show
later.
2.12
Finding methods
Sometimes you can guess the name of a method, or at least part of the name
of a method, more easily than the name of a class. For example, if you are
interested in the current time, you might expect that there would be a method
called ”now”, or containing ”now” as a substring. But where might it be?
Spotter and Finder can help you.
Spotter
With Spotter you can also find methods. Either by getting a class and navigating or using category such as:
• #implementor a method name will display all the methods that are
implemented and have the same name. For example you will get all the
do: methods.
• #selector a method name will display all the selectors that matches
this name
With Finder
Select World Menu > Tools > Finder. Type now in the top left search box,
click Search (or just press the Enter key). You should see a list of results
similar to the one in Figure 2.14.
25
A quick tour of Pharo
Figure 2.14: The Finder showing all classes defining a method named now.
The Finder will display a list of all the method names that contain the substring ”now”. To scroll to now itself, move the cursor to the list and type ”n”;
this type-ahead trick works in all scrolling windows. Expanding the ”now”
item shows you the classes that implement a method with this name. Selecting
any one of them will display the source code for the implementation in the
code pane on the bottom.
Finding methods using examples
At other times, you may have a good idea that a method exists, but will have
no idea what it might be called. The Finder can still help! For example, suppose that you would like to find a method that turns a string into upper case
(for example, transforming 'eureka' into 'EUREKA'). We can give the inputs
and expected output of a method and the Finder will try to find it for you.
The Finder has a really powerful functionality: you can give the receiver,
arguments and expected result and the finder tries to find the corresponding
message.
Trying Finder
In the Finder, select the Examples mode using the second combo-box (the one
that shows Selectors by default).
26
2.13 Defining a new method
Type 'eureka' . 'EUREKA' into the search box and press the Enter key.
The Finder will then suggest a method that does what you were looking for,
as well as display a list of classes that implement methods with the same
name. In this case, it determined that the asUppercase method is the one that
performed the operation that fit your example.
Click on the 'eureka' asUppercase --> 'EUREKA' expression, to show the
list of classes that implement that method.
An asterisk at the beginning of a line in the list of classes indicates that this
method is the one that was actually used to obtain the requested result. So,
the asterisk in front of String lets us know that the method asUppercase
defined in the class String was executed and returned the result we wanted.
The classes that do not have an asterisk are just other implementors of asUppercase, which share the method name but were not used to return the
wanted result. So the method Character>>asUppercase was not executed
in our example, because 'eureka' is not a Character instance (but is instead a
String).
You can also use the Finder to search for methods by arguments and results.
For example, if you are looking for a method that will find the greatest common factor of two integers, you might try 25 . 35 . 5 as an example. You
can also give the method finder multiple examples to narrow the search space;
the help text in the bottom pane explains how.
2.13
Defining a new method
The advent of Test Driven Development (TDD) has changed the way we write
code. The idea behind TDD is that we write a test that defines the desired
behaviour of our code before we write the code itself. Only then do we write
the code that satisfies the test.
Suppose that our assignment is to write a method that ”says something loudly
and with emphasis”. What exactly could that mean? What would be a good
name for such a method? How can we make sure that programmers who may
have to maintain our method in the future have an unambiguous description of what it should do? We can answer all of these questions by giving an
example.
Our goal is to define a new method named shout in the class String. The idea
is that this message should turn a string into its uppercase version as shown in
the example below:
'No panic' shout
>>> 'NO PANIC!'
However, before creating the shout method itself, we must first create a test
method! In the next section, we can use the ”No Panic” example to create our
test method.
27
A quick tour of Pharo
Figure 2.15: Defining a test method in the class StringTest.
Defining a new test method
How do we create a new method in Pharo? First, we have to decide which
class the method should belong to. In this case, the shout method that we are
testing will go in class String, so the corresponding test will, by convention,
go in a class called StringTest.
First, open a browser on the class StringTest, and select an appropriate
protocol for our method, in this case 'tests - converting'. The highlighted
text in the bottom pane is a template that reminds you what a Pharo method
looks like. Delete this template code (remember, you can either click on the
beginning or the end of the text, or press CMD-a, to ”Select All”), and start
typing your method. We can turn our ”No Panic” code example into the test
method itself:
testShout
self assert: ('No panic' shout = 'NO PANIC!')
Once you have typed the text into the browser, notice that the corner is orange. This is a reminder that the pane contains unsaved changes. So, select
Accept (s) by right clicking in the bottom pane, or just type CMD-s, to compile and save your method. You should see a situation similar to the one
depicted in Figure 2.15.
If this is the first time you have accepted any code in your image, you will
likely be prompted to enter your name. Since many people have contributed
code to the image, it is important to keep track of everyone who creates or
modifies methods. Simply enter your first and last names, without any spaces.
Because there is as yet no method called shout, the automatic code checker
(Quality Assitance) in the lower browser pane will inform you, that the mes28
2.13 Defining a new method
Figure 2.16: Looking at the error in the debugger.
sage shout is sent but not implemented. This can be quite useful if you have
merely made a typing mistake, but in this case, we really do mean shout, since
that is the method we are about to create. We confirm this by selecting the
first option from the menu of choices.
Running your test method
Run your newly created test: open the Test Runner from the World Menu (or
press on the circle icon in front of the method name this is faster and cooler).
In the Test Runner the leftmost two panes are a bit like the top panes in the
System Browser. The left pane contains a list of packages, but it’s restricted to
those packages that contain test classes.
Select CollectionsTests-Strings package, and the pane to the right will
show all of the test classes in it, which includes the class StringTest. Class
names are already selected, so click Run Selected to run all these tests.
You should see the upper right pane turn red, which indicates that there was
an error in running the tests. The list of tests that gave rise to errors is shown
in the bottom right pane. As you can see, the method StringTest>>testShout is the culprit. (Note that StringTest>>testShout is the Pharo way of
identifying the testShout method of the StringTest class.) If you click on
that method in the bottom right pane, the erroneous test will run again, this
time in such a way that you see the error happen: MessageNotUnderstood:
ByteString>>shout (see Figure 2.16).
29
A quick tour of Pharo
Figure 2.17: Pressing the Create button in the debugger prompts you to select in
which class to create the new method.
Figure 2.18: The automatically created shout method waiting for a real definition.
The window that opens with the error message is the Pharo debugger. We will
look at the debugger and how to use it in Chapter: The Pharo Environment.
Implementing the tested method
The error is, of course, exactly what we expected: running the test generates
an error because we haven’t yet written a method that tells strings how to
shout. Nevertheless, it’s good practice to make sure that the test fails because
this confirms that we have set up the testing machinery correctly and that the
new test is actually being run. Once you have seen the error, you can Abandon
the running test, which will close the debugger window.
Coding in the debugger
Instead of pressing Abandon, you can define the missing method using the
Create button right in the debugger. This will prompt you to select a class in
30
2.14
Chapter summary
which to define the new method (see Figure 2.17), then prompt you to select a
protocol for that method, and finally take you to a code editor window in the
debugger, in which you can edit the code for this new method. Note that since
the system cannot implement the method for you, it creates a generic method
that is tagged as to be implemented (see Figure 2.18).
Now let’s define the method that will make the test succeed! Right inside the
debugger edit the shout method with this definition:
shout
^ self asUppercase,'!'
The comma is the string concatenation operation, so the body of this method
appends an exclamation mark to an upper-case version of whatever String
object the shout message was sent to. The ^ tells Pharo that the expression
that follows is the answer to be returned from the method, in this case the
new concatenated string.
When you’ve finished implementing the method, do not forget to compile it
using CMD-s and you can press Proceed and continue with the tests. Note that
Proceed simply continues on running the test suite, and does not re-run the
failed method.
Let’s run the tests and see. Click on Run Selected
again in the Test Runner, and this time you should see a green bar and text
indicating that all of the tests ran with no failures and no errors. When you
get to a green bar, it’s a good idea to save your work by saving the image
(World Menu > Save), and take a break. So, do that right now!
Does this method work?
2.14
Chapter summary
This chapter has introduced you to the Pharo environment and shown you
how to use some of the major tools, such as the System Browser, Spotter,
the Finder, the Debugger, and the Test Runner. You have also seen a little of
Pharo’s syntax, even though you may not understand it all yet.
• A running Pharo system consists of a virtual machine, a .sources file,
and .image and .changes files. Only these last two change, as they
record a snapshot of the running system.
• When you open a Pharo image, you will find yourself in exactly the same
state (i.e., with exactly the same running objects) that you had when you
last saved that image.
• You can click on the Pharo background to bring up the World Menu and
launch various tools.
• A Playground is a tool for writing and evaluating snippets of code. You
can also use it to store arbitrary text.
31
A quick tour of Pharo
• You can use keyboard shortcuts on text in the playground, or any other
tool, to evaluate code. The most important of these are Do it (CMD-d),
Print it (CMD-p), Inspect it (CMD-i), and Browse it (CMD-b).
• The System Browser is the main tool for browsing Pharo code and for
developing new code.
• The Test runner is a tool for running unit tests, and aids in Test Driven
Development.
• The Debugger allows you to examine errors and exceptions (such as
errors or failures encountered when running tests). You can even create
new methods right in the debugger.
32
CHAPTER
3
A first application
In this chapter, we will develop a simple game: LightsOut (http://en.wikipedia.
org/wiki/Lights_Out_(game)). Along the way we will show most of the tools that
Pharo programmers use to construct and debug their programs, and show
how programs are shared with other developers. We will see the browser, the
object inspector, the debugger and the Monticello package browser.
In Pharo you can develop in a traditional way, by defining a class, then its instance variables, then its methods. However, in Pharo your development flow
can be much more productive than that! You can define instance variables and
methods on the fly. You can also code in the debugger using the exact context
of currently executed objects. This chapter will sketch such alternate way and
show you how you can be really productive.
Figure 3.1: The Lights Out game board
33
A first application
Figure 3.2: Create a Package and class template
3.1
The Lights Out game
To show you how to use Pharo’s programming tools, we will build a simple
game called Lights Out. The game board consists of a rectangular array of
light yellow cells. When you click on one of the cells, the four surrounding
cells turn blue. Click again, and they toggle back to light yellow. The object of
the game is to turn blue as many cells as possible.
Lights Out is made up of two kinds of objects: the game board itself, and 100
individual cell objects. The Pharo code to implement the game will contain
two classes: one for the game and one for the cells. We will now show you how
to define these classes using the Pharo programming tools.
3.2 Creating a new Package
We have already seen the browser in Chapter : A Quick Tour of Pharo where
we learned how to navigate to packages, classes and methods, and saw how to
define new methods. Now we will see how to create packages and classes.
From the World menu, open a System Browser. Right-click on an existing
package in the Package pane and select Add package... from the menu.
Type the name of the new package (we use PBE-LightsOut) in the dialog box
and click OK (or just press the return key). The new package is created, and
positioned alphabetically in the list of packages (see Figure 3.2).
Hints: You can type PBE in the filter to get your package filtered out the other
ones (See Figure 3.3).
34
3.3 Defining the class LOCell
Figure 3.3: Filtering our package to work more efficiently
3.3 Defining the class LOCell
At this point there are, of course, no classes in the new package. However, the
main editing pane displays a template to make it easy to create a new class
(see Figure 3.3).
This template shows us a Pharo expression that sends a message to a class
called Object, asking it to create a subclass called NameOfSubClass. The new
class has no variables, and should belong to the category (package) PBELightsOut.
Creating a new class
We simply edit the template to create the class that we really want. Modify the
class creation template as follows:
• Replace Object with SimpleSwitchMorph.
• Replace NameOfSubClass with LOCell.
• Add mouseAction to the list of instance variables.
You should get the following class definition:
SimpleSwitchMorph subclass: #LOCell
instanceVariableNames: 'mouseAction'
classVariableNames: ''
package: 'PBE-LightsOut'
This new definition consists of a Pharo expression that sends a message to the
existing class SimpleSwitchMorph, asking it to create a subclass called LOCell.
(Actually, since LOCell does not exist yet, we passed the symbol #LOCell as an
argument, representing the name of the class to create.) We also tell it that
instances of the new class should have a mouseAction instance variable, which
35
A first application
Figure 3.4: The newly-created class LOCell
we will use to define what action the cell should take if the mouse should click
on it.
At this point you still have not created anything. Note that the top right of
the panel changed to orange. This means that there are unsaved changes. To
actually send this subclass message, you must save (accept) the source code.
Either right-click and select Accept, or use the shortcut CMD-s (for ”Save”).
The message will be sent to SimpleSwitchMorph, which will cause the new
class to be compiled. You should get the situation depicted in Figure 3.4.
Once the class definition is accepted, the class is created and appears in the
class pane of the browser (see Figure 3.4). The editing pane now shows the
class definition. Below you get the Quality Assistant’s feedback: It runs automatically quality rules on your code and reports them.
About comments
Pharoers put a very high value on the readability of their code, but also good
quality comments.
People have the tendency to believe that it is not necessary to comment well written methods: it is plain wrong and encourages
sloppiness. Of course, bad code should renamed and refactored. Obviously
commenting trivial methods makes no sense. A comment should not be the
code written in english but an explanation of what the method is doing, its
Method comments.
36
3.4
Adding methods to a class
context, or the rationale behind its implementation. When reading a comment, the reader should be comforted that his hypotheses are correct.
For the class comment, the Pharo class comment template
gives a good idea of a strong class comment. Read it! It is based on CRC for
Class Responsibility Collaborators. So in a nutshell the comments state the
responsibility of the class in a couple of sentences and how it collaborates
with other classes to achieve this responsibilities. In addition we can state
the API (main messages an object understands), give an example (usually in
Pharo we define examples as class methods), and some details about internal
representation or implementation rationale.
Class comments.
Select the comment button and define a class comment following this template
On categories vs. packages
Historically, Pharo packages were implemented as ”categories” (a group
of classes). With the newer versions of Pharo, the term category is being
deprecated, and replaced exclusively by package.
If you use an older version of Pharo or an old tutorial, the class template will
be as follow:
SimpleSwitchMorph subclass: #LOCell
instanceVariableNames: 'mouseAction'
classVariableNames: ''
category: 'PBE-LightsOut'
It is equivalent to the one we mentioned earlier. In this book we only use the
term package. The Pharo package is also what you will be using to version
your source code using the Monticello versioning tool.
3.4
Adding methods to a class
Now let’s add some methods to our class. Select the protocol 'no messages'
in the protocol pane. You will see a template for method creation in the
editing pane. Select the template text, and replace it by the following (Do not
forget to compile it):
initialize
super initialize.
self label: ''.
self borderWidth: 2.
bounds := 0 @ 0 corner: 16 @ 16.
offColor := Color paleYellow.
onColor := Color paleBlue darker.
self useSquareCorners.
37
A first application
Figure 3.5: The newly-created method initialize
self turnOff
Note that the characters '' on line 3 are two separate single quotes with
nothing between them, not a double quote! '' denotes the empty string.
Another way to create an empty string is String new. Do not forget to accept
this method definition.
Notice that the method is called initialize. The name
is very significant! By convention, if a class defines a method named initialize, it is called right after the object is created. So, when we execute LOCell
new, the message initialize is sent automatically to this newly created object. initialize methods are used to set up the state of objects, typically to
set their instance variables; this is exactly what we are doing here.
Initialize methods.
The first thing that this method does (line
2) is to execute the initialize method of its superclass, SimpleSwitchMorph.
The idea here is that any inherited state will be properly initialized by the
initialize method of the superclass. It is always a good idea to initialize
inherited state by sending super initialize before doing anything else. We
don’t know exactly what SimpleSwitchMorph’s initialize method will do
(and we don’t care), but it’s a fair bet that it will set up some instance variables
to hold reasonable default values. So we had better call it, or we risk starting
in an unclean state.
Invoking superclass initialization.
38
3.5 Inspecting an object
The rest of the method sets up the state of this object. Sending self label:
'', for example, sets the label of this object to the empty string.
The expression 0@0 corner: 16@16
probably needs some explanation. 0@0 represents a Point object with x and
y coordinates both set to 0. In fact, 0@0 sends the message @ to the number
0 with argument 0. The effect will be that the number 0 will ask the Point
class to create a new instance with coordinates (0,0). Now we send this
newly created point the message corner: 16@16, which causes it to create a
Rectangle with corners 0@0 and 16@16. This newly created rectangle will be
assigned to the bounds variable, inherited from the superclass.
About point and rectangle creation.
Note that the origin of the Pharo screen is the top left, and the y coordinate
increases downwards.
The rest of the method should be self-explanatory. Part of
the art of writing good Pharo code is to pick good method names so that the
code can be read like a kind of pidgin English. You should be able to imagine
the object talking to itself and saying ”Self, use square corners!”, ”Self, turn
off!”.
About the rest.
Notice that there is a little green arrow next to your method (see Figure 3.5).
This means the method exists in the superclass and is overridden in your class.
3.5 Inspecting an object
You can immediately test the effect of the code you have written by creating a
new LOCell object and inspecting it: Open a Playground, type the expression
LOCell new, and Inspect it (using the menu item with the same name).
The left-hand column of the inspector shows a list of instance variables and
the value of the instance variable is shown in the right column (see Figure 3.6).
If you click on an instance variable the inspector will open a new pane with
the detail of the instance variable (see Figure 3.7).
The bottom pane of the inspector is a mini-playground.
It’s useful because in this playground the pseudo-variable self is bound to the
object selected.
Executing expressions.
Go to that Playground at the bottom of the pane and type the text self
bounds: (200@200 corner: 250@250) Do it. To refresh the values, click
on the update button (the blue little circle) at the top right of the pane. The
bounds variable should change in the inspector. Now type the text self openInWorld in the mini-playground and Do it.
39
Figure 3.6: The inspector used to examine a LOCell object
Figure 3.7: When we click on an instance variable, we inspect its value (another
object)
3.6 Defining the class LOGame
Figure 3.8: An LOCell open in world
The cell should appear near the top left-hand corner of the screen (as shown
in Figure 3.8) and exactly where its bounds say that it should appear. Metaclick on the cell to bring up the Morphic halo. Move the cell with the brown
(next to top-right) handle and resize it with the yellow (bottom-right) handle.
Notice how the bounds reported by the inspector also change. (You may have
to click refresh to see the new bounds value.) Delete the cell by clicking on the
x in the pink handle.
3.6 Defining the class LOGame
Now let’s create the other class that we need for the game, which we will name
LOGame.
Class creation
Make the class definition template visible in the browser main window. Do
this by clicking on the package name (or right-clicking on the Class pane and
selecting Add Class). Edit the code so that it reads as follows, and Accept it.
BorderedMorph subclass: #LOGame
instanceVariableNames: ''
classVariableNames: ''
package: 'PBE-LightsOut'
41
A first application
Figure 3.9: Declaring cells as a new instance variable
Here we subclass BorderedMorph. Morph is the superclass of all of the graphical shapes in Pharo, and (unsurprisingly) a BorderedMorph is a Morph with a
border. We could also insert the names of the instance variables between the
quotes on the second line, but for now, let’s just leave that list empty.
Initializing our game
Now let’s define an initialize method for LOGame. Type the following into
the browser as a method for LOGame and try to Accept it.
initialize
| sampleCell width height n |
super initialize.
n := self cellsPerSide.
sampleCell := LOCell new.
width := sampleCell width.
height := sampleCell height.
self bounds: (5 @ 5 extent: (width * n) @ (height * n) + (2 * self
borderWidth)).
cells := Matrix new: n tabulate: [ :i :j | self newCellAt: i at: j ]
Pharo will complain that it doesn’t know the meaning of cells (see Figure 3.9).
It will offer you a number of ways to fix this.
Choose Declare new instance variable, because we want cells to be an instance variable.
Taking advantage of the debugger
At this stage if you open a Playground, type LOGame new, and Do it, Pharo will
complain that it doesn’t know the meaning of some of the terms (see Figure
3.10). It will tell you that it doesn’t know of a message cellsPerSide, and will
open a debugger. But cellsPerSide is not a mistake; it is just a method that
we haven’t yet defined. We will do so, shortly.
Now let us do it: type LOGame new and Do it. Do not close the debugger. Click
on the button Create of the debugger, when prompted, select LOGame, the
class which will contain the method, click on ok, then when prompted for
42
3.6 Defining the class LOGame
Figure 3.10: Pharo detecting an unknown selector
a method protocol enter accessing. The debugger will create the method
cellsPerSide on the fly and invoke it immediately. As there is no magic,
the created method will simply raise an exception and the debugger will
stop again (as shown in Figure 3.11) giving you the opportunity to define the
behavior of the method.
Here you can write your method. This method could hardly be simpler: it
answers the constant 10. One advantage of representing constants as methods
is that if the program evolves so that the constant then depends on some other
features, the method can be changed to calculate this value.
cellsPerSide
"The number of cells along each side of the game"
^ 10
Define the method cellsPerSide in the debugger. Do not forget to compile the method definition by using Accept. You should obtain a situation
as shown by Figure 3.12. If you press the button Proceed the program will
continue its execution - here it will stop since we did not define the method
newCellAt:. We could use the same process but for now we stop to explain
a bit what we did so far. Close the debugger, and look at the class definition
once again (which you can do by clicking on LOGame on the second pane of
the System Browser), you will see that the browser has modified it to include
the instance variable cells.
43
A first application
Figure 3.11: The system created a new method with a body to be defined.
Studying the initialize method
Let us now study the method initialize.
1 initialize
2 | sampleCell width height n |
3 super initialize.
4 n := self cellsPerSide.
5 sampleCell := LOCell new.
6 width := sampleCell width.
7 height := sampleCell height.
8 self bounds: (5 @ 5 extent: (width * n) @ (height * n) + (2 * self
borderWidth)).
9 cells := Matrix new: n tabulate: [ :i :j | self newCellAt: i at: j ]
• At line 2, the expression | sampleCell width height n | declares 4
temporary variables. They are called temporary variables because their
scope and lifetime are limited to this method. Temporary variables with
explanatory names are helpful in making code more readable. Lines 4-7
set the value of these variables.
• How big should our game board be? Big enough to hold some integral
44
3.6 Defining the class LOGame
Figure 3.12: Defining cellsPerSide in the debugger
number of cells, and big enough to draw a border around them. How
many cells is the right number? 5? 10? 100? We don’t know yet, and
if we did, we would probably change our minds later. So we delegate
the responsibility for knowing that number to another method, which
we name cellsPerSide, and which we will write in a minute or two.
Don’t be put off by this: it is actually good practice to code by referring
to other methods that we haven’t yet defined. Why? Well, it wasn’t
until we started writing the initialize method that we realized that
we needed it. And at that point, we can give it a meaningful name, and
move on, without interrupting our flow.
• The fourth line uses this method, n := self cellsPerSide. sends the
message cellsPerSide to self, i.e., to this very object. The response,
which will be the number of cells per side of the game board, is assigned
to n.
• The next three lines create a new LOCell object, and assign its width
and height to the appropriate temporary variables.
• Line 8 sets the bounds of the new object. Without worrying too much
about the details just yet, believe us that the expression in parentheses
45
A first application
creates a square with its origin (i.e., its top-left corner) at the point (5,5)
and its bottom-right corner far enough away to allow space for the right
number of cells.
• The last line sets the LOGame object’s instance variable cells to a newly
created Matrix with the right number of rows and columns. We do this
by sending the message new: tabulate: to the Matrix class (classes are
objects too, so we can send them messages). We know that new: tabulate: takes two arguments because it has two colons (:) in its name.
The arguments go right after the colons. If you are used to languages
that put all of the arguments together inside parentheses, this may seem
weird at first. Don’t panic, it’s only syntax! It turns out to be a very good
syntax because the name of the method can be used to explain the roles
of the arguments. For example, it is pretty clear that Matrix rows: 5
columns: 2 has 5 rows and 2 columns, and not 2 rows and 5 columns.
• Matrix new: n tabulate: [ :i :j | self newCellAt: i at: j ]
creates a new n X n matrix and initializes its elements. The initial value
of each element will depend on its coordinates. The (i,j)th element will
be initialized to the result of evaluating self newCellAt: i at: j.
3.7 Organizing methods into protocols
Before we define any more methods, let’s take a quick look at the third pane
at the top of the browser. In the same way that the first pane of the browser
lets us categorize classes into packages, the protocol pane lets us categorize
methods so that we are not overwhelmed by a very long list of method names
in the method pane. These groups of methods are called ”protocols”.
The browser also offers us the --all-- virtual protocol, which, you will not be
surprised to learn, contains all of the methods in the class.
If you have followed along with this example, the protocol pane may well
contain the protocol as yet unclassified. Right-click in the protocol pane
and select categorize all uncategorized to fix this, and move the initialize
method to a new protocol called initialization.
How does the System Browser know that this is the right protocol? Well, in
general Pharo can’t know exactly, but in this case there is also an initialize
method in the superclass, and it assumes that our initialize method should
go in the same protocol as the one that it overrides.
Pharoers frequently use the notation Class >>
method to identify the class to which a method belongs. For example, the
cellsPerSide method in class LOGame would be referred to as LOGame >>
cellsPerSide. Just keep in mind that this is not Pharo syntax exactly, but
merely a convenient notation to indicate ”the instance method cellsPerSide
A typographic convention.
46
3.8 Finishing the game
which belongs to the class LOGame”. (Incidentally, the corresponding notation
for a class-side method would be LOGame class >> someClassSideMethod.)
From now on, when we show a method in this book, we will write the name
of the method in this form. Of course, when you actually type the code into
the browser, you don’t have to type the class name or the >>; instead, you just
make sure that the appropriate class is selected in the class pane.
3.8 Finishing the game
Now let’s define the other method that are used by LOGame >> initialize.
Let’s define LOGame >> newCellAt: at: in the initialization protocol.
LOGame >> newCellAt: i at: j
"Create a cell for position (i,j) and add it to my on-screen
representation at the appropriate screen position. Answer the
new cell"
| c origin |
c := LOCell new.
origin := self innerBounds origin.
self addMorph: c.
c position: ((i - 1) * c width) @ ((j - 1) * c height) + origin.
c mouseAction: [ self toggleNeighboursOfCellAt: i at: j ]
As you can see there are some tabulation and empty lines. In
order to keep the same convention you can right-click on the method edit
area and click on Format (or use CMD-Shift-f shortcut). This will format your
method.
Formatting.
The method defined above created a new LOCell, initialized to position (i,
j) in the Matrix of cells. The last line defines the new cell’s mouseAction to
be the block [ self toggleNeighboursOfCellAt: i at: j ]. In effect, this
defines the callback behaviour to perform when the mouse is clicked. The
corresponding method also needs to be defined.
LOGame >> toggleNeighboursOfCellAt:
i > 1
ifTrue: [ (cells at: i - 1 at:
i < self cellsPerSide
ifTrue: [ (cells at: i + 1 at:
j > 1
ifTrue: [ (cells at: i at: j j < self cellsPerSide
ifTrue: [ (cells at: i at: j +
i at: j
j) toggleState ].
j) toggleState ].
1) toggleState ].
1) toggleState ]
The method toggleNeighboursOfCellAt:at: toggles the state of the four
cells to the north, south, west and east of cell (i, j). The only complication is
47
A first application
Figure 3.13: Drag a method to a protocol
that the board is finite, so we have to make sure that a neighboring cell exists
before we toggle its state.
Place this method in a new protocol called game logic. (Right-click in the
protocol pane to add a new protocol.) To move (re-classify) the method, you
can simply click on its name and drag it to the newly-created protocol (see
Figure 3.13).
To complete the Lights Out game, we need to define two more methods in class
LOCell this time to handle mouse events.
LOCell >> mouseAction: aBlock
^ mouseAction := aBlock
The method above does nothing more than set the cell’s mouseAction variable
to the argument, and then answers the new value. Any method that changes
the value of an instance variable in this way is called a setter method; a method
that answers the current value of an instance variable is called a getter method.
If you are used to getters and setters in other
programming languages, you might expect these methods to be called setMouseAction and getMouseAction. The Pharo convention is different. A
getter always has the same name as the variable it gets, and a setter is named
similarly, but with a trailing ”:”, hence mouseAction and mouseAction:. Collectively, setters and getters are called accessor methods, and by convention
they should be placed in the accessing protocol. In Pharo, all instance variables are private to the object that owns them, so the only way for another
object to read or write those variables is through accessor methods like this
one. In fact, the instance variables can be accessed in subclasses too.
Setter/Getter convention.
48
3.9 Let’s try our code
Go to the class LOCell, define LOCell >> mouseAction: and put it in the
accessing protocol.
Finally, we need to define a method mouseUp:. This will be called automatically by the GUI framework if the mouse button is released while the cursor is
over this cell on the screen. Add the method LOCell >> mouseUp: and then
Categorize automatically the methods.
LOCell >> mouseUp: anEvent
mouseAction value
What this method does is to send the message value to the object stored in
the instance variable mouseAction. In LOGame >> newCellAt: i at: j we
created the block [self toggleNeighboursOfCellAt: i at: j ] which is
toggling all the neighbours of a cell and we assigned this block to the mouseAction of the cell. Therefore sending the value message causes this block to be
evaluated, and consequently the state of the cells will toggle.
3.9 Let’s try our code
That’s it: the Lights Out game is complete! If you have followed all of the steps,
you should be able to play the game, consisting of just 2 classes and 7 methods.
In a Playground, type LOGame new openInWorld and Do it .
The game will open, and you should be able to click on the cells and see how
it works. Well, so much for theory... When you click on a cell, a debugger will
appear. In the upper part of the debugger window you can see the execution
stack, showing all the active methods. Selecting any one of them will show, in
the middle pane, the Smalltalk code being executed in that method, with the
part that triggered the error highlighted.
Click on the line labeled LOGame >> toggleNeighboursOfCellAt: at: (near
the top). The debugger will show you the execution context within this
method where the error occurred (see Figure 3.14).
At the bottom of the debugger is a variable zone. You can inspect the object
that is the receiver of the message that caused the selected method to execute,
so you can look here to see the values of the instance variables. You can also
see the values of the method arguments.
Using the debugger, you can execute code step by step, inspect objects in
parameters and local variables, evaluate code just as you can in a playground,
and, most surprisingly to those used to other debuggers, change the code
while it is being debugged! Some Pharoers program in the debugger almost all
the time, rather than in the browser. The advantage of this is that you see the
method that you are writing as it will be executed, with real parameters in the
actual execution context.
In this case we can see in the first line of the top panel that the toggleState
message has been sent to an instance of LOGame, while it should clearly have
49
A first application
Figure 3.14: The debugger, with the method toggleNeighboursOfCell:at:
selected
been an instance of LOCell. The problem is most likely with the initialization
of the cells matrix. Browsing the code of LOGame >> initialize shows that
cells is filled with the return values of newCellAt: at:, but when we look
at that method, we see that there is no return statement there! By default,
a method returns self, which in the case of newCellAt: at: is indeed an
instance of LOGame. The syntax to return a value from a method in Pharo is ^.
Close the debugger window. Add the expression ^ c to the end of the method
LOGame >> newCellAt:at: so that it returns c.
LOGame >> newCellAt: i at: j
"Create a cell for position (i,j) and add it to my on-screen
representation at the appropriate screen position. Answer the
new cell"
| c origin |
c := LOCell new.
origin := self innerBounds origin.
self addMorph: c.
c position: ((i - 1) * c width) @ ((j - 1) * c height) + origin.
c mouseAction: [ self toggleNeighboursOfCellAt: i at: j ].
^ c
Often, you can fix the code directly in the debugger window and click Proceed
50
3.10
Saving and sharing Pharo code
to continue running the application. In our case, because the bug was in the
initialization of an object, rather than in the method that failed, the easiest
thing to do is to close the debugger window, destroy the running instance of
the game (with the halo CMD-Alt-Shift and click), and create a new one.
Execute LOGame new openInWorld again because if you use the old game
instance it will still contain the block with the old logic.
Now the game should work properly... or nearly so. If we happen to move the
mouse between clicking and releasing, then the cell the mouse is over will also
be toggled. This turns out to be behavior that we inherit from SimpleSwitchMorph. We can fix this simply by overriding mouseMove: to do nothing:
LOCell >> mouseMove: anEvent
Finally we are done!
By default when an error occurs in Pharo, the system
displays a debugger. However, we can fully control this behavior. For example
we can write the error in a file. We can even serialize the execution stack in
a file, zip and reopen it in another image. Now when we are in development
mode the debugger is available to let us go as fast as possible. In production
system, developers often control the debugger to hide their mistakes from
their clients.
About the debugger.
3.10
Saving and sharing Pharo code
Now that you have Lights Out working, you probably want to save it somewhere so that you can archive it and share it with your friends. Of course, you
can save your whole Pharo image, and show off your first program by running
it, but your friends probably have their own code in their images, and don’t
want to give that up to use your image. What you need is a way of getting
source code out of your Pharo image so that other programmers can bring it
into theirs.
Note We’ll be discussing the various ways to save and share code in a
future chapter, Chapter : Sharing Code and Source Control. For now, here
is an overview of some of the available methods.
Saving plain code
The simplest way of doing this is by ”filing out” the code. The right-click menu
in the Package pane will give you the option to File Out the whole of package
PBE-LightsOut. The resulting file is more or less human readable, but is really
intended for computers, not humans. You can email this file to your friends,
and they can file it into their own Pharo images using the file list browser.
51
A first application
Figure 3.15: File Out our PBE-LightsOut
Figure 3.16: Import your code with the file browser
Right-click on the PBE-LightsOut package and file out the contents (see
Figure 3.15). You should now find a file named PBE-LightsOut.st in the same
folder on disk where your image is saved. Have a look at this file with a text
editor.
Open a fresh Pharo image and use the File Browser tool (Tools --> File
Browser) to file in the PBE-LightsOut.st fileout (see Figure 3.16) and fileIn.
Verify that the game now works in the new image.
Monticello packages
Although fileouts are a convenient way of making a snapshot of the code you
have written, they are definitevely ”old school”. Just as most open-source
projects find it much more convenient to maintain their code in a repository
using SVN or Git, so Pharo programmers find it more convenient to manage
their code using Monticello packages. These packages are represented as files
52
3.10
Saving and sharing Pharo code
Figure 3.17: Monticello browser. The package PBE-LightsOut is not save yet
with names ending in .mcz. They are actually zip-compressed bundles that
contain the complete code of your package.
Using the Monticello package browser, you can save packages to repositories
on various types of server, including FTP and HTTP servers. You can also just
write the packages to a repository in a local file system directory. A copy of
your package is also always cached on your local disk in the package-cache
folder. Monticello lets you save multiple versions of your program, merge
versions, go back to an old version, and browse the differences between
versions. In fact, Monticello is a distributed revision control system. This
means it allows developers to save their work on different places, not on a
single repository as it is the case with CVS or Subversion.
You can also send a .mcz file by email. The recipient will have to place it in her
package-cache folder; she will then be able to use Monticello to browse and
load it.
Monticello Browser
Open the Monticello browser from the World menu (see Figure 3.17).
In the right-hand pane of the browser is a list of Monticello repositories,
which will include all of the repositories from which code has been loaded
into the image that you are using. The top list item is a local directory called
the package-cache. It caches copies of the packages that you have loaded or
published over the network. This local cache is really handy because it lets
you keep your own local history. It also allows you to work in places where
you do not have internet access, or where access is slow enough that you do
not want to save to a remote repository very frequently.
53
A first application
Saving and loading code with Monticello
On the left-hand side of the Monticello browser is a list of packages that have a
version loaded into the image. Packages that have been modified since they
were loaded are marked with an asterisk. (These are sometimes referred to as
dirty packages.) If you select a package, the list of repositories is restricted to
just those repositories that contain a copy of the selected package.
Add the PBE-LightsOut package to your Monticello browser using the + Package button and type PBE-LightsOut.
SmalltalkHub: a Github for Pharo
We think that the best way to save your code and share it is to create an
account for your project in SmalltalkHub. SmalltalkHub is like GitHub: it is a
web front-end to a HTTP Monticello server that lets you manage your projects.
There is a public server at http://www.smalltalkhub.com/.
To be able to use SmalltalkHub you will need an account. Open the site in your
browser. Then, click on the Join link and follow the instructions to create a
new account. Finally, login to your account. Click on the + New project to
create a new project. Fill in the information requested and click Register
project button. You will be sent to your profile page, on the right side you will
see the list of your projects and projects you watch by other coders. Click on
the project you just created.
Under Monticello registration title label you will see a box containing a
smalltalk message similar to
MCHttpRepository
location: 'http://www.smalltalkhub.com/mc/UserName/ProjectName/main'
user: 'YourName'
password: 'Optional_Password'
Copy the contents and go back to Pharo. Once your project has been created
on SmalltalkHub, you have to tell your Pharo system to use it. With the PBELightsOut package selected, click the +Repository button in the Monticello
browser. You will see a list of the different Repository types that are available. To add a SmalltalkHub repository select smalltalkhub.com. You will be
presented with a dialog in which you can provide the necessary information
about the server. You should paste the code snippet you have copied from
SmalltalkHub (see Figure 3.18). This message tells Monticello where to find
your project online. You can also provide your user name and password. If
you do not, then Pharo will ask for them each time you try to save into your
online repository at SmalltalkHub.
Once you have accepted, your new repository should be listed on the righthand side of the Monticello browser. Click on the Save button to save a first
version of your Lights Out game on SmalltalkHub. Don’t forget to put a com54
3.10
Saving and sharing Pharo code
Figure 3.18: Create your first repository on SmalltalkHub
Figure 3.19: Do your first commit
ment so that you, or others, will have an idea of what your commit contains
(see Figure 3.19).
To load a package into your image, open Monticello , select the repository
where the package is located and click open button. It will open a new window
with two columns, left one is the Package to be loaded, the right one is the
version of the package to be loaded. Select the Package and the version you
want to load and click the load button.
Open the PBE-LightsOut repository you have just saved. You should see the
package you just saved.
55
A first application
Monticello has many more capabilities, which will be discussed in depth in
Chapter : Sharing Code and Source Control.
If you are already familiar with Git and Github there are several
solutions to manage your projects with Git. This blog posts provides a summary: http://blog.yuriy.tymch.uk/2015/07/pharo-and-github-versioning-revision-2.
html#! There is a chapter in preparation on about how to use Git with Pharo.
Request it on the Pharo mailing-list.
About Git.
3.11
Chapter summary
In this chapter you have seen how to create packages, classes and methods. In
addition, you have learned how to use the System browser, the inspector, the
debugger and the Monticello browser.
• Packages are groups of related classes.
• A new class is created by sending a message to its superclass.
• Protocols are groups of related methods inside a class.
• A new method is created or modified by editing its definition in the
browser and then accepting the changes.
• The inspector offers a simple, general-purpose GUI for inspecting and
interacting with arbitrary objects.
• The browser detects usage of undeclared variables, and offers possible
corrections.
• The initialize method is automatically executed after an object is
created in Pharo. You can put any initialization code there.
• The debugger provides a high-level GUI to inspect and modify the state
of a running program.
• You can share source code by filing out a package, class or method.
• A better way to share code is to use Monticello to manage an external
repository, for example defined as a SmalltalkHub project.
56
CHAPTER
4
Syntax in a nutshell
Pharo adopts a syntax very close to that of its ancestor, Smalltalk. The syntax
is designed so that program text can be read aloud as though it were a kind of
pidgin English. The following method of the class Week shows an example of
the syntax. It checks whether DayNames already contains the argument, i.e. if
this argument represents a correct day name. If this is the case, it will assign it
to the variable StartDay.
startDay: aSymbol
(DayNames includes: aSymbol)
ifTrue: [ StartDay := aSymbol ]
ifFalse: [ self error: aSymbol, ' is not a recognised day name' ]
Pharo’s syntax is minimal. Essentially there is syntax only for sending messages (i.e. expressions). Expressions are built up from a very small number of
primitive elements (message sends, assignments, closures, returns...). There
are only 6 keywords, and there is no syntax for control structures or declaring
new classes. Instead, nearly everything is achieved by sending messages to
objects. For instance, instead of an if-then-else control structure, conditionals
are expressed as messages (such as ifTrue:) sent to Boolean objects. New
(sub-)classes are created by sending a message to their superclass.
4.1
Syntactic elements
Expressions are composed of the following building blocks:
1. The six reserved keywords, or pseudo-variables: self, super, nil, true,
false, and thisContext
57
Syntax in a nutshell
2. Constant expressions for literal objects including numbers, characters,
strings, symbols and arrays
3. Variable declarations
4. Assignments
5. Block closures
6. Messages
We can see examples of the various syntactic elements in the Table below.
Syntax
What it represents
startPoint
Transcript
self
1
2r101
1.5
2.4e7
$a
'Hello'
#Hello
#(1 2 3)
{ 1 . 2 . 1 + 2 }
"a comment"
| x y |
x := 1
[:x | x + 2 ]
<primitive: 1>
3 factorial
3 + 4
2 raisedTo: 6 modulo: 10
^ true
x := 2 . x := x + x
Transcript show: 'hello'; cr
a variable name
a global variable name
pseudo-variable
decimal integer
binary integer
floating point number
exponential notation
the character 'a'
the string 'Hello'
the symbol #Hello
a literal array
a dynamic array
a comment
declaration of variables x and y
assign 1 to x
a block that evaluates to x + 2
virtual machine primitive or annotation
unary message factorial
binary message +
keyword message raisedTo:modulo:
return the value true
expression separator (.)
message cascade (;)
Local variables. startPoint is a variable name, or identifier. By convention,
identifiers are composed of words in ”camelCase” (i.e., each word except the
first starting with an upper case letter). The first letter of an instance variable,
method or block argument, or temporary variable must be lower case. This
indicates to the reader that the variable has a private scope.
Shared variables. Identifiers that start with upper case letters are global
variables, class variables, pool dictionaries or class names. Transcript is a
global variable, an instance of the class ThreadSafeTranscript.
The receiver. self is a keyword that refers to the object inside which the
current method is executing. We call it ”the receiver” because this object
58
4.1
Syntactic elements
has received the message that caused the method to execute. self is called a
”pseudo-variable” since we cannot assign to it.
Integers. In addition to ordinary decimal integers like 42, Pharo also provides
a radix notation. 2r101 is 101 in radix 2 (i.e., binary), which is equal to decimal
5.
Floating point numbers. can be specified with their base-ten exponent:
2.4e7 is 2.4 X 10^7.
Characters. A dollar sign introduces a literal character: $a is the literal for
the character 'a'. Instances of non-printing characters can be obtained
by sending appropriately named messages to the Character class, such as
Character space and Character tab.
Strings. Single quotes ' ' are used to define a literal string. If you want a
string with a single quote inside, just double the quote, as in 'G''day'.
Symbols. Symbols are like Strings, in that they contain a sequence of characters. However, unlike a string, a literal symbol is guaranteed to be globally
unique. There is only one Symbol object #Hello but there may be multiple
String objects with the value 'Hello'.
Compile-time arrays. are defined by #( ), surrounding space-separated
literals. Everything within the parentheses must be a compile-time constant.
For example, #(27 (true false) abc) is a literal array of three elements:
the integer 27, the compile-time array containing the two booleans, and the
symbol #abc. (Note that this is the same as #(27 #(true false) #abc).)
Run-time arrays. Curly braces { } define a (dynamic) array at run-time.
Elements are expressions separated by periods. So { 1. 2. 1 + 2 } defines
an array with elements 1, 2, and the result of evaluating 1+2.
Comments. are enclosed in double quotes ” ”. ”hello” is a comment, not a
string, and is ignored by the Pharo compiler. Comments may span multiple
lines.
Local variable definitions. Vertical bars | | enclose the declaration of one or
more local variables in a method (and also in a block).
Assignment. := assigns an object to a variable.
Blocks. Square brackets [ ] define a block, also known as a block closure
or a lexical closure, which is a first-class object representing a function. As
we shall see, blocks may take arguments ([:i | ...]) and can have local
variables.
Primitives. < primitive: ... > denotes an invocation of a virtual machine
primitive. For example, < primitive: 1 > is the VM primitive for SmallInteger. Any code following the primitive is executed only if the primitive
fails. The same syntax is also used for method annotations (pragmas).
59
Syntax in a nutshell
Unary messages. These consist of a single word (like factorial) sent to
a receiver (like 3). In 3 factorial, 3 is the receiver, and factorial is the
message selector.
Binary messages. These are message with an argument and whose selector
looks like mathematical expressions (for example: +) sent to a receiver, and
taking a single argument. In 3 + 4, the receiver is 3, the message selector is +,
and the argument is 4.
Keyword messages. They consist of multiple keywords (like raisedTo: modulo:), each ending with a colon and taking a single argument. In the expression 2 raisedTo: 6 modulo: 10, the message selector raisedTo:modulo:
takes the two arguments 6 and 10, one following each colon. We send the
message to the receiver 2.
Method return. ^ is used to return a value from a method.
Sequences of statements. A period or full-stop (.) is the statement separator.
Putting a period between two expressions turns them into independent
statements.
Cascades. Semicolons ( ; ) can be used to send a cascade of messages to a
single receiver. In Transcript show: 'hello'; cr we first send the keyword
message show: 'hello' to the receiver Transcript, and then we send the
unary message cr to the same receiver.
The classes Number, Character, String and Boolean are described in more
detail in Chapter : Basic Classes.
4.2
Pseudo-variables
In Pharo, there are 6 reserved keywords, or pseudo-variables: nil, true,
false, self, super, and thisContext. They are called pseudo-variables because they are predefined and cannot be assigned to. true, false, and nil are
constants, while the values of self, super, and thisContext vary dynamically
as code is executed.
• true and false are the unique instances of the Boolean classes True and
False. See Chapter : Basic Classes for more details.
• self always refers to the receiver of the currently executing method.
• super also refers to the receiver of the current method, but when you
send a message to super, the method-lookup changes so that it starts
from the superclass of the class containing the method that uses super.
For further details see Chapter : The Pharo Object Model.
• nil is the undefined object. It is the unique instance of the class UndefinedObject. Instance variables, class variables and local variables are
initialized to nil.
60
4.3
Message sends
• thisContext is a pseudo-variable that represents the top frame of
the execution stack. thisContext is normally not of interest to most
programmers, but it is essential for implementing development tools
like the debugger, and it is also used to implement exception handling
and continuations.
4.3
Message sends
There are three kinds of messages in Pharo. This distinction has been made to
reduce the number of mandatory parentheses.
1. Unary messages take no argument. 1 factorial sends the message
factorial to the object 1.
2. Binary messages take exactly one argument. 1 + 2 sends the message +
with argument 2 to the object 1.
3. Keyword messages take an arbitrary number of arguments. 2 raisedTo:
6 modulo: 10 sends the message consisting of the message selector
raisedTo:modulo: and the arguments 6 and 10 to the object 2.
Unary message selectors consist of alphanumeric characters, and start with a lower case letter.
Unary messages.
Binary message selectors consist of one or more characters
from the following set:
Binary messages.
+ - / \ * ~ < > = @ % | & ? ,
Keyword message selectors consist of a series of
alphanumeric keywords, where each keyword starts with a lower-case letter
and ends with a colon.
Keyword message selectors.
Unary messages have the highest precedence, then
binary messages, and finally keyword messages, so:
Message precedence.
2 raisedTo: 1 + 3 factorial
>>> 128
First we send factorial to 3, then we send + 6 to 1, and finally we send
raisedTo: 7 to 2. Recall that we use the notation expression >>> to show the
result of evaluating an expression.
Precedence aside, execution is strictly from left to right, so:
1 + 2 * 3
>>> 9
return 9 and not 7. Parentheses must be used to alter the order of evaluation:
61
Syntax in a nutshell
1 + (2 * 3)
>>> 7
Message sends may be composed with periods
and semi-colons. A period separated sequence of expressions causes each
expression in the series to be evaluated as a statement, one after the other.
Periods and semi-colons.
Transcript cr.
Transcript show: 'hello world'.
Transcript cr
This will send cr to the Transcript object, then send it show: 'hello world',
and finally send it another cr.
When a series of messages is being sent to the same receiver, then this can be
expressed more succinctly as a cascade. The receiver is specified just once, and
the sequence of messages is separated by semi-colons:
Transcript
cr;
show: 'hello world';
cr
This has precisely the same effect as the previous example.
4.4
Method syntax
Whereas expressions may be evaluated anywhere in Pharo (for example, in a
playground, in a debugger, or in a browser), methods are normally defined in
a browser window, or in the debugger. Methods can also be filed in from an
external medium, but this is not the usual way to program in Pharo.
Programs are developed one method at a time, in the context of a given class.
A class is defined by sending a message to an existing class, asking it to create
a subclass, so there is no special syntax required for defining classes.
Here is the method lineCount in the class String. The usual convention is
to refer to methods as ClassName»methodName. Here the method is then
String»lineCount. Note that ClassName»methodName is not part of the Pharo
syntax just a convention used in books to clearly define a method.
String >> lineCount
"Answer the number of lines represented by the receiver, where every
cr adds one line."
| cr count |
cr := Character cr.
count := 1 min: self size.
self do: [:c | c == cr ifTrue: [count := count + 1]].
^ count
62
4.5
Block syntax
Syntactically, a method consists of:
1. the method pattern, containing the name (i.e., lineCount) and any
arguments (none in this example)
2. comments (these may occur anywhere, but the convention is to put one
at the top that explains what the method does)
3. declarations of local variables (i.e., cr and count); and
4. any number of expressions separated by dots (here there are four)
The execution of any expression preceded by a ^ (a caret or upper arrow,
which is Shift-6 for most keyboards) will cause the method to exit at that
point, returning the value of that expression. A method that terminates
without explicitly returning some expression will implicitly return self.
Arguments and local variables should always start with lower case letters.
Names starting with upper-case letters are assumed to be global variables.
Class names, like Character, for example, are simply global variables referring
to the object representing that class.
4.5
Block syntax
Blocks (lexical closures) provide a mechanism to defer the execution of expressions. A block is essentially an anonymous function with a definition context.
A block is executed by sending it the message value. The block answers the
value of the last expression in its body, unless there is an explicit return (with
^) in which case it returns the value of the returned expression.
[ 1 + 2 ] value
>>> 3
[ 3 = 3 ifTrue: [ ^ 33 ]. 44 ] value
>>> 33
Blocks may take parameters, each of which is declared with a leading colon. A
vertical bar separates the parameter declaration(s) from the body of the block.
To evaluate a block with one parameter, you must send it the message value:
with one argument. A two-parameter block must be sent value:value:, and
so on, up to 4 arguments.
[ :x | 1 + x ] value: 2
>>> 3
[ :x :y | x + y ] value: 1 value: 2
>>> 3
If you have a block with more than four parameters, you must use valueWithArguments: and pass the arguments in an array. (A block with a large number
of parameters is often a sign of a design problem.)
63
Syntax in a nutshell
Blocks may also declare local variables, which are surrounded by vertical bars,
just like local variable declarations in a method. Locals are declared after any
arguments:
[ :x :y |
| z |
z := x + y.
z ] value: 1 value: 2
>>> 3
Blocks are actually lexical closures, since they can refer to variables of the
surrounding environment. The following block refers to the variable x of its
enclosing environment:
| x |
x := 1.
[ :y | x + y ] value: 2
>>> 3
Blocks are instances of the class BlockClosure. This means that they are
objects, so they can be assigned to variables and passed as arguments just like
any other object.
4.6
Conditionals and loops in a nutshell
Pharo offers no special syntax for control constructs. Instead, these are
typically expressed by sending messages to booleans, numbers and collections,
with blocks as arguments.
Some conditionals
Conditionals are expressed by sending one of the messages ifTrue:, ifFalse:
or ifTrue:ifFalse: to the result of a boolean expression. See Chapter : Basic
Classes, for more about booleans.
(17 * 13 > 220)
ifTrue: [ 'bigger' ]
ifFalse: [ 'smaller' ]
>>>'bigger'
Some loops
Loops are typically expressed by sending messages to blocks, integers or
collections. Since the exit condition for a loop may be repeatedly evaluated,
it should be a block rather than a boolean value. Here is an example of a very
procedural loop:
n := 1.
[ n < 1000 ] whileTrue: [ n := n*2 ].
64
4.6
Conditionals and loops in a nutshell
n
>>> 1024
whileFalse: reverses the exit condition.
n := 1.
[ n > 1000 ] whileFalse: [ n := n*2 ].
n
>>> 1024
timesRepeat: offers a simple way to implement a fixed iteration:
n := 1.
10 timesRepeat: [ n := n*2 ].
n
>>> 1024
We can also send the message to:do: to a number which then acts as the
initial value of a loop counter. The two arguments are the upper bound, and a
block that takes the current value of the loop counter as its argument:
result := String new.
1 to: 10 do: [:n | result := result, n printString, ' '].
result
>>> '1 2 3 4 5 6 7 8 9 10 '
High-order iterators
Collections comprise a large number of different classes, many of which support the same protocol. The most important messages for iterating over collections include do:, collect:, select:, reject:, detect: and inject:into:.
These messages define high-level iterators that allow one to write very compact code.
An Interval is a collection that lets one iterate over a sequence of numbers
from the starting point to the end. 1 to: 10 represents the interval from 1 to
10. Since it is a collection, we can send the message do: to it. The argument is
a block that is evaluated for each element of the collection.
result := String new.
(1 to: 10) do: [:n | result := result, n printString, ' '].
result
>>> '1 2 3 4 5 6 7 8 9 10 '
collect: builds a new collection of the same size, transforming each element.
(You can think of collect: as the Map in the MapReduce programming
model).
(1 to:10) collect: [ :each | each * each ]
>>> #(1 4 9 16 25 36 49 64 81 100)
select: and reject: build new collections, each containing a subset of the
elements satisfying (or not) the boolean block condition.
65
Syntax in a nutshell
detect: returns the first element satisfying the condition. Don’t forget that
strings are also collections (of characters), so you can iterate over all the
characters.
'hello there' select: [ :char | char isVowel ]
>>> 'eoee'
'hello there' reject: [ :char | char isVowel ]
>>> 'hll thr'
'hello there' detect: [ :char | char isVowel ]
>>> $e
Finally, you should be aware that collections also support a functional-style
fold operator in the inject:into: method. You can also think of it as the
Reduce in the MapReduce programming model. This lets you generate a
cumulative result using an expression that starts with a seed value and injects
each element of the collection. Sums and products are typical examples.
(1 to: 10) inject: 0 into: [ :sum :each | sum + each]
>>> 55
This is equivalent to 0+1+2+3+4+5+6+7+8+9+10.
More about collections can be found in Chapter : Collections.
4.7
Primitives and pragmas
In Pharo everything is an object, and everything happens by sending messages.
Nevertheless, at certain points we hit rock bottom. Certain objects can only
get work done by invoking virtual machine primitives.
For example, the following are all implemented as primitives: memory allocation (new, new:), bit manipulation (bitAnd:, bitOr:, bitShift:), pointer and
integer arithmetic (+, -, <, >, *, /, =, ==...), and array access (at:, at:put:).
Primitives are invoked with the syntax <primitive: aNumber>. A method
that invokes such a primitive may also include Pharo code, which will be
executed only if the primitive fails.
Here we see the code for SmallInteger>>+. If the primitive fails, the expression super + aNumber will be executed and returned.
+ aNumber
"Primitive. Add the receiver to the argument and answer with the
result
if it is a SmallInteger. Fail if the argument or the result is not a
SmallInteger Essential No Lookup. See Object documentation
whatIsAPrimitive."
<primitive: 1>
^ super + aNumber
66
4.8
Chapter summary
In Pharo, the angle bracket syntax is also used for method annotations called
pragmas.
4.8
Chapter summary
• Pharo has only six reserved identifiers (also called pseudo-variables):
true, false, nil, self, super, and thisContext.
• There are five kinds of literal objects: numbers (5, 2.5, 1.9e15, 2r111),
characters ($a), strings ('hello'), symbols (#hello), and arrays (#('hello'
#hi) or { 1 . 2 . 1 + 2 } )
• Strings are delimited by single quotes, comments by double quotes. To
get a quote inside a string, double it.
• Unlike strings, symbols are guaranteed to be globally unique.
• Use #( ... ) to define a literal array. Use { ... } to define a dynamic
array. Note that #(1+2) size >>> 3, but {1+2} size >>> 1
• There are three kinds of messages: unary (e.g., 1 asString, Array new),
binary (e.g., 3 + 4, 'hi', ' there'), and keyword (e.g., 'hi' at: 2
put: $o)
• A cascaded message send is a sequence of messages sent to the same target, separated by semi-colons: OrderedCollection new add: #calvin;
add: #hobbes; size >>> 2
• Local variables are declared with vertical bars. Use := for assignment.
|x| x := 1
• Expressions consist of message sends, cascades and assignments, evaluated left to right (and optionally grouped with parentheses). Statements
are expressions separated by periods.
• Block closures are expressions enclosed in square brackets. Blocks may
take arguments and can contain temporary variables. The expressions
in the block are not evaluated until you send the block a value message
with the correct number of arguments. [ :x | x + 2 ] value: 4
• There is no dedicated syntax for control constructs, just messages that
conditionally evaluate blocks.
67
CHAPTER
5
Understanding message syntax
Although Pharo’s message syntax is simple, it is unconventional and can take
some time getting used to. This chapter offers some guidance to help you get
acclimatized to this special syntax for sending messages. If you already feel
comfortable with the syntax, you may choose to skip this chapter, or come
back to it later.
5.1
Identifying messages
In Pharo, except for the syntactic elements listed in Chapter : Syntax in a
Nutshell (:= ^ . ; # () {} [ : | ] < >), everything is a message send.
There is no operators, just messages sent to objects. Therefore you can define
a message named + in your class but there is also no precedence because Pharo
always takes the simplest form of definition.
The order in which messages are sent is determined by the type of message.
There are just three types of messages: unary, binary, and keyword messages. Pharo distinguishes such three types of messages to minimize the
number of parentheses. Unary messages are always sent first, then binary
messages and finally keyword ones. As in most languages, parentheses can be
used to change the order of execution. These rules make Pharo code as easy
to read as possible. And most of the time you do not have to think about the
rules.
Message structure
As most computation in Pharo is done by message passing, correctly identifying messages is key to avoiding future mistakes. The following terminology
69
Understanding message syntax
Figure 5.1: Two message sends composed of a receiver, a method selector, and a
set of arguments.
Figure 5.2: aMorph color: Color yellow is composed of two message sends:
Color yelow and aMorph color: Color yellow.
will help us:
• A message is composed of the message selector and the optional message arguments.
• A message is sent to a receiver.
• The combination of a message and its receiver is called a message send as
shown in Figure 5.1.
A message is always sent to a receiver, which can be a single literal, a block or
a variable or the result of evaluating another message.
To help you identify the receiver of a message, we will underline it for you. We
will also surround each message send with an ellipse, and number the message
sends starting from the first one to help you see the send order in Figure 5.2.
Figure 5.2 represents two message sends, Color yellow and aMorph color:
Color yellow, hence there are two ellipses. The message send Color yellow
is executed first, so its ellipse is numbered 1. There are two receivers: 1)
aMorph which receives the message color: ... and 2) Color which receives
the message yellow. Both receivers are underlined.
70
5.2 Three types of messages
A receiver can be the first element of a message, such as 100 in the message
send 100 + 200 or Color in the message send Color yellow. However, a
receiver can also be the result of other messages. For example in the message
Pen new go: 100, the receiver of the message go: 100 is the object returned
by the message send Pen new. In all the cases, a message is sent to an object
called the receiver which may be the result of another message send.
Message send
Message type
Action
Color yellow
aPen go: 100
100 + 20
Browser open
Pen new go: 100
aPen go: 100 + 20
unary
keyword
binary
unary
un. and keyw.
keyw. and bin.
Creates a color
Forwards pen 100 pixels
100 receives the message +
Opens a new browser
Creates and moves pen 100 px
Pen moves forward 120 px
The table shows several characteristics of message sends.
• You should note that not all message sends have arguments. Unary
messages like open do not have arguments. Single keyword and binary
messages like go: 100 and + 20 each have one argument.
• There are also simple messages and composed ones. Color yellow and
100 + 20 are simple: a message is sent to an object, while the message
send aPen go: 100 + 20 is composed of two messages: + 20 is sent to
100 and go: is sent to aPen with the argument being the result of the
first message.
• A receiver can be an expression (such as an assignment, a message send
or a literal) which returns an object. In Pen new go: 100, the message
go: 100 is sent to the object that results from the execution of the
message send Pen new.
5.2 Three types of messages
Pharo distinguishes between three kinds of messages to reduce mandatory
parentheses. A few simple rules based on the such different message determine the order in which the messages are sent.
There are three different types of messages:
• Unary messages are messages that are sent to an object without any
other information. For example, in 3 factorial, factorial is a unary
message.
• Binary messages are messages consisting of operators (often arithmetic).
They are binary because they always involve only two objects: the
receiver and the argument object. For example in 10 + 20, + is a binary
message sent to the receiver 10 with argument 20.
71
Understanding message syntax
• Keyword messages are messages consisting of one or more keywords,
each ending with a colon (:) and taking an argument. For example in
anArray at: 1 put: 10, the keyword at: takes the argument 1 and
the keyword put: takes the argument 10.
Unary messages
Unary messages are messages that do not require any argument. They follow the syntactic template: receiver messageName. The selector is simply
made up of a succession of characters not containing : (e.g., factorial, open,
class).
89 sin
>>> 0.860069405812453
3 sqrt
>>> 1.732050807568877
Float pi
>>> 3.141592653589793
'blop' size
>>> 4
true not
>>> false
Object class
>>> Object class "The class of Object is Object class (BANG)"
Important Unary messages are messages that do not require any argument. They follow the syntactic template: receiver selector.
Binary messages
Binary messages are messages that require exactly one argument and whose
selector consists of a sequence of one or more characters from the set: +, -, *,
/, &, =, >, |, <, ~, and @. Note that -- (double minus) is not allowed for parsing
reasons.
100@100
>>> 100@100 "creates a Point object"
3 + 4
>>> 7
10 - 1
>>> 9
4 <= 3
>>> false
72
5.2 Three types of messages
(4/3) * 3 == 4
>>> true "equality is just a binary message, and Fractions are exact"
(3/4) == (3/4)
>>> false "two equal Fractions are not the same object"
Important Binary messages are messages that require exactly one argument and whose selector is composed of a sequence of characters from:
+, -, *, /, &, =, >, |, <, ~, and @. -- is not possible. They follow the syntactic
template: receiver selector argument.
Keyword messages
Keyword messages are messages that require one or more arguments and
whose selector consists of one or more keywords each ending in :. Keyword
messages follow the syntactic template: receiver selectorWordOne: argumentOne wordTwo: argumentTwo.
Each keyword takes an argument. Hence r:g:b: is a method with three
arguments, playFileNamed: and at: are methods with one argument, and
at:put: is a method with two arguments. To create an instance of the class
Color one can use the method r:g:b: (as in Color r: 1 g: 0 b: 0), which
creates the color red. Note that the colons are part of the selector.
Note In Java or C++, the method invocation Color r: 1 g: 0 b: 0
would be written Color.rgb(1, 0, 0).
1 to: 10
>>> (1 to: 10) "creates an interval"
Color r: 1 g: 0 b: 0
>>> Color red "creates a new color"
12 between: 8 and: 15
>>> true
nums := Array newFrom: (1 to: 5).
nums at: 1 put: 6.
nums
>>> #(6 2 3 4 5)
Important Keyword messages are messages that require one or more arguments. Their selector consists of one or more keywords each ending in a
colon (:). They follow the syntactic template: receiver selectorWordOne:
argumentOne wordTwo: argumentTwo.
73
Understanding message syntax
Figure 5.3: Unary messages are sent first so Color yellow is sent. This returns a
color
5.3 Message composition
The three types of messages each have different precedence, which allows
them to be composed in an elegant way and with few parentheses.
• Unary messages are always sent first, then binary messages and finally,
keyword messages.
• Messages in parentheses are sent prior to any other messages.
• Messages of the same kind are evaluated from left to right.
These rules lead to a very natural reading order. If you want to be sure that
your messages are sent in the order that you want, you can always add more
parentheses, as shown in Figure 5.3. In this figure, the message yellow is
an unary message and the message color: a keyword message, therefore
the message send Color yellow is sent first. However as message sends
in parentheses are sent first, putting (unnecessary) parentheses around
Color yellow just emphasizes that it will be sent first. The rest of the section
illustrates each of these points.
Unary > Binary > Keywords
Unary messages are sent first, then binary messages, and finally keyword
messages. We also say that unary messages have a higher priority over the
other types of messages.
Important Rule one. Unary messages are sent first, then binary messages,
and finally keyword based messages. Unary > Binary > Keyword
As these examples show, Pharo’s syntax rules generally ensure that message
sends can be read in a natural way:
1000 factorial / 999 factorial
>>> 1000
2 raisedTo: 1 + 3 factorial
>>> 128
Unfortunately the rules are a bit too simplistic for arithmetic message sends,
so you need to introduce parentheses whenever you want to impose a priority
over binary operators:
74
5.3 Message composition
Figure 5.4: Binary messages are sent before keyword
1 + 2 * 3
>>> 9
1 + (2 * 3)
>>> 7
In the message aPen color: Color yellow, there is one unary
message yellow sent to the class Color and a keyword message color: sent to
aPen. Unary messages are sent first so the message send Color yellow is sent
(1). This returns a color object which is passed as argument of the message
aPen color: aColor (2) as shown in the following script. Figure 5.3 shows
graphically how messages are sent.
Example 1.
aPen color: Color yellow
"unary message is sent first"
(1) Color yellow
>>> aColor
"keyword message is sent next"
(2) aPen color: aColor
Example 2. In the message aPen go: 100 + 20, there is a binary message
+ 20 and a keyword message go:. Binary messages are sent prior to keyword
messages so 100 + 20 is sent first (1): the message + 20 is sent to the object
100 and returns the number 120. Then the message aPen go: 120 is sent with
120 as argument (2). The following example shows how the message send is
executed:
aPen go: 100 + 20
"binary message first"
(1) 100 + 20
>>> 120
"then keyword message"
(2) aPen go: 120
As an exercise we let you decompose the evaluation of the message Pen new go: 100 + 20 which is composed of one unary, one keyword
and one binary message (see Figure 5.5).
Example 3.
75
Understanding message syntax
Figure 5.5: Decomposing Pen new go: 100 +
Figure 5.6: Decomposing Pen new
Parentheses first
Parenthesised messages are sent prior to other messages.
3 + 4 factorial
>>> 27 "(not 5040)"
(3 + 4) factorial
>>> 5040
Here we need the parentheses to force sending lowMajorScaleOn: before
play.
(FMSound lowMajorScaleOn: FMSound clarinet) play
"(1) send the message clarinet to the FMSound class to create a
clarinet sound.
(2) send this sound to FMSound as argument to the lowMajorScaleOn:
keyword message.
(3) play the resulting sound."
Important Rule two. Parenthesised messages are sent prior to other
messages. (Msg) > Unary > Binary > Keyword
The message (65@325 extent: 134@100) center returns the
center of a rectangle whose top left point is (65, 325) and whose size is 134x100.
The following script shows how the message is decomposed and sent. First the
message between parentheses is sent. It contains two binary messages, 65@325
and 134@100, that are sent first and return points, and a keyword message
extent: which is then sent and returns a rectangle. Finally the unary message
center is sent to the rectangle and a point is returned. Evaluating the message
without parentheses would lead to an error because the object 100 does not
understand the message center.
Example 4.
(65@325 extent: 134@100) center
"Expression within parentheses then binary"
76
5.3 Message composition
(1) 65@325
>>> aPoint
"binary"
(2)134@100
>>> anotherPoint
"keyword"
(3) aPoint extent: anotherPoint
>>> aRectangle
"unary"
(4) aRectangle center
>>> 132@375
From left to right
Now we know how messages of different kinds or priorities are handled. The
final question to be addressed is how messages with the same priority are
sent. They are sent from the left to the right. Note that you already saw this
behaviour in the previous script where the two point creation messages (@)
were sent first.
The following script shows that execution from left to right for messages of
the same type reduces the need for parentheses.
1.5 tan rounded asString = (((1.5 tan) rounded) asString)
>>> true
Important Rule three. When the messages are of the same kind, the order
of evaluation is from left to right.
In the message sends Pen new down all messages are unary
messages, so the leftmost one, Pen new, is sent first. This returns a newly
created pen to which the second message down is sent, as shown in Figure 5.6.
Example 5.
Arithmetic inconsistencies
The message composition rules are simple but they result in inconsistency
for the execution of arithmetic message sends expressed in terms of binary
messages. Here we see the common situations where extra parentheses are
needed.
3 + 4 * 5
>>> 35 "(not 23) Binary messages sent from left to right"
3 + (4 * 5)
>>> 23
1 + 1/3
>>> (2/3) "and not 4/3"
77
Understanding message syntax
Figure 5.7: The two messages + and * are of the same kind so their execution is
from left to right.
1 + (1/3)
>>> (4/3)
1/3 + 2/3
>>> (7/9) "and not 1"
(1/3) + (2/3)
>>> 1
In the message sends 20 + 2 * 5, there are only binary messages + and *. However in Pharo there is no special priority for the operations
+ and *. They are just binary messages, hence * does not have priority over +.
Here the leftmost message + is sent first, and then the * is sent to the result as
shown in the following script and Figure 5.7.
Example 6.
"As there is no priority among binary messages, the leftmost message +
is evaluated first
even if by the rules of arithmetic the * should be sent first."
20 + 2 * 5
(1) 20 + 2
>>> 22
(2) 22 * 5
>>> 110
As shown in the previous example the result of this message send is not 30
but 110. This result is perhaps unexpected but follows directly from the rules
used to send messages. This is the price to pay for the simplicity of the Pharo
model. To get the correct result, we should use parentheses. When messages
are enclosed in parentheses, they are evaluated first. Hence the message send
20 + (2 * 5) returns the result as shown by the following script and Figure
5.8.
"The messages surrounded by parentheses are evaluated first therefore *
is sent prior to + which produces the correct behaviour."
20 + (2 * 5)
(1) (2 * 5)
>>> 10
(2) 20 + 10
>>> 30
78
5.4
Hints for identifying keyword messages
Figure 5.8: Parenthesed expressions are executed first.
Figure 5.9: Equivalent messages using parentheses.
Figure 5.10: Equivalent messages using parentheses.
Note Arithmetic operators such as + and * do not have different priority.
+ and * are just binary messages, therefore * does not have priority over +.
Use parentheses to obtain the desired result.
Table : Message sends and their fully parenthesized equivalents
Implicit precedence
Explicitly parenthesized equivalent
aPen color: Color yellow
aPen go: 100 + 20
aPen penSize: aPen penSize + 2
2 factorial + 4
aPen color: (Color yellow)
Pen go: (100 + 20)
aPen penSize: ((aPen penSize) + 2)
(2 factorial) + 4
Note that the first rule stating that unary messages are sent prior to binary
and keyword messages avoids the need to put explicit parentheses around
them. Table above shows message sends written following the rules and
equivalent message sends if the rules would not exist. Both message sends
result in the same effect or return the same value.
5.4
Hints for identifying keyword messages
Often beginners have problems understanding when they need to add parentheses. Let’s see how keywords messages are recognized by the compiler.
79
Understanding message syntax
Parentheses or not?
The characters [, ], ( and ) delimit distinct areas. Within such an area, a keyword message is the longest sequence of words terminated by : that is not cut
by the characters ., or ;. When the characters [, ], ( and { surround some words
with colons, these words participate in the keyword message local to the area
defined.
In this example, there are two distinct keyword messages: rotateBy:magnify:smoothing: and at:put:.
aDict
at: (rotatingForm
rotateBy: angle
magnify: 2
smoothing: 1)
put: 3
Note The characters [, ], ( and ) delimit distinct areas. Within such an
area, a keyword message is the longest sequence of words terminated by
: that is not cut by the characters ., or ;.When the characters [, ], ( and )
surround some words with colons, these words participate in the keyword
message local to the area defined.
Precedence hints
If you have problems with these precedence rules, you may start simply by
putting parentheses whenever you want to distinguish two messages having
the same precedence.
The following piece of code does not require parentheses because the message
send x isNil is unary, and is sent prior to the keyword message ifTrue:.
(x isNil)
ifTrue: [...]
The following code requires parentheses because the messages includes: and
ifTrue: are both keyword messages.
ord := OrderedCollection new.
(ord includes: $a)
ifTrue: [...]
Without parentheses the unknown message includes:ifTrue: would be sent
to the collection!
When to use [ ] or ( )
You may also have problems understanding when to use square brackets
(blocks) rather than parentheses. The basic principle is that you should use
80
5.5 Expression sequences
[ ] when you do not know how many times, potentially zero, an expression
should be executed. [ expression ] will create a block closure (an object,
as always) from expression, which may be executed zero or more times,
depending on the context. (Recall from Chapter : Syntax in a Nutshell that an
expression can either be a message send, a variable, a literal, an assignment or
a block.)
Following this principle, the conditional branches of ifTrue: or ifTrue:ifFalse: require blocks. Similarly, both the receiver and the argument of the
whileTrue: message require the use of square brackets since we do not know
how many times either the receiver (the loop conditional) or the argument
(the ”loop body”) will be executed.
Parentheses, on the other hand, only affect the order of sending messages. So
in ( expression ), the expression will always be executed exactly once.
"both the receiver and the argument must be blocks"
[ x isReady ] whileTrue: [ y doSomething ]
"the argument is evaluated more than once, so must be a block"
4 timesRepeat: [ Beeper beep ]
"receiver is evaluated once, so is not a block"
(x isReady) ifTrue: [ y doSomething ]
5.5 Expression sequences
Expressions (i.e., message sends, assignments and so on) separated by periods
are evaluated in sequence. Note that there is no period between a variable
definition (the | box | section) and the following expressions. The value
of a sequence is the value of the last expression. The values returned by all
the expressions except the last one are ignored. Note that the period is a
separator between expressions, and not a terminator. Therefore a final period
is optional.
| box |
box := 20@30 corner: 60@90.
box containsPoint: 40@50
>>> true
5.6 Cascaded messages
Pharo offers a way to send multiple messages to the same receiver using a
semicolon (;). This is called the cascade in Pharo jargon. It follows the pattern:
expression msg1; msg2
Transcript show: 'Pharo is '.
Transcript show: 'fun '.
Transcript cr.
81
Understanding message syntax
is equivalent to:
Transcript
show: 'Pharo is';
show: 'fun ';
cr
Note that the object receiving the cascaded messages can itself be the result
of a message send. In fact, the receiver of all the cascaded messages is the
receiver of the first message involved in a cascade. In the following example,
the first cascaded message is setX:setY since it is followed by a cascade.
The receiver of the cascaded message setX:setY: is the newly created point
resulting from the evaluation of Point new, and not Point. The subsequent
message isZero is sent to that same receiver.
Point new setX: 25 setY: 35; isZero
>>> false
Important expression msg1. expression msg2 is equivalent to expression msg1; msg2
5.7 Chapter summary
• A message is always sent to an object named the receiver, which itself
may be the result of other message sends.
• There are three types of messages: unary, binary, and keyword.
• Unary messages are messages that do not require any argument. They
are of the form of receiver selector.
• Binary messages are messages that involve two objects, the receiver and
another object, whose selector is composed of one or more characters
from the following list: +, -, *, /, |, &, =, >, <, ~, and @. They are of the
form: receiver selector argument.
• Keyword messages are messages that involve more than one object
and that contain at least one colon character (:). They are of the form:
receiver selectorWordOne: argumentOne wordTwo: argumentTwo.
• Rule One. Unary messages are sent first, then binary messages, and
finally keyword messages.
• Rule Two. Messages in parentheses are sent before any others.
• Rule Three. When the messages are of the same type, the order of
evaluation is from left to right.
• In Pharo, traditional arithmetic operators such as + and * have the
same priority. + and * are just binary messages, therefore * does not
82
5.7 Chapter summary
have priority over +. You must use parentheses to obtain the desired
arithmetical order of operations.
83
CHAPTER
6
The Pharo object model
The Pharo programming model is heavily inspired by the one of Smalltalk. It is
simple and uniform: everything is an object, and objects communicate only
by sending each other messages. Instance variables are private to the object.
Methods are all public and dynamically looked up (late-bound). In this chapter
we present the core concepts of the Pharo object model. We revisit concepts
such as self and super and define precisely their semantics. Then we discuss
the consequences of representing classes as objects. This will be extended in
Chapter: Classes and Metaclasses.
6.1
The rules of the model
The object model is based on a set of simple rules that are applied uniformly.
The rules are as follows:
Rule 1. Everything is an object.
Rule 2. Every object is an instance of a class.
Rule 3. Every class has a superclass.
Rule 4. Everything happens by sending messages.
Rule 5. Method lookup dynamically follows the inheritance chain.
Let us look at each of these rules in some detail.
6.2 Everything is an Object
The mantra everything is an object is highly contagious. After only a short
while working with Pharo, you will start to be surprised at how this rule
85
The Pharo object model
simplifes everything you do. Integers, for example, are truly objects, so you
can send messages to them, just as you do to any other object. At the end of
this chapter, we added an implementation note on the object implementation
for the curious reader.
"send '+ 4' to 3, yielding 7"
3 + 4
>>> 7
"send factorial, yielding a big number"
20 factorial
>>> 2432902008176640000
The object 7 is different than the object returned by 20 factorial, but because they are both polymorphic objects, none of the code, not even the
implementation of factorial, needs to know about this.
Coming back to everything is an object rule, perhaps the most fundamental
consequence of this rule is that classes are objects too. Classes are not secondclass objects: they are really first-class objects that you can send messages to,
inspect, and change. This means that Pharo is a truly reflective system, which
gives a great deal of expressive power to developers.
Important
Classes are objects too.
6.3 Every object is an instance of a class
Every object has a class; you can find out which one by sending it the message
class.
1 class
>>> SmallInteger
20 factorial class
>>> LargePositiveInteger
'hello' class
>>> ByteString
(4@5) class
>>> Point
Object new class
>>> Object
A class defines the structure of its instances via instance variables, and the
behavior of its instances via methods. Each method has a name, called its
selector, which is unique within the class.
Since classes are objects, and every object is an instance of a class, it follows that
classes must also be instances of classes. A class whose instances are classes
86
6.4
Instance structure and behavior
is called a metaclass. Whenever you create a class, the system automatically
creates a metaclass for you. The metaclass defines the structure and behavior
of the class that is its instance. 99% of the time you will not need to think
about metaclasses, and may happily ignore them. (We will have a closer look
at metaclasses in Chapter : Classes and Metaclasses.)
6.4
Instance structure and behavior
Now we will briefly present how we specify the structure and behavior of
instances.
Instance variables
Instance variables in Pharo are private to the instance itself. This is in contrast
to Java and C++, which allow instance variables (also known as fields or member
variables) to be accessed by any other instance that happens to be of the same
class. We say that the encapsulation boundary of objects in Java and C++ is the
class, whereas in Pharo it is the instance.
In Pharo, two instances of the same class cannot access each other’s instance
variables unless the class defines accessor methods. There is no language syntax that provides direct access to the instance variables of any other object.
(Actually, a mechanism called reflection does provide a way to ask another
object for the values of its instance variables; meta-programming is intended
for writing tools like the object inspector, whose sole purpose is to look inside
other objects.)
Instance variables can be accessed by name in any of the instance methods of
the class that defines them, and also in the methods defined in its subclasses.
This means that Pharo instance variables are similar to protected variables in
C++ and Java. However, we prefer to say that they are private, because it is
considered bad style in Pharo to access an instance variable directly from a
subclass.
Instance encapsulation example
The method Point>>dist: computes the distance between the receiver and
another point. The instance variables x and y of the receiver are accessed
directly by the method body. However, the instance variables of the other
point must be accessed by sending it the messages x and y.
Point >> dist: aPoint
"Answer the distance between aPoint and the receiver."
| dx dy |
dx := aPoint x - x.
dy := aPoint y - y.
87
The Pharo object model
^ ((dx * dx) + (dy * dy)) sqrt
1@1 dist: 4@5
>>> 5.0
The key reason to prefer instance-based encapsulation to class-based encapsulation is that it enables different implementations of the same abstraction
to coexist. For example, the method dist: need not know or care whether
the argument aPoint is an instance of the same class as the receiver. The
argument object might be represented in polar coordinates, or as a record in
a database, or on another computer in a distributed system. As long as it can
respond to the messages x and y, the code of method dist (shown above) will
still work.
Methods
All methods are public and virtual (i.e., dynamically looked up). Methods are
grouped into protocols that indicate their intent. Some common protocol
names have been established by convention, for example, accessing for all
accessor methods, and initialization for establishing a consistent initial
state for the object. The protocol private is sometimes used to group methods that should not be seen from outside. Nothing, however, prevents you
from sending a message that is implemented by such a ”private” method.
Methods can access all instance variables of the object. Some developers
prefer to access instance variables only through accessors. This practice
has some value, but it also clutters the interface of your classes, and worse,
exposes its private state to the world.
6.5 The instance side and the class side
Since classes are objects, they can have their own instance variables and their
own methods. We call these class instance variables and class methods, but they
are really no different from ordinary instance variables and methods: They
simply operate on different objects (classes in this case). An instance variable
describes instance state and a method describes instance behavior. Similarly,
class instance variables are just instance variables defined by a metaclass (and
describe the state of classes - instances of metaclasses), and class methods are
just methods defined by a metaclass (and that will be executed on classes).
A class and its metaclass are two separate classes, even though the former
is an instance of the latter. However, this is largely irrelevant to you as a
programmer: you are concerned with defining the behavior of your objects
and the classes that create them.
For this reason, the browser helps you to browse both class and metaclass as if
they were a single thing with two ”sides”: the instance side and the class side, as
shown in Figure 6.1. By default, when you select a class in the browser, you’re
88
6.5 The instance side and the class side
Figure 6.1: Browsing a class and its metaclass
browsing the instance side (i.e., the methods that are executed when messages
are sent to an instance of Color). Clicking on the Class side button switches
you over to the class side (the methods that will be executed when messages
are sent to the class Color itself).
For example, Color blue sends the message blue to the class Color. You will
therefore find the method blue defined on the class side of Color, not on the
instance side.
"Class-side method blue (convenient instance creation method)"
aColor := Color blue.
>>> Color blue
"Color instances are self-evaluating"
"Instance-side accessor method red (returns the red RGB value)"
Color blue red
>>> 0.0
"Instance-side accessor method blue (returns the blue RGB value)"
Color blue blue
>>> 1.0
You define a class by filling in the template proposed on the instance side.
When you accept this template, the system creates not just the class that you
defined, but also the corresponding metaclass (which you can then edit by
clicking on the Class side button). The only part of the metaclass creation
template that makes sense for you to edit directly is the list of the metaclass’s
instance variable names.
Once a class has been created, browsing its instance side (Class side unchecked) lets you edit and browse the methods that will be possessed by
instances of that class (and of its subclasses).
89
The Pharo object model
Class methods
Class methods can be quite useful; browse Color class for some good examples. You will see that there are two kinds of methods defined on a class:
instance creation methods, like Color class>>blue, and those that perform a
utility function, like Color class>>wheel:. This is typical, although you will
occasionally find class methods used in other ways.
It is convenient to place utility methods on the class side because they can be
executed without having to create any additional objects first. Indeed, many
of them will contain a comment designed to make it easy to execute them.
Browse method Color class>>wheel:, double-click just at the beginning of
the comment "(Color wheel: 12) inspect" and press CMD-d. You will see
the effect of executing this method.
For those familiar with Java and C++, class methods may seem similar to static
methods. However, the uniformity of the Pharo object model (where classes
are just regular objects) means that they are somewhat different: whereas
Java static methods are really just statically-resolved procedures, Pharo class
methods are dynamically-dispatched methods. This means that inheritance,
overriding and super-sends work for class methods in Pharo, whereas they
don’t work for static methods in Java.
Class instance variables
With ordinary instance variables, all the instances of a class have the same
set of variable names (though each instance has its own private set of values),
and the instances of its subclasses inherit those names. The story is exactly
the same with class instance variables: each class has its own private class
instance variables. A subclass will inherit those class instance variables, but
the subclass will have its own private copies of those variables. Just as objects don’t
share instance variables, neither do classes and their subclasses share class
instance variables.
For example, you could use a class instance variable called count to keep track
of how many instances you create of a given class. However, any subclass
would have its own count variable, so subclass instances would be counted
separately.
Example: Class instance variables and subclasses
Suppose we define the class Dog, and its subclass Hyena. Suppose that we add a
count class instance variable to the class Dog (i.e. we define it on the metaclass
Dog class). Hyena will naturally inherit the class instance variable count
from Dog.
Object subclass: #Dog
instanceVariableNames: ''
90
6.5 The instance side and the class side
classVariableNames: ''
package: 'PBE-CIV'
Dog class
instanceVariableNames: 'count'
Dog subclass: #Hyena
instanceVariableNames: ''
classVariableNames: ''
package: 'PBE-CIV'
Now suppose we define class methods for Dog to initialize its count to 0, and
to increment it when new instances are created:
Dog class >> initialize
count := 0.
Dog class >> new
count := count +1.
^ super new
Dog class >> count
^ count
Now when we create a new Dog, the count value of the class Dog is incremented, and so is that of the class Hyena (but the hyenas are counted separately).
Side note: Notice the use of initialize on the classes, in the following code. In
Pharo, when you instantiate an object such as Dog new, initialize is called
automatically as part of the new message send (you can see for yourself by
browsing Behavior>>new). But with classes, simply defining them does not
automatically call initialize, and so we have to call it explicitly here. By
default class initialize methods are automatically executed only when classes
are loaded. See also the discussion about lazy initialization, below.
Dog initialize.
Hyena initialize.
Dog count
>>> 0
Hyena count
>>> 0
| aDog |
aDog := Dog new.
Dog count
>>> 1 "Incremented"
Hyena count
>>> 0 "Still the same"
Class instance variables are private to a class in exactly the same way that
instance variables are private to an instance. Since classes and their instances
are different objects, this has the following consequences:
91
The Pharo object model
1. A class does not have access to the instance variables of its own instances.
So, the class Color does not have access to the variables of an object instantiated from it, aColorRed. In other words, just because a class was used to
create an instance (using new or a helper instance creation method like Color
red), it doesn’t give the class any special direct access to that instance’s variables. The class instead has to go through the accessor methods (a public
interface) just like any other object.
2. The reverse is also true: an instance of a class does not have access to the
class instance variables of its class. In our example above, aDog (an individual
instance) does not have direct access to the count variable of the Dog class
(except, again, through an accessor method).
Important A class does not have access to the instance variables of its
own instances. An instance of a class does not have access to the class
instance variables of its class.
For this reason, instance initialization methods must always be defined on the
instance side, the class side has no access to instance variables, and so cannot
initialize them! All that the class can do is to send initialization messages,
using accessors, to newly created instances.
Java has nothing equivalent to class instance variables. Java and C++ static
variables are more like Pharo class variables (discussed in Section 6.9), since in
those languages all of the subclasses and all of their instances share the same
static variable.
Example: Defining a Singleton
The Singleton pattern provides a typical example of the use of class instance
variables and class methods. Imagine that we would like to implement a class
WebServer, and to use the Singleton pattern to ensure that it has only one
instance.
We define the class WebServer as follow.
Object subclass: #WebServer
instanceVariableNames: 'sessions'
classVariableNames: ''
package: 'Web'
Then, clicking on the Class side button, we add the (class) instance variable
uniqueInstance.
WebServer class
instanceVariableNames: 'uniqueInstance'
As a result, the class WebServer class will have a new instance variable (in
addition to the variables that it inherits from Behavior, such as superclass
92
6.5 The instance side and the class side
and methodDict). It means that the value of this extra instance variable will
describe the instance of the class WebServer class i.e. the class WebServer.
Object class allInstVarNames
>>> "#('superclass' 'methodDict' 'format' 'layout' 'instanceVariables'
'organization' 'subclasses' 'name' 'classPool' 'sharedPools'
'environment' 'category' 'traitComposition' 'localSelectors')"
WebServer class allInstVarNames
>>>"#('superclass' 'methodDict' 'format' 'layout' 'instanceVariables'
'organization' 'subclasses' 'name' 'classPool' 'sharedPools'
'environment' 'category' 'traitComposition' 'localSelectors'
#uniqueInstance)"
We can now define a class method named uniqueInstance, as shown below.
This method first checks whether uniqueInstance has been initialized. If it
has not, the method creates an instance and assigns it to the class instance
variable uniqueInstance. Finally the value of uniqueInstance is returned.
Since uniqueInstance is a class instance variable, this method can directly
access it.
WebServer class >> uniqueInstance
uniqueInstance ifNil: [ uniqueInstance := self new ].
^ uniqueInstance
The first time that WebServer uniqueInstance is executed, an instance of the
class WebServer will be created and assigned to the uniqueInstance variable.
The next time, the previously created instance will be returned instead of
creating a new one. (This pattern, checking if a variable is nil in an accessor
method, and initializing its value if it is nil, is called lazy initialization).
Note that the instance creation code in the code above.
is written as self new and not as WebServer new. What is the difference?
Since the uniqueInstance method is defined in WebServer class, you might
think that there is no difference. And indeed, until someone creates a subclass of WebServer, they are the same. But suppose that ReliableWebServer
is a subclass of WebServer, and inherits the uniqueInstance method. We
would clearly expect ReliableWebServer uniqueInstance to answer a ReliableWebServer. Using self ensures that this will happen, since self will
be bound to the respective receiver, here the classes WebServer and ReliableWebServer. Note also that WebServer and ReliableWebServer will each
have a different value for their uniqueInstance instance variable.
Do not over-use the lazy initialization pattern. The
setting of initial values for instances of objects generally belongs in the initialize method. Putting initialization calls only in initialize helps from a
readability perspective – you don’t have to hunt through all the accessor methods to see what the initial values are. Although it may be tempting to instead
A note on lazy initialization.
93
The Pharo object model
initialize instance variables in their respective accessor methods (using ifNil:
checks), avoid this unless you have a good reason.
For example, in our uniqueInstance method above, we used lazy initialization
because users won’t typically expect to call WebServer initialize. Instead,
they expect the class to be ”ready” to return new unique instances. Because
of this, lazy initialization makes sense. Similarly, if a variable is expensive to
initialize (opening a database connection or a network socket, for example),
you will sometimes choose to delay that initialization until you actually need
it.
6.6 Every class has a superclass
Each class in Pharo inherits its behaviour and the description of its structure
from a single superclass. This means that Smalltalk has single inheritance.
SmallInteger superclass
>>> Integer
Integer superclass
>>> Number
Number superclass
>>> Magnitude
Magnitude superclass
>>> Object
Object superclass
>>> ProtoObject
ProtoObject superclass
>>> nil
Traditionally the root of an inheritance hierarchy is the class Object (since
everything is an object). In Pharo, the root is actually a class called ProtoObject, but you will normally not pay any attention to this class. ProtoObject
encapsulates the minimal set of messages that all objects must have and ProtoObject is designed to raise as many as possible errors (to support proxy
definition). However, most classes inherit from Object, which defines many
additional messages that almost all objects understand and respond to. Unless you have a very good reason to do otherwise, when creating application
classes you should normally subclass Object, or one of its subclasses.
A new class is normally created by sending the message subclass: instanceVariableNames: ... to an existing class. There are a few other methods to
create classes. To see what they are, have a look at Class and its subclass
creation protocol.
Although Pharo does not provide multiple inheritance, it supports a mechanism called Traits for sharing behaviour across unrelated classes. Traits are
94
6.6 Every class has a superclass
collections of methods that can be reused by multiple classes that are not
related by inheritance. Using traits allows one to share code between different
classes without duplicating code.
Abstract methods and abstract classes
An abstract class is a class that exists to be subclassed, rather than to be
instantiated. An abstract class is usually incomplete, in the sense that it does
not define all of the methods that it uses. The ”placeholder” methods, those
that the other methods assume to be (re)defined are called abstract methods.
Pharo has no dedicated syntax to specify that a method or a class is abstract.
Instead, by convention, the body of an abstract method consists of the expression self subclassResponsibility. This indicates that subclasses have the
responsibility to define a concrete version of the method. self subclassResponsibility methods should always be overridden, and thus should never
be executed. If you forget to override one, and it is executed, an exception will
be raised.
Similarly, a class is considered abstract if one of its methods is abstract. Nothing actually prevents you from creating an instance of an abstract class;
everything will work until an abstract method is invoked.
Example: the abstract class Magnitude
Magnitude is an abstract class that helps us to define objects that can be compared to each other. Subclasses of Magnitude should implement the methods
<, = and hash. Using such messages, Magnitude defines other methods such
as >, >=, <=, max:, min: between:and: and others for comparing objects. Such
methods are inherited by subclasses. The method Magnitude>>< is abstract,
and defined as shown in the following script.
< aMagnitude
"Answer whether the receiver is less than the argument."
^self subclassResponsibility
By contrast, the method >= is concrete, and is defined in terms of <.
>= aMagnitude
"Answer whether the receiver is greater than or equal to the
argument."
^(self < aMagnitude) not
The same is true of the other comparison methods (they are all defined in
terms of the abstract method <).
95
The Pharo object model
Character is a subclass of Magnitude; it overrides the < method (which, if you
recall, is marked as abstract in Magnitude by the use of self subclassResponsibility) with its own version (see the method definition below).
Character also explicitly defines methods = and hash; it inherits from Magnitude the methods >=, <=, ~= and others.
< aCharacter
"Answer true if the receiver's value < aCharacter's value."
^self asciiValue < aCharacter asciiValue
Traits
A trait is a collection of methods that can be included in the behaviour of a
class without the need for inheritance. This makes it easy for classes to have
a unique superclass, yet still share useful methods with otherwise unrelated
classes.
To define a new trait, simply right-click in the class pane and select Add
Trait, or replace the subclass creation template by the trait creation template, below.
Trait named: #TAuthor
uses: { }
package: 'PBE-LightsOut'
Here we define the trait TAuthor in the package PBE-LightsOut. This trait
does not use any other existing traits. In general we can specify a trait composition expression of other traits to use as part of the uses: keyword argument.
Here we simply provide an empty array.
Traits may contain methods, but no instance variables. Suppose we would like
to be able to add an author method to various classes, independent of where
they occur in the hierarchy.
We might do this as follows:
TAuthor >> author
"Returns author initials"
^ 'on' "oscar nierstrasz"
Now we can use this trait in a class that already has its own superclass, for
instance the LOGame class that we defined in Chapter : A First Application.
We simply modify the class creation template for LOGame to include a uses:
keyword argument that specifies that TAuthor should be used.
BorderedMorph subclass: #LOGame
uses: TAuthor
instanceVariableNames: 'cells'
classVariableNames: ''
96
6.7 Everything happens by sending messages
package: 'PBE-LightsOut'
If we now instantiate LOGame, it will respond to the author message as expected.
LOGame new author
>>> 'on'
Trait composition expressions may combine multiple traits using the + operator. In case of conflicts (i.e., if multiple traits define methods with the same
name), these conflicts can be resolved by explicitly removing these methods (with -), or by redefining these methods in the class or trait that you are
defining. It is also possible to alias methods (with @), providing a new name for
them.
Traits are used in the system kernel. One good example is the class Behavior.
Object subclass: #Behavior
uses: TPureBehavior @
{#basicAddTraitSelector:withMethod:->#addTraitSelector:withMethod:}
instanceVariableNames: 'superclass methodDict format'
classVariableNames: 'ObsoleteSubclasses'
package: 'Kernel-Classes'
Here we see that the method addTraitSelector:withMethod: defined in
the trait TPureBehavior has been aliased to basicAddTraitSelector:withMethod:.
6.7 Everything happens by sending messages
This rule captures the essence of programming in Pharo.
In procedural programming (and in some static features of some objectoriented languages such as Java), the choice of which piece of code to execute
when a procedure is called is made by the caller. The caller chooses the procedure to execute statically, by name.
In Pharo, we do not ”invoke methods”. Instead, we send messages. This is
just a terminology point but it is significant. It implies that this is not the
responsibility of the client to select the method to be executed, it is the one of
the receiver of the message.
When sending a message, we do not decide which method will be executed.
Instead, we tell an object to do something for us by sending it a message. A
message is nothing but a name and a list of arguments. The receiver then
decides how to respond by selecting its own method for doing what was asked.
Since different objects may have different methods for responding to the
same message, the method must be chosen dynamically, when the message is
received.
97
The Pharo object model
3 + 4
>>> 7 "send message + with argument 4 to integer 3"
(1@2) + 4
>>> 5@6 "send message + with argument 4 to point (1@2)"
As a consequence, we can send the same message to different objects, each of
which may have its own method for responding to the message. We do not tell
the SmallInteger 3 or the Point (1@2) how to respond to the message + 4.
Each has its own method for +, and responds to + 4 accordingly.
One of the consequences of Pharo’s model of message sending is that it encourages a style in which objects tend to have very small methods and delegate
tasks to other objects, rather than implementing huge, procedural methods
that assume too much responsibility. Joseph Pelrine expresses this principle
succinctly as follows:
”Don’t do anything that you can push off onto someone else.”
Many object-oriented languages provide both static and dynamic operations
for objects. In Pharo there are only dynamic message sends. For example,
instead of providing static class operations, we simply send messages to
classes (which are simply objects).
Ok, so nearly everything in Pharo happens by sending messages. At some point
action must take place:
Variable declarations are not message sends. In fact, variable declarations are
not even executable. Declaring a variable just causes space to be allocated for
an object reference.
Assignments are not message sends. An assignment to a variable causes that
variable name to be freshly bound in the scope of its definition.
Returns are not message sends. A return simply causes the computed result to
be returned to the sender.
Primitives (and Pragmas/annotations) are not message sends. They are implemented in the virtual machine.
Other than these few exceptions, pretty much everything else does truly
happen by sending messages. In particular, since there are no public fields
in Pharo, the only way to update an instance variable of another object is to
send it a message asking that it update its own field. Of course, providing
setter and getter methods for all the instance variables of an object is not good
object-oriented style. Joseph Pelrine also states this very nicely:
”Don’t let anyone else play with your data.”
98
6.8 Method lookup follows the inheritance chain
6.8 Method lookup follows the inheritance chain
What exactly happens when an object receives a message? This is a two step
process: method lookup and method execution.
Lookup. First, the method having the same name as the message is looked up.
Method Execution. Second, the found method is applied to the receiver with
the message arguments: When the method is found, the arguments are bound
to the parameters of the method, and the virtual machine executes it.
The lookup process is quite simple:
1. The class of the receiver looks up the method to use to handle the message.
2. If this class does not have that method method defined, it asks its superclass, and so on, up the inheritance chain.
It is essentially as simple as that. Nevertheless there are a few questions that
need some care to answer:
• What happens when a method does not explicitly return a value?
• What happens when a class reimplements a superclass method?
• What is the difference between self and super sends?
• What happens when no method is found?
The rules for method lookup that we present here are conceptual; virtual
machine implementors use all kinds of tricks and optimizations to speed up
method lookup. That’s their job, but you should never be able to detect that
they are doing something different from our rules.
First let us look at the basic lookup strategy, and then consider these further
questions.
Method lookup
Suppose we create an instance of EllipseMorph.
anEllipse := EllipseMorph new.
If we now send this object the message defaultColor, we get the result Color
yellow.
anEllipse defaultColor
>>> Color yellow
The class EllipseMorph implements defaultColor, so the appropriate method
is found immediately.
EllipseMorph >> defaultColor
"Answer the default color/fill style for the receiver"
^ Color yellow
99
The Pharo object model
Figure 6.2: Method lookup follows the inheritance hierarchy
In contrast, if we send the message openInWorld to anEllipse, the method
is not immediately found, since the class EllipseMorph does not implement
openInWorld. The search therefore continues in the superclass, BorderedMorph, and so on, until an openInWorld method is found in the class Morph
(see Figure 6.2).
Morph >> openInWorld
"Add this morph to the world."
self openInWorld: self currentWorld
Returning self
Notice that EllipseMorph>>defaultColor explicitly returns Color yellow,
whereas Morph>>openInWorld does not appear to return anything.
Actually a method always answers a message with a value (which is, of course,
an object). The answer may be defined by the ^ construct in the method, but if
execution reaches the end of the method without executing a ^, the method
still answers a value – it answers the object that received the message. We
usually say that the method answers self, because in Pharo the pseudo-variable
self represents the receiver of the message, much like the keyword this in
Java. Other languages, such as Ruby, by default return the value of the last
statement in the method. Again, this is not the case in Pharo, instead you can
imagine that a method without an explicit return ends with ^ self.
Important
self represents the receiver of the message.
This suggests that openInWorld is equivalent to openInWorldReturnSelf,
defined below.
100
6.8 Method lookup follows the inheritance chain
Morph >> openInWorld
"Add this morph to the world."
self openInWorld: self currentWorld
^ self
Why is explicitly writing ^ self not a so good thing to do? When you return
something explicitly, you are communicating that you are returning something of interest to the sender. When you explicitly return self, you are
saying that you expect the sender to use the returned value. This is not the
case here, so it is best not to explicitly return self. We only return self on
special case to stress that the receiver is returned.
This is a common idiom in Pharo, which Kent Beck refers to as Interesting
return value:
”Return a value only when you intend for the sender to use the value.”
Important By default (if not specified differently) a method returns the
message receiver.
Overriding and extension
If we look again at the EllipseMorph class hierarchy in Figure 6.2, we see that
the classes Morph and EllipseMorph both implement defaultColor. In fact,
if we open a new morph (Morph new openInWorld) we see that we get a blue
morph, whereas an ellipse will be yellow by default.
We say that EllipseMorph overrides the defaultColor method that it inherits
from Morph. The inherited method no longer exists from the point of view of
anEllipse.
Sometimes we do not want to override inherited methods, but rather extend
them with some new functionality, that is, we would like to be able to invoke
the overridden method in addition to the new functionality we are defining
in the subclass. In Pharo, as in many object-oriented languages that support
single inheritance, this can be done with the help of super sends.
A frequent application of this mechanism is in the initialize method. Whenever a new instance of a class is initialized, it is critical to also initialize any
inherited instance variables. However, the knowledge of how to do this is
already captured in the initialize methods of each of the superclass in
the inheritance chain. The subclass has no business even trying to initialize
inherited instance variables!
It is therefore good practice whenever implementing an initialize method
to send super initialize before performing any further initialization:
BorderedMorph >> initialize
"initialize the state of the receiver"
super initialize.
101
The Pharo object model
self borderInitialize
We need super sends to compose inherited behaviour that would otherwise be
overridden.
Important It is a good practice that an initialize method start by
sending super initialize.
Self sends and super sends
self represents the receiver of the message and the lookup of the method
starts in the class of the receiver. Now what is super? super is not the superclass! It is a common and natural mistake to think this. It is also a mistake to
think that lookup starts in the superclass of the class of the receiver.
Important self represents the receiver of the message and the method
lookup starts in the class of the receiver.
How do self sends differ from super sends?
Like self, super represents the receiver of the message. Yes you read it well!
The only thing that changes is the method lookup. Instead of lookup starting
in the class of the receiver, it starts in the superclass of the class of the method
where the super send occurs.
Important super represents the receiver of the message and the method
lookup starts in the superclass of the class of the method where the super
send occurs.
We shall see with the following example precisely how this works. Imagine
that we define the following three methods:
First we define the method fullPrintOn: on class Morph that just adds to the
stream the name of the class followed by the string ’ new’ - the idea is that
we could execute the resulting string and gets back an instance similar to the
receiver.
Morph >> fullPrintOn: aStream
aStream nextPutAll: self class name, ' new'
Second we define the method constructorString that send the message
fullPrintOn:.
Morph >> constructorString
^ String streamContents: [ :s | self fullPrintOn: s ].
Finally, we define the method fullPrintOn: on the class BorderedMorph superclass of EllipseMorph. This new method extends the superclass behavior:
it invokes it and adds extra behavior.
102
6.8 Method lookup follows the inheritance chain
Figure 6.3: self and super sends
BorderedMorph >> fullPrintOn: aStream
aStream nextPutAll: '('.
super fullPrintOn: aStream.
aStream
nextPutAll: ') setBorderWidth: ';
print: borderWidth;
nextPutAll: ' borderColor: ', (self colorString: borderColor)
Consider the message constructorString sent to an instance of EllipseMorph:
EllipseMorph new constructorString
>>> '(EllipseMorph new) setBorderWidth: 1 borderColor: Color black'
How exactly is this result obtained through a combination of self and super
sends? First, anEllipse constructorString will cause the method constructorString to be found in the class Morph, as shown in Figure 6.3.
The method Morph>>constructorString performs a self send of fullPrintOn:. The message fullPrintOn: is looked up starting in the class EllipseMorph, and the method BorderedMorph>>fullPrintOn: is found in
BorderedMorph (see Figure 6.3). What is critical to notice is that the self send
causes the method lookup to start again in the class of the receiver, namely
the class of anEllipse.
At this point, BorderedMorph>>fullPrintOn: does a super send to extend
the fullPrintOn: behaviour it inherits from its superclass. Because this is
a super send, the lookup now starts in the superclass of the class where the
super send occurs, namely in Morph. We then immediately find and evaluate
Morph>>fullPrintOn:.
103
The Pharo object model
Stepping back
A self send is dynamic in the sense that by looking at the method containing
it, we cannot predict which method will be executed. Indeed an instance of a
subclass may receive the message containing the self expression and redefine
the method in that subclass. Here EllipseMorph could redefine the method
fullPrintOn: and this method would be executed by method constructorString. Note that by only looking at the method constructorString, we
cannot predict which fullPrintOn: method (either the one of EllipseMorph,
BorderedMorph, or Morph) will be executed when executing the method constructorString, since it depends on the receiver the constructorString
message.
Important A self send triggers a method lookup starting in the class of the
receiver. A self send is dynamic in the sense that by looking at the method
containing it, we cannot predict which method will be executed.
Note that the super lookup did not start in the superclass of the receiver.
This would have caused lookup to start from BorderedMorph, resulting in an
infinite loop!
If you think carefully about super send and Figure 6.3, you will realize that
super bindings are static: all that matters is the class in which the text of the
super send is found. By contrast, the meaning of self is dynamic: it always
represents the receiver of the currently executing message. This means that
all messages sent to self are looked up by starting in the receiver’s class.
Important A super send triggers a method lookup starting in the superclass of the class of the method performing the super send. We say that super
sends are static because just looking at the method we know the class
where the lookup should start (the class above the class containing the
method).
Message not understood
What happens if the method we are looking for is not found?
Suppose we send the message foo to our ellipse. First the normal method
lookup would go through the inheritance chain all the way up to Object (or
rather ProtoObject) looking for this method. When this method is not found,
the virtual machine will cause the object to send self doesNotUnderstand:
#foo. (See Figure 6.4.)
Now, this is a perfectly ordinary, dynamic message send, so the lookup starts
again from the class EllipseMorph, but this time searching for the method
doesNotUnderstand:. As it turns out, Object implements doesNotUnderstand:. This method will create a new MessageNotUnderstood object which is
capable of starting a Debugger in the current execution context.
104
6.9 Shared variables
Figure 6.4: Message foo is not understood
Why do we take this convoluted path to handle such an obvious error? Well,
this offers developers an easy way to intercept such errors and take alternative action. One could easily override the method Object>>doesNotUnderstand: in any subclass of Object and provide a different way of handling the
error.
In fact, this can be an easy way to implement automatic delegation of messages from one object to another. A Delegator object could simply delegate
all messages it does not understand to another object whose responsibility it is
to handle them, or raise an error itself!
6.9 Shared variables
Now we will look at an aspect of Pharo that is not so easily covered by our five
rules: shared variables.
Pharo provides three kinds of shared variables:
1. Globally shared variables.
2. Class variables: variables shared between instances and classes. (Not to be
confused with class instance variables, discussed earlier).
3. Pool variables: variables shared amongst a group of classes,
The names of all of these shared variables start with a capital letter, to warn us
that they are indeed shared between multiple objects.
Global variables
In Pharo, all global variables are stored in a namespace called Smalltalk,
which is implemented as an instance of the class SystemDictionary. Global
variables are accessible everywhere. Every class is named by a global variable.
In addition, a few globals are used to name special or commonly useful objects.
105
The Pharo object model
The variable Processor names an instance of ProcessScheduler, the main
process schedler of Pharo
Processor class
>>> ProcessorScheduler
Other useful global variables
Smalltalk is the instance of SmalltalkImage. It contains many functionality
to manage the system. In particular it holds a reference to the main namespace Smalltalk globals. This namespace includes Smalltalk itself since it is
a global variable. The keys to this namespace are the symbols that name the
global objects in Pharo code. So, for example:
Smalltalk globals at: #Boolean
>>> Boolean
Since Smalltalk is itself a global variable:
Smalltalk globals at: #Smalltalk
>>> Smalltalk
(Smalltalk globals at: #Smalltalk) == Smalltalk
>>> true
World is an instance of PasteUpMorph that represents the screen. World
bounds answers a rectangle that defines the whole screen space; all Morphs on
the screen are submorphs of World.
ActiveHand is the current instance of HandMorph, the graphical representation
of the cursor. ActiveHand’s submorphs hold anything being dragged by the
mouse.
Undeclared is another dictionary, which contains all the undeclared variables.
If you write a method that references an undeclared variable, the browser will
normally prompt you to declare it, for example as a global or as an instance
variable of the class. However, if you later delete the declaration, the code will
then reference an undeclared variable. Inspecting Undeclared can sometimes
help explain strange behaviour!
Using globals in your code
The recommended practice is to strictly limit the use of global variables. It is
usually better to use class instance variables or class variables, and to provide
class methods to access them. Indeed, if Pharo were to be implemented from
scratch today, most of the global variables that are not classes would be
replaced by singletons.
The usual way to define a global is just to perform Do it on an assignment to a
capitalized but undeclared identifier. The parser will then offer to declare the
106
6.9 Shared variables
Figure 6.5: Instance and class methods accessing different variables
global for you. If you want to define a global programmatically, just execute
Smalltalk globals at: #AGlobalName put: nil. To remove it, execute
Smalltalk globals removeKey: #AGlobalName.
Class variables
Sometimes we need to share some data amongst all the instances of a class
and the class itself. This is possible using class variables. The term class variable
indicates that the lifetime of the variable is the same as that of the class.
However, what the term does not convey is that these variables are shared
amongst all the instances of a class as well as the class itself, as shown in
Figure 6.5. Indeed, a better name would have been shared variables since this
expresses more clearly their role, and also warns of the danger of using them,
particularly if they are modified.
In Figure 6.5 we see that rgb and cachedDepth are instance variables of Color,
hence only accessible to instances of Color. We also see that superclass,
subclass, methodDict and so on are class instance variables, i.e., instance
variables only accessible to the Color class.
But we can also see something new: ColorRegistry and CachedColormaps are
class variables defined for Color. The capitalization of these variables gives us
a hint that they are shared. In fact, not only may all instances of Color access
these shared variables, but also the Color class itself, and any of its subclasses.
Both instance methods and class methods can access these shared variables.
A class variable is declared in the class definition template. For example,
the class Color defines a large number of class variables to speed up color
creation; its definition is shown below.
Object subclass: #Color
instanceVariableNames: 'rgb cachedDepth cachedBitPattern alpha'
classVariableNames: 'BlueShift CachedColormaps ColorRegistry
ComponentMask GrayToIndexMap GreenShift HalfComponentMask
107
The Pharo object model
IndexedColors
MaskingMap RandomStream RedShift'
package: 'Graphics-Primitives'
The class variable ColorRegistry is an instance of IdentityDictionary
containing the frequently-used colors, referenced by name. This dictionary is
shared by all the instances of Color, as well as the class itself. It is accessible
from all the instance and class methods.
Class initialization
The presence of class variables raises the question: how do we initialize them?
One solution is lazy initialization (discussed earlier in this chapter). This can
be done by introducing an accessor method which, when executed, initializes
the variable if it has not yet been initialized. This implies that we must use the
accessor all the time and never use the class variable directly. This furthermore imposes the cost of the accessor send and the initialization test. It also
arguably defeats the point of using a class variable, since in fact it is no longer
shared.
Another solution is to override the class method initialize (we’ve seen this
before in the Dog example).
Color class >> initialize
...
self initializeColorRegistry.
...
If you adopt this solution, you will need to remember to invoke the initialize method after you define it (by evaluating Color initialize). Although
class side initialize methods are executed automatically when code is
loaded into memory (from a Monticello repository, for example), they are
not executed automatically when they are first typed into the browser and
compiled, or when they are edited and re-compiled.
Pool variables
Pool variables are variables that are shared between several classes that may
not be related by inheritance. Pool variables were originally stored in pool
dictionaries; now they should be defined as class variables of dedicated classes
(subclasses of SharedPool). Our advice is to avoid them; you will need them
only in rare and specific circumstances. Our goal here is therefore to explain
pool variables just enough so that you can understand them when you are
reading code.
A class that accesses a pool variable must mention the pool in its class definition. For example, the class Text indicates that it is using the pool dictionary
TextConstants, which contains all the text constants such as CR and LF. This
108
6.10
Internal object implementation note
dictionary has a key #CR that is bound to the value Character cr, i.e., the
carriage return character.
ArrayedCollection subclass: #Text
instanceVariableNames: 'string runs'
classVariableNames: ''
poolDictionaries: 'TextConstants'
package: 'Collections-Text'
This allows methods of the class Text to access the keys of the dictionary in
the method body directly, i.e., by using variable syntax rather than an explicit
dictionary lookup. For example, we can write the following method.
Text >> testCR
^ CR == Character cr
Once again, we recommend that you avoid the use of pool variables and pool
dictionaries.
6.10
Internal object implementation note
Here is an implementation note for people that really want to go deep inside
the way Pharo represents internally objects. The implementation distinguished between two different kinds of objects:
1. Objects with zero or more fields that are passed by reference and exist
on the Pharo heap.
2. Immediate objects that are passed by value. Depending on version,
these are a range of the integers called SmallInteger, all Character
objects and possibly a sub-range of 64-bit floating-point numbers called
SmallFloat64. In the implementation, such immediate objects occupy
an object pointer, most of whose bits encode the immediate’s value and
some of the bits encode the object’s class.
The first kind of object, an ordinary object, comes in a number of varieties:
1. Normal objects that have zero or more named instance variables, such
as Point which has an x and a y instance variable. Each instance variable holds an object pointer, which can be a reference to another ordinary object or an immediate.
2. Indexable objects like arrays that have zero or more indexed instance
variables numbered from 1 to N. Each indexed instance variable holds
an object pointer, which can be a reference to another ordinary object
or an immediate. Indexable objects are accessed using the messages at:
and at:put:. For example ((Array new: 1) at: 1 put: 2; at: 1)
answers 2.
109
The Pharo object model
3. Objects like Closure or Context that have both named instance variables
and indexed instance variables. In the object, the indexed instance
variables follow the named instance variables.
4. Objects like ByteString or Bitmap that have indexed instance variables
numbered from 1 to N that contain raw data. Each datum may occupy 8,
16 or 32-bits, depending on its class definition. The data can be accessed
as either integers, characters or floating-point numbers, depending on
how methods at: and at:put: are implemented. The at: and at:put:
methods convert between Pharo objects and raw data, hiding the internal representation, but allowing the system to represent efficiently data
such as strings, and bitmaps.
A beauty of Pharo is that you normally don’t need to care about the differences between these three kinds of object.
6.11
Chapter summary
The object model of Pharo is both simple and uniform. Everything is an object,
and pretty much everything happens by sending messages.
• Everything is an object. Primitive entities like integers are objects, but
also classes are first-class objects.
• Every object is an instance of a class. Classes define the structure of
their instances via private instance variables and the behaviour of their
instances via public methods. Each class is the unique instance of its
metaclass. Class variables are private variables shared by the class and
all the instances of the class. Classes cannot directly access instance variables of their instances, and instances cannot access instance variables
of their class. Accessors must be defined if this is needed.
• Every class has a superclass. The root of the single inheritance hierarchy
is ProtoObject. Classes you define, however, should normally inherit
from Object or its subclasses. There is no syntax for defining abstract
classes. An abstract class is simply a class with an abstract method (one
whose implementation consists of the expression self subclassResponsibility). Although Pharo supports only single inheritance, it is
easy to share implementations of methods by packaging them as traits.
• Everything happens by sending messages. We do not call methods, we
send messages. The receiver then chooses its own method for responding
to the message.
• Method lookup follows the inheritance chain; self sends are dynamic
and start the method lookup in the class of the receiver, whereas super
sends start the method lookup in the superclass of class in which the
super send is written. From this perspective super sends are more
static than self sends.
110
6.11
Chapter summary
• There are three kinds of shared variables. Global variables are accessible
everywhere in the system. Class variables are shared between a class,
its subclasses and its instances. Pool variables are shared between a
selected set of classes. You should avoid shared variables as much as
possible.
111
CHAPTER
7
Some of the key tools of the
Pharo environment
The goal of this chapter is to present the main tools of the Pharo programming
environment. You have already seen how to define methods and classes using
the browser; this chapter will show you more of its features, and introduce
you to some of the other browsers.
Of course, very occasionally you may find that your program does not work as
you expect. Pharo has an excellent debugger, but like most powerful tools, it
can be confusing on first use. We will walk you through a debugging session
and demonstrate some of the features of the debugger.
One of the unique features of Pharo (and its ancestors) is that while you
are programming, you are living in a world of live objects, not in a world of
static program text. This makes it possible to get very rapid feedback while
programming, which makes you more productive. There is a tool that let you
look at, and indeed change, live objects: the inspector.
7.1
Pharo environment overview
The System Browser is the central development tool. You will use it to create,
define, and organize your classes and methods. Using it you can also navigate
through all the library classes. Unlike other environments where the source
code is stored in separate files, in Pharo all classes and methods are contained
in the image.
The Finder tool will let you find methods, classes, pragmas, and more. You
can look for a method’s name, a class name, some source code, a pragma’s
113
Some of the key tools of the Pharo environment
name or even look for methods by providing an example!
The Monticello Browser is the starting point for loading code from, and
saving code in, Monticello packages. It is discussed in more detail in Chapter :
Sharing Code and Source Control.
The Process Browser provides a view of all of the processes (threads) executing in Smalltalk.
The Test Runner lets you run and debug SUnit tests, and is described in more
detail in Chapter : SUnit.
The Transcript is a window on the Transcript output stream, which is
useful for writing log messages.
The Playground is a window into which you can type input. It can be used for
any purpose, but is most often used for typing Pharo expressions and executing them via Do it. We have already briefly encountered the Playground (and
the Transcript) in Chapter : A Quick Tour of Pharo.
The Debugger has an obvious role, but you will discover that it has a more central place compared to debuggers for other programming languages, because
in Pharo you can program in the debugger. The debugger is not launched from
a menu; it is normally entered by running a failing test, by typing CMD-. to
interrupt a running process, or by inserting a Halt now expression in code.
Window groups
Managing multiple windows within a Pharo image can become a tedious
process. Window Groups are windows that offer tab support similar to the one
you are used to on your web browser. To create a window group, click on the
down arrow which appears on the top right corner of every window within
Pharo, and select Create window group. This will turn that window into a
window group with a tab bar, containing as its first tab the original contents of
the window. You can add other windows to the group (within the Pharo image
only, of course), by dragging and dropping their title bars onto the tab bar,
next to existing tabs. Each new window that you drag onto the tab bar will be
added as a new tab.
Themes and icon sets
If, like some of us, you are not a fan of the default bright theme, there is also
a dark theme that can be used with Pharo. It can be found in the Settings
browser in the World Menu (World > System > Settings), in the Appearance section. The User interface theme pulldown allows you to switch
between the default Pharo3 theme and the Pharo3 Dark theme. You can also
experiment with available icon sets in the next pulldown in that section.
Please note that you may have to close and reopen existing windows to redraw
them correctly with the new dark theme.
114
7.2 The main code browser
Figure 7.1: Window Group with two tabs, one with Playground and one with System
Browser
You can use the Catalog Browser to install additional themes. Open World
> Tools > Catalog Browser and search for theme in the search box. For
example, if you want something more colorful, you can download the Nireas
theme, which is a blue theme inspired by classic home computers like Amiga
500 and Amstrad CPC 6128. Nireas comes with a GUI tool that allows you to
customize the theme’s colors to your liking, if blue is not your thing.
Of course, if you are feeling especially adventurous, you can even make your
own themes using existing ones as templates.
7.2 The main code browser
Many different class browsers have been developed over the years for Pharo.
Pharo simplifies this story by offering a single browser that integrates various
views. Figure 7.2 shows the browser as it appears when you first open it.
The four small panes at the top of the browser represent a hierarchic view of
the methods in the system, much in the same way as the Mac OS X Finder in
column mode provide a view of the files on the disk. The leftmost pane lists
packages of classes; select one (say Kernel) and the pane immediately to the
right will then show all of the classes in that package.
Similarly, if you select one of the classes in the second pane, say, Boolean
(see Figure 7.3), the third pane will show all of the protocols defined for that
class, as well as a virtual protocol --all--. Protocols are a way of categorizing
methods; they make it easier to find and think about the behaviour of a class
by breaking it up into smaller, conceptually coherent pieces. The fourth pane
shows the names of all of the methods defined in the selected protocol. If you
then select a method name, the source code of the corresponding method
115
Figure 7.2: The main code browser.
Figure 7.3: The browser with the class Boolean selected.
7.2 The main code browser
Figure 7.4: Browsing the or: method in class Boolean.
appears in the large pane at the bottom of the browser, where you can view
it, edit it, and save the edited version. If you select class Boolean, protocol
controlling and the method or:, the browser should look like Figure 7.4.
Unlike directories in a file browser, the four top panes of the browser are not
quite equal. Whereas classes and methods are part of the Smalltalk language,
packages and protocols are not: they are a convenience introduced by the
browser to limit the amount of information that needs to be shown in each
pane. For example, if there were no protocols, the browser would have to
show a list of all of the methods in the selected class; for many classes this list
would be too large to navigate conveniently.
Because of this, the way that you create a new package or a new protocol is
different from the way that you create a new class or a new method. To create
a new package, right-click in the package pane and select Add package....
To create a new protocol, right-click in the protocol pane and select Add
protocol.... Enter the name of the new thing in the dialog, and you are
done: there is nothing more to a package or a protocol than its name and its
contents.
In contrast, to create a new class or a new method, you will actually have to
write some Smalltalk code. If you click the currently selected package (in
the left-most pane), the bottom browser pane will display a class creation
117
Some of the key tools of the Pharo environment
Figure 7.5: Browser showing the class-creation template
template (Figure 7.5). You create a new class by editing this template: replace
Object by the name of the existing class of which you wish to create a subclass, replace NameOfSubclass by the name that you would like to give to your
new subclass, and fill in the instance variable names if you know them. The
package for the new class is by default the currently selected package, but you
can change this too if you like. If you already have the browser focused on the
class that you wish to subclass, you can get the same template with slightly
different initialization by right-clicking in the class pane, and selecting Add
Class. You can also just edit the definition of an existing class, changing the
class name to something new. In all cases, when you accept the new definition, the new class (the one whose name follows the #) is created (as is the
corresponding metaclass). Creating a class also creates a global variable that
references the class, which is why you can refer to all of the existing classes by
using their names.
Can you see why the name of the new class has to appear as a Symbol (i.e.,
prefixed with #) in the class creation template, but after the class is created,
code can refer to the class by using the name as an identifier (without the #)?
The process of creating a new method is similar. First select the class in which
you want the method to live, and then select a protocol. The browser will
display a method-creation template, as shown in Figure 7.6, which you can
fill-in or edit.
118
7.2 The main code browser
Figure 7.6: Showing the method-creation template.
Navigating the code space
The browser provides several tools for exploring and analyzing code. These
tools can be accessed by right-clicking in the various contextual menus, or, in
the case of the most frequently used tools, by means of keyboard shortcuts.
Opening a new browser window
Sometimes you want to open multiple browser windows. When you are writing code you will almost certainly need at least two: one for the method that
you are typing, and another to browse around the system to see how things
work. You can open a browser on a class named by any selected text using the
CMD-b keyboard shortcut.
To do Try this: In a playground window, type the name of a class (for
instance Morph), select it, and then press CMD-b. This trick is often useful; it
works in any text window.
Message senders
While writing new code, refactoring existing code, or while trying to learn
how to use unfamiliar libraries, you will frequently want to know the senders
119
Some of the key tools of the Pharo environment
Figure 7.7: The Senders Of...(b,n) menu item.
and implementors of various messages.
There are several ways of discovering where in the codebase a message is used,
by listing its senders:
1. From the method pane. Select a method in the method pane of the browser.
You can then right-click on it and select Senders of...(b,n) in the context menu. Alternatively, you can also use the shortcut CMD-b CMD-n to do
the same thing (that’s what the b,n in the menu item stands for). To help
remember this shortcut, think: browse senders.
2. From the code pane. Highlight a particular message in the source code. This
can be done in a code pane of a browser, in a Playground window, or in any
text window. If you want to find the senders of a keyword message, you
highlight all of the keywords in the message, including arguments. Then,
you can right-click on the highlighted selector and choose Code search... >
senders of it (n). Alternatively, you can use the shortcut CMD-n instead of
right-clicking.
3. Using Spotter. Bring up a particular method in Spotter (press SHIFT-Enter
to bring up the Spotter search box, type in the message selector, arrow down
to a particular Implementor of that message, and press CMD-right arrow to
focus the search on it.) A list of Senders now appears in the search results.
120
7.2 The main code browser
Figure 7.8: The Senders Browser showing that the Canvas>>draw method sends
the drawOn: message to its argument.
Only a handful of senders are shown by default, but you can view the full list
by clicking on the arrow next to the Senders category (or arrow down to the
Senders list and expand it by pressing CMD-SHIFT-right arrow).
Let’s try some of these in action.
Open a browser on the Morph class, select the Morph>>drawOn: method in the
method pane. If you now press CMD-b CMD-n (or right-click in the method
pane and select Senders of... (Figure 7.7)), a browser will open with the list
of all methods in the image that send the selected message (Figure 7.8).
Now look at the third sender on the list, Canvas>>draw:. You can see that
this method sends drawOn: to whatever object is passed to it as an argument,
which could potentially be an instance of any class at all. Dataflow analysis
can help figure out the class of the receiver of some messages, but in general,
there is no simple way for the browser to know which message-sends might
cause which methods to be executed. For this reason, the Senders browser
shows exactly what its name suggests: all of the senders of the message with
the chosen selector. The senders browser is nevertheless extremely useful
when you need to understand how you can use a method: it lets you navigate
quickly through example uses. Since all of the methods with the same selector
121
Some of the key tools of the Pharo environment
Figure 7.9: Finding senders of a keyword message in the code pane.
should be used in the same way, all of the uses of a given message ought to be
similar.
To do Switch to the Dictionary class in the browser (remember, you
can right-click in the package or class pane and select Find class..., or
just use the CMD-f CMD-c shortcut), and select the addAll: method in the
method pane.
Looking at the source code, suppose you wanted to see all of the senders of the
at: key put: value message. You can simply highlight the whole message
send, and press CMD-n (or right-click and select Code Search > senders of
it (n)), to bring up the list of senders (see Figure 7.9).
Message implementors
Similarly, you may come across a message, and want to see how it’s implemented. This is what the Implementors browser is for. It works in the same
way as the Senders browser, but instead lists all of the classes that implement
a method with the same selector.
1. From the method pane. Select a method in the method pane. You can then
bring up the Implementors browser by right-clicking on the method and
selecting Implementors of...(b,m) in the context menu (or use the shortcut
CMD-b CMD-m). To help remember this shortcut, think: browse implementors.
2. From the code pane. Highlight a particular message in the source code (or any
text window). If you want to find the implementors of a keyword message, you
highlight all of the keywords in the message. Then, you can right-click on the
highlited selector and choose Code search... > implementors of it (m)
from the menu (or just use the shortcut CMD-n).
3. Using Spotter. Bring up a method in Spotter (press SHIFT-Enter to bring up
the Spotter search box, and start typing the message selector). The Implemen122
7.2 The main code browser
tors category will show up in the search results, showing the first handful of
implementors. To see the full list, click on the arrow to the right of Implementors category (or arrow down to Implementors and press SHIFT-CMD-right
arrow).
Try this out: Press SHIFT-Enter and type drawOn: in the Spotter search box.
You should see a list showing 5 out of 100 implementors of that method. It
shouldn’t be all that surprising that so many classes implement this method:
drawOn: is the message that is understood by every object that is capable of
drawing itself on the screen.
Notice that if you only typed drawOn and left out the colon (:), the number of
implementors in the search results is larger. This is because Spotter is doing
a partial search, and including any methods that have ’drawOn’ in the name,
such as drawOn:offset:, drawOnAthensCanvas:, and so on. This is useful for
when you want to find a method but can only remember a part of its name.
Method inheritance and overriding
The inheritance browser displays all the methods overridden by the displayed
method. To see how it works, select the ImageMorph>>drawOn: method in the
browser. Note the arrow icons next to the method name (Figure 7.10). The
upward-pointing arrow tells you that ImageMorph>>drawOn: overrides an
inherited method (i.e., Morph>>drawOn:), and the downward-pointing arrow
tells you that it is overridden by subclasses. (You can also click on the icons
to navigate to these methods.) Now right-click on it in the method pane, and
select Inheritance. The inheritance browser shows you the hierarchy of
overridden methods (see Figure 7.10).
Hierarchy view
By default, the browser presents a list of packages in the leftmost pane. However it is possible to switch to a class hierarchy view. Simply select a particular
class of interest, such as ImageMorph and then click on the Hierarchy button
(Hier.). You will then see in the second pane a class hierarchy displaying all
superclasses and subclasses of the selected class.
Notice that the package pane is disabled, and the packages are greyed out.
When you are in Hierarchy view, you cannot change packages. To be able
to change them again, toggle out of the Hierarchy view by clicking on the
Hierarchy button again.
In Figure 7.11, the hierarchy view reveals that the direct superclass of ImageMorph is Morph.
123
Some of the key tools of the Pharo environment
Figure 7.10: ImageMorph>>drawOn: and the hierarchy of classes overriding it.
Finding variable references
By right-clicking on a class in the class pane, and selecting Analyze > Inst
var references... or Analyze > Class var refs..., you can find out
where an instance variable or a class variable is used. You can also have access
to those views by clicking on the Variables button, above the package list.
Once you click on the button or select the menu item, you will be presented
with a dialog that invites you to choose a variable from all of the variables
defined in the current class, and all of the variables that it inherits. The list is
in inheritance order; it can often be useful to bring up this list just to remind
yourself of the name of an instance variable. If you click outside the list, it
will go away and no variable browser will be created. If you click on a variable,
bounds for example, a Message Browser will be created (Figure 7.12).
You can use a similar method to look at direct variable assignments (that is,
places that modify the variable without using accessor methods). Right-click
on the class and select Analyze > Inst var assignments.
124
7.2 The main code browser
Figure 7.11: A hierarchy view of ImageMorph.
Bytecode source
You have the possibility of browsing the bytecode of a method. To do that,
right-click on your method and select Toggle Bytecodes, or use the shortcut
CMD-b CMD-b (see Figure 7.13). Reselect the method again to get back to the
normal view.
Refactorings
The contextual menus offer a large number of standard refactorings. Simply
right-click in any of the four panes to see the currently available refactoring
operations. See Figure 7.14.
Refactoring was formerly available only in a special browser called the refactoring browser, but it can now be accessed from any browser.
Browser menus
Many additional functions are available by right-clicking in the browser
panes. Even if the labels on the menu items are the same, their meaning may
be context dependent. For example, the package pane, the class pane, the
protocol pane and the method pane all have a File out menu item. However,
they do different things: the package pane’s File out menu item files out the
125
Some of the key tools of the Pharo environment
Figure 7.12: A Message Browser for accesses to bounds variable of Morph.
whole package, the class pane’s item files out the whole class, the protocol
pane’s item files out the whole protocol, and the method pane’s item files out
just the displayed method.
Although this may seem obvious, it can be a source of confusion for beginners.
Possibly the most useful menu item is Find class... (f,c) in the package
or class panes. Most of us do not know the package contents of the whole
system, and it is much faster to type CMD-f CMD-c followed by the first few
characters of the name of a class than to guess which package it might be in.
The History Navigator pulldown, found above the protocol and method
panes, can also help you quickly go back to a class or method that you have
browsed recently, even if you have forgotten its name.
Another useful method in the class pane is Find method (CMD-f CMD-m), which
brings up a menu of all the methods in the class and gives you a search box.
Alternatively, if you are searching for a particular method of the selected class,
it is often quicker to browse the --all-- protocol, place the mouse in the
method pane, and type the first letter of the name of the method that you are
looking for. This will usually scroll the pane so that the sought-for method
126
Figure 7.13: Bytecode of the ImageMorph»#DrawOn: compiled method.
Figure 7.14: Refactoring operations.
Some of the key tools of the Pharo environment
name is visible.
To do
Try both ways of navigating to OrderedCollection>>removeAt:
There are many other options available in the menus. It pays to spend a few
minutes working with the browser and seeing what is there.
Browsing programmatically
The class SystemNavigation provides a number of utility methods that are
useful for navigating around the system. Many of the functions offered by the
system browser are implemented by SystemNavigation.
Open a playground and evaluate the following code to browse the senders of
drawOn::
SystemNavigation default browseAllSendersOf: #drawOn:
To restrict the search for senders to the methods of a specific class:
SystemNavigation default browseAllSendersOf: #drawOn: localTo:
ImageMorph
Because the development tools are objects, they are completely accessible
from programs and you can develop your own tools or adapt the existing tools
to your needs.
The programmatic equivalent to the Implementors of... menu item is:
SystemNavigation default browseAllImplementorsOf: #drawOn:
To do To learn more about what is available, explore the class SystemNavigation
with the browser.
7.3 The inspector
One of the things that makes Pharo so different from many other programming environments is that it provides you with a window onto a world of live
objects, not a world of static code. Any of those objects can be examined by
the programmer, and even modified (although some care is necessary when
changing the basic objects that support the system). By all means experiment,
but save your image first!
As an illustration of what you can do with an inspector, type DateAndTime now
in a playground, and then right-click and choose Inspect it (CMD-i) or Do it
and go (CMD-g) (the latter opens an inspector inside the playground window).
128
7.3 The inspector
Figure 7.15: Inspecting DateAndTime now.
Note that it’s often not necessary to select the text before using the menu; if
no text is selected, the menu operations work on the whole of the current line.
A window like that shown in Figure 7.15 will appear. This is an inspector, and
can be thought of as a window onto the internals of a particular object – in
this case, the particular instance of DateAndTime that was created when you
evaluated the expression DateAndTime now. The title bar of the window shows
the printable representation of the object that is being inspected.
In the default view (the Raw tab), instance variables can be explored by selecting them in the variable list in the Variable column. As you select a variable,
its printable representation is shown in the Value column. More importantly,
a separate Inspector view for the selected variable opens in the right hand
pane.
For variables that are simple types (booleans, integers, etc), the nested inspector view is not much different from the printable representation in the
Value column (although it is a full-fledged Inspector). But for most instance
variables, the nested Inspector view on the right has its own Raw tab, with its
own list of instance variables. (You can also see that list in the left pane, by
expanding the triangle next to a variable’s name.)
You can keep drilling down into the hierarchy of instance variables, with
more nested Inspector panes opening to the right of the parent. However,
to prevent the multiple panes from being impractical, the panes ”scroll”
to the right, within the overall inspector window. You can keep track of
129
Some of the key tools of the Pharo environment
which ”page” you’re on, and also back-track to the original instance that you
were inspecting, by using the pagination dots at the bottom of the inspector
window.
There are special variants of the inspector for Dictionaries, OrderedCollections, CompiledMethods and a few other classes. These variants have other
tabs, in addition to the Raw view, that make it easier to examine the contents
of these special objects. For example, an inspector on a Dictionary instance,
has an additional Items tab that shows that dictionary’s keys and values in an
intuitive fashion.
The horizontal pane at the bottom of the inspector is a small playground
window. It is useful because in this window, the pseudo-variable self is bound
to the object that you have selected in the left pane. That means you can write
and evaluate arbitrary code expressions that use the selected variable’s self,
in that bottom pane.
For example, take the inspector on DateAndTime now that you opened earlier
in this section. You can select its bottom playground pane, and evaluate the
expression self - DateAndTime today. The result will be a Duration object
that represents the time interval between midnight today and the instant
at which you evaluated DateAndTime now and created the DateAndTime instance that you are inspecting. You can also try evaluating DateAndTime now
- self; this will tell you how long you have spent reading this section of this
book!
The bottom pane is especially useful if you wanted to change the instance
variables of the object being inspected. Provided that you have accessor
methods defined for those variables, you can send messages to the root self
and change its variables via those accessor methods.
7.4
The debugger
The debugger is arguably the most powerful tool in the Pharo tool suite. It is
used not just for debugging, but also for writing new code. To demonstrate the
debugger, let’s start by creating a bug!
Using the browser, add the following method to the class String:
suffix
"assumes that I'm a file name, and answers my suffix, the part after
the last dot"
| dot dotPosition |
dot := '.'.
dotPosition := (self size to: 1 by: -1) detect: [ :i | (self at: i)
= dot ].
^ self copyFrom: dotPosition to: self size
130
7.4
The debugger
Figure 7.16: A PreDebugWindow notifies us of a bug.
Of course, we are sure that such a trivial method will work, so instead of
writing an SUnit test, we just type 'readme.txt' suffix in a playground and
Print it (p). What a surprise! Instead of getting the expected answer 'txt',
a PreDebugWindow pops up, as shown in Figure 7.16.
The PreDebugWindow has a title bar that tells us what error occurred, and
shows us a stack trace of the messages that led up to the error. Starting from
the bottom of the trace, UndefinedObject>>DoIt represents the code that
was compiled and run when we selected 'readme.txt' suffix in the playground and asked Pharo to Print it. This code, of course, sent the message
suffix to a ByteString object ('readme.txt'). This caused the inherited suffix method in class String to execute; all this information is encoded in the
next line of the stack trace, ByteString(String)>>suffix. Working up the
stack, we can see that suffix sent detect:... and eventually detect:ifNone
sent errorNotFound:.
To find out why the dot was not found, we need the debugger itself, so click
on Debug. You can also open the debugger by clicking on any of the lines on
the stack trace. If you do this, the debugger will open already focused on the
corresponding method.
The debugger is shown in Figure 7.17; it looks intimidating at first, but it is
quite easy to use. The title bar and the top pane are very similar to those
that we saw in the PreDebugWindow. However, the debugger combines the
stack trace with a method browser, so when you select a line in the stack
trace, the corresponding method is shown in the pane below. It’s important
to realize that the execution that caused the error is still in your image, but
in a suspended state. Each line of the stack trace represents a frame on the
131
Some of the key tools of the Pharo environment
Figure 7.17: The debugger showing the execution stack and the state of various
objects.
execution stack that contains all of the information necessary to continue the
execution. This includes all of the objects involved in the computation, with
their instance variables, and all of the temporary variables of the executing
methods.
In Figure 7.17 we have selected the detect:ifFound:IfNone: method in
the top pane. The method body is displayed in the center pane; the blue
highlight around the message value shows that the current method has sent
the message value and is waiting for an answer.
The Variables pane at the bottom of the debugger is actually like a inspector
(without playground pane). You can select one variable and open another inspector pane with the well-known inspector panes (Raw, ...), and an Evaluator
acting as a Playground for evaluating code within the context of the selected
variable. The variables list has up to four types variables it shows.
• parameter any parameter passed to this method.
• temp any temporaries used in this method.
• attribute any instance variable accessible from the context of the current receiver.
• implicit pseudo variables (self, thisContext, stackTop) used in the
132
7.4
The debugger
current context.
As you select different stack frames, the identity of self may change, and
so will the contents of the Variables list. If you click on self in the bottomleft pane, you will see that self is the interval (10 to: 1 by -1), which is
what we expect. You can always select self and select the Evaluator pane to
evaluate some code in the content of the current receiver. But because all of
the variables are also in scope in the method pane; you should feel free to type
or select expressions directly in the method pane and evaluate them. You can
always Cancel (l) your changes using the menu or CMD-l.
Selecting thisContext from the list of (implicit) variables, shows the current
context object.
As we can see one method lower in the stack trace, the exceptionBlock is
[self errorNotFound: ...], so, it is not surprising that we see the corresponding error message.
Incidentally, if you want to open a full inspector on one of the variables shown
in the mini-inspectors, just double-click on the name of the variable, or select
the name of the variable and right-click to choose Inspect (i). This can be
useful if you want to watch how a variable changes while you execute other
code.
Looking back at the method window, we see that we expected the penultimate
line of the method to find '.' in the string 'readme.txt', and that execution
should never have reached the final line. Pharo does not let us run an execution backwards, but it does let us start a method again, which works very well
in code that does not mutate objects, but instead creates new ones.
Click Restart, and you will see that the focus of execution returns to the first
statement of the current method. The blue highlight shows that the next
message to be sent will be do: (see Figure 7.18).
The Into and Over buttons give us two different ways to step through the
execution. If you click Over, Pharo executes the current message-send (in
this case the do:) in one step, unless there is an error. So Over will take us to
the next message-send in the current method, which is value – this is exactly
where we started, and not much help. What we need to do is to find out why
the do: is not finding the character that we are looking for.
After clicking Over, click Restart to get back to the situation shown in Figure
7.18.
Click Into two times; Pharo will go into the method corresponding to the
highlighted message-send, in this case, Interval>>do:.
However, it turns out that this is not much help either; we can be fairly confident that Interval>>do: is not broken. The bug is much more likely to be
in what we asked Pharo to do. Through is the appropriate button to use in this
133
Some of the key tools of the Pharo environment
Figure 7.18: The debugger after restarting the detect:ifFound:IfNone:
method.
case: we want to ignore the details of the do: itself and focus on the execution
of the argument block.
Select the detect:ifFound:IfNone: method again and Restart to get back to
the state shown in Figure 7.18. Now click on Through a few times. Select each
in the context window as you do so. You should see each count down from 10
as the do: method executes.
When each is 7 we expect the ifTrue: block to be executed, but it isn’t. To see
what is going wrong, go Into the execution of value: as illustrated in Figure
7.19.
After clicking Into, we find ourselves in the position shown in Figure 7.20. It
looks at first that we have gone back to the suffix method, but this is because
we are now executing the block that suffix provided as argument to detect:.
If you select dot in the context inspector, you will see that its value is '.'.
And now you see why they are not equal: the seventh character of 'readme.txt'
134
7.4
The debugger
Figure 7.19: The debugger after stepping Through the do: method several times.
is of course a Character, while dot is a String.
Now that we see the bug, the fix is obvious: we have to convert dot to a character before starting to search for it.
Change the code right in the debugger so that the assignment reads dot := $.
and accept the change.
Because we are executing code inside a block that is inside a detect:, several
stack frames will have to be abandoned in order to make this change. Pharo
asks us if this is what we want (see Figure 7.21), and, assuming that we click
yes, will save (and compile) the new method.
The evaluation of the expression 'readme.txt' suffix will complete, and
print the answer '.txt'.
Is the answer correct? Unfortunately, we can’t say for sure. Should the suffix
be .txt or txt? The method comment in suffix is not very precise. The way
to avoid this sort of problem is to write an SUnit test that defines the answer.
135
Some of the key tools of the Pharo environment
Figure 7.20: The debugger showing why 'readme.txt' at: 7 is not equal to dot.
testSuffixFound
self assert: 'readme.txt' suffix = 'txt'
The effort required to do that was little more than to run the same test in the
playground, but using SUnit saves the test as executable documentation, and
makes it easy for others to run. Moreover, if you add testSuffix to the class
StringTest and run that test suite with SUnit, you can very quickly get back
to debugging the error. SUnit opens the debugger on the failing assertion,
but you need only go back down the stack one frame, Restart the test and
go Into the suffix method, and you can correct the error, as we are doing
in Figure 7.22. It is then only a second of work to click on the Run Failures
button in the SUnit Test Runner, and confirm that the test now passes.
Here is a better test:
testSuffixFound
self assert: 'readme.txt' suffix = 'txt'.
self assert: 'read.me.txt' suffix = 'txt'
136
7.4
The debugger
Figure 7.21: Changing the suffix method in the debugger: asking for confirmation
of the exit from an inner block.
Why is this test better? Because it tells the reader what the method should do
if there is more than one dot in the target String.
There are a few other ways to get into the debugger in addition to catching
errors and assertion failures. If you execute code that goes into an infinite
loop, you can interrupt it and open a debugger on the computation by typing
CMD-. (that’s a full stop or a period, depending on where you learned English).
(It is also useful to know that you can bring up an emergency debugger at
any time by typing CMD-SHIFT-.) You can also just edit the suspect code to
insert Halt now.. So, for example, we might edit the suffix method to read
as follows:
suffix
"assumes that I'm a file name, and answers my suffix, the part after
the last dot"
| dot dotPosition |
dot := FileDirectory dot first.
dotPosition := (self size to: 1 by: -1) detect: [ :i | (self at: i)
= dot ].
Halt now.
^ self copyFrom: dotPosition to: self size
137
Some of the key tools of the Pharo environment
Figure 7.22: Changing the suffix method in the debugger: fixing the off-by-one
error after an SUnit assertion failure.
When we run this method, the execution of the Halt now will bring up the
pre-debugger, from where we can either proceed, or go into the debugger (and
from there look at variables, step through the computation, and edit the code).
That’s all there is to the debugger, but it’s not all there is to the suffix
method. The initial bug should have made you realize that if there is no
dot in the target string, the suffix method will raise an error. This isn’t the
behaviour that we want, so let’s add a second test to specify what should
happen in this case.
testSuffixNotFound
self assert: 'readme' suffix = ''
Lastly, add testNoSuffix to the test suite in class StringTest, and watch the
test raise an error. Enter the debugger by selecting the erroneous test in SUnit,
and edit the code so that the test passes. The easiest and clearest way to do
this is to replace the detect: message by detect:ifNone:, where the second
argument is a block that simply returns the string size.
We will learn more about SUnit in Chapter : SUnit.
138
7.5 The process browser
Figure 7.23: The Process Browser showing all the active thread by priority.
7.5 The process browser
Pharo is a multi-threaded system, and there are many lightweight processes
(also known as threads) running concurrently in your image. In the future the
Pharo virtual machine may take advantage of multiple processors when they
are available, but at present, concurrency is implemented by time-slicing.
The Process Browser is a cousin of the debugger that lets you look at the
various processes running inside Pharo. You can open it using the World Menu,
by selecting Tools > Process Browser (figure 7.23 shows a screenshot). The
top-left pane lists all of the processes in Pharo, in priority order, from the
timer interrupt watcher at priority 80 to the idle process at priority 10. Of
course, on a uniprocessor, the only process that can be running when you look
is the UI process; all others will be waiting for some kind of event.
By default, the display of processes is static; it can be updated by right-clicking
and selecting Turn on auto-update (a).
If you select a process in the top-left pane, its stack trace is displayed in the
top-right pane, just as with the debugger. If you select a stack frame, the
corresponding method is displayed in the bottom pane. The process browser
is not equipped with mini-inspectors for self and thisContext, but rightclicking on the stack frames provide equivalent functionality.
139
Some of the key tools of the Pharo environment
7.6 Finding methods
The Finder is one of several code search tools in Pharo to help you find methods by name (or even functionality). We’ve discussed it in some length in
Chapter : A Quick Tour of Pharo.
7.7 Chapter summary
In order to develop effectively with Pharo, it is important to invest some effort
into learning the tools available in the environment.
• The standard browser is your main interface for browsing existing
packages, classes, method protocols and methods, and for defining new
ones.
• The browser offers several useful shortcuts to directly jump to senders
or implementors of a message, versions of a method, and so on.
• From any of the tools, you can highlight the name of a class or a method
and immediately jump to a browser by using the keyboard shortcut
CMD-b.
• You can also browse the Pharo system programmatically by sending
messages to SystemNavigation default.
• The Inspector is a tool that is useful for exploring and interacting with
live objects in your image.
• You can even inspect tools by meta-clicking to bring up their morphic
halo and selecting the debug handle.
• The Debugger is a tool that not only lets you inspect the run-time stack
of your program when an error is raised, but it also enables you to
interact with all of the objects of your application, including the source
code. In many cases you can modify your source code from the debugger
and continue executing. The debugger is especially effective as a tool to
support test-first development in tandem with SUnit (Chapter : SUnit).
• The Process Browser lets you monitor, query and interact with the processes current running in your image.
• The Finder is a tool for locating methods.
140
CHAPTER
8
Sharing code and source control
The consequence of programming in a world of live objects rather than with
files and a text editor is that you have to do something explicit to export your
program from your Pharo image. The old way of doing this is by creating
a fileout or a change set, which are essentially encoded text files that can be
imported into another system. The preferred way of sharing code in Pharo
is to save packages and share them using a versioned repository on a server.
Up to Pharo 50, this is done using a tool called Monticello, and is a much more
powerful and effective way to work, especially when working in a team. In the
future Pharo will officially offer support for Git and hosting on servers such as
GitHub, BitBucket and more.
We gave you a quick overview of Monticello, Pharo’s packaging tool, in Chapter : A First Application. However, Monticello has many more features than
were discussed there. Because Monticello manages packages, before telling
you more about Monticello, it’s important that we first explain exactly what a
package is.
8.1
Packages: groups of classes and methods
We have pointed out earlier (in Chapter : A First Application) packages are
more or less a group of classes and methods. Now we will see exactly what the
relationship is. The package system is a simple, lightweight way of organizing
Pharo source code that exploits a simple naming convention.
Let’s explain this using an example. Suppose that you are developing a framework to facilitate the use of relational databases from Pharo. You have decided to call your framework PharoLink, and have created a couple of classes
such as OracleConnection, MySQLConnection, PostgresConnection but also
141
Sharing code and source control
DBTable, DBRow, DBQuery, and so on. These classes are placed inside a package
called PharoLink. However, not all of your code will reside in these classes.
For example, you may also have a series of methods to convert objects into an
SQL-friendly format: Object>>#asSQL, String>>#asSQL, or Date>>#asSQL.
These methods belong in the same package as the classes. But clearly the
whole of class Object does not belong in your package! So you need a way of
putting certain methods in a package, even though the rest of the class is in
another package.
The way that you do this is by placing those methods in a protocol (of Object,
String, Date, and so on) named *PharoLink (note the initial asterisk). The
*PharoLink protocols added to the package named PharoLink. To be precise,
the rules for what goes in a package are as follows.
A package named Foo contains: 1. All class definitions of classes in the package
Foo.
2. All methods in any class in protocols named *Foo or *foo (When performing
this comparison, the case of the letters in the names is ignored.), or whose
name starts with *Foo- or *foo-.
3. All methods in classes in the package Foo except for those methods in protocols whose names start with *: Because classes of package Foo can be also
extended by other packages.
A consequence of these rules is that each class definition and each method
belongs to exactly one package.
Accessing packages
The class RPackage represents packages. The class RPackageOrganizer implements a singleton that holds all the Pharo packages. The following expressions
are examples of the queries that you can perform.
(RPackageOrganizer default packageNamed: 'AST-Core') definedClassNames
(RPackageOrganizer default packageNamed: 'AST-Core') extendedClasses
Basic Monticello
When you open the Monticello browser, you will see two list panes and a row
of buttons, as shown in Figure 8.1. The left-hand pane lists all of the packages
that have been loaded into the image that you are running; the particular
version of the package is shown in parentheses after the name.
The right-hand pane lists all of the source code repositories that Monticello
knows about, usually because it has loaded code from them. If you select a
142
8.1
Packages: groups of classes and methods
Figure 8.1: The Monticello browser.
package in the left pane, the right pane is filtered to show only those repositories that contain versions of the selected package.
The package-cache
One of the repositories is a directory named package-cache, which is a subdirectory of the directory in which your image is running. When you load code
from or write code to a remote repository, a copy is also saved in the package
cache. This can be useful if the network is not available and you need to access
a package. Also, if you are given a Monticello (.mcz) file directly, for example
as an email attachment, the most convenient way to access it is to place it in
the package-cache directory.
Adding repositories
To add a new repository to the list, click the +Repository, and choose the
kind of repository from the pop-up menu. Let’s add an HTTP repository.
To do Open Monticello, click on +Repository, and select HTTP. Edit the
dialog to read:
MCHttpRepository
location: 'http://squeaksource.com/PharoByExample'
user: ''
password: ''
Click on Open to open a repository browser on this repository. You should
see something like Figure 8.2. On the left is a list of all of the packages in the
repository; if you select one, then the pane on the right will show all of the
versions of the selected package in this repository.
143
Sharing code and source control
Figure 8.2: A Repository browser.
Browsing versions
If you select one of the versions, you can Browse it (without loading it into
your image), Load it, or look at the Changes that will be made to your image
by loading the selected version. You can also make a Copy of a version of a
package, which you can then write to another repository.
As you can see, the names of versions contain the name of the package, the
name of the author of the version, and a version number. The version name is
also the name of the file in the repository. Never change these names; correct
operation of Monticello depends on them! Monticello version files are just zip
archives, and if you are curious you can unpack them with a zip tool, but the
best way to look at their contents is using Monticello itself.
Creating a package
To create a package with Monticello, you have to do two things: write some
code, and tell Monticello about it.
To do Create a package called PBE, and put a couple of classes in it, as
shown in Figure 8.3. Also, create a method in an existing class, such as
Object, and put it in the same package as your classes – see Figure 8.4.
Committing a package
When you add a package via the class browser, Monticello automatically add
the package to its list. PBE is now in list of packages; the package entry will
be marked with an asterisk to show that the version in the image has not yet
been written to any repository (It is said to be dirty).
144
Figure 8.3: Two classes and one extension in the PBE package.
Figure 8.4: An extension method that is also be in the PBE package.
Sharing code and source control
Figure 8.5: The as-yet-unsaved PBE package in Monticello.
Initially, the only repository associated with this package will be your package
cache, as shown in Figure 8.5. That’s OK: you can still save the code, which
will cause it to be written to the package cache. Just click Save and you will be
invited to provide a log message for the version of the package that you are
about to save, as shown in Figure 8.6; when you accept the message, Monticello will save your package. To indicate this, the asterisk decorating the name
in Monticello’s package pane will be removed, and the version number added.
Introducing a change
If you then make a change to the package – say by adding a method to one
of the classes – the asterisk will re-appear, showing that you have unsaved
changes. If you open a repository browser on the package cache, you can
select the saved version, and use Changes and the other buttons. You can
of course save the new version to the repository too; once you Refresh the
repository view, it should look like Figure 8.7.
To save the new package to a repository other than the package cache, you
need to first make sure that Monticello knows about the repository, adding
it if necessary. Then you can use the Copy in the package-cache repository
browser, and select the repository to which the package should be copied. You
can also associate the desired repository with the package by right-clicking on
the repository and selecting add to package... and select the corresponding
package. Once the package knows about a repository, you can save a new
version by selecting the repository and the package in the Monticello Browser,
and clicking Save. Of course, you must have permission to write to a repository. The PharoByExample repository on SqueakSource is world readable but
not world writable, so if you try and save there, you will see an error message.
However, you can create your own repository and save your code there.
Nowadays the favorite code repositories are http://www.smalltalkhub.org
and http://ss3.gemstone.com. This is especially useful as a mechanism to share
146
Figure 8.6: Providing a log message for a new version of a package.
Figure 8.7: Two versions of our package are now in the package cache.
Sharing code and source control
Figure 8.8: The versions browser showing two versions of the ObjTest>>testIVOffset method.
your code with others and to make sure that you can use automated build
systems such as Jenkins or Travis.
If you do try and save to a repository where you don’t have write permission, a
version will nevertheless be written to the package-cache. So you can recover
by editing the repository information (right-click in the Monticello Browser)
or choosing a different repository, and then using Copy from the packagecache browser.
8.2 Source control
Versions of a method
When you save a new version of a method, the old one is not lost. Pharo keeps
all of the old versions (assuming that you are using the same image and that
you saved it), and allows you to compare different versions and to go back
(revert) to an old version.
The browse > versions (v) menu item gives access to the successive modifications made to the selected method. In Figure 8.8 we can see two versions of
a method.
148
8.2 Source control
The top pane displays one line for each version of the method, listing the
name of the programmer who wrote it, the date and time at which it was
saved, the names of the class and the method, and the protocol in which it was
defined. The current (active) version is at the top of the list.
The existence of the versions browser means that you never have to worry
about preserving code that you think might no longer be needed. If you
find that you do need it, you can always revert to the old version, or copy
the needed code fragment out of the old version and paste it into a another
method. Get into the habit of using versions; commenting out code that is no
longer needed is a bad practice because it makes the current code harder to
read. Pharoers rate code readability extremely highly.
Hint: What if you delete a method entirely, and then decide that you want
it back? You can find the deletion in a change set, where you can ask to see
versions by right-clicking. The change set browser is described in section 8.2.
Change sets and the changesorter
Whenever you are working in Pharo, any changes that you make to methods
and classes are recorded in a change set. This includes creating new classes, renaming classes, changing categories, adding methods to existing classes – just
about everything of significance. However, arbitrary Do its are not included.
At any time, many change sets exist, but only one of them – ChangeSet current – is collecting the changes that are being made to the image. You can see
which change set is current and can examine all of the change sets using the
change sorter, available by selecting World > Tools... > Change Sorter.
Figure 8.9 shows the dual change sorter browser. The title bar shows which
change set is current, and this change set is selected when the change sorter
opens.
Other change sets can be selected in the top-left pane; the right-click menu
allows you to make a different change set current, or to create a new change
set. The next pane lists all of the classes affected by the selected change set
(with their categories). Selecting one of the classes displays the names of its
methods that are also in the change set (not all of the methods in the class)
in the left central pane, and selecting a method name displays the method
definition in the bottom pane.
The change sorter also lets you delete classes and methods from the change
set using the right-click menu on the corresponding items.
The change sorter allows you to simultaneously view two change sets, one on
the left hand side and the other on the right. This layout supports the change
sorter’s main feature, which is the ability to move or copy changes from one
change set to another, as shown by the right-click menu in Figure 8.9. It is also
possible to copy individual methods from one side to the other.
149
Sharing code and source control
Figure 8.9: The Change Sorter showing all the changes of one changeset and
offering actions to move changes to other changesets.
You may be wondering why you should care about the composition of a
change set. the answer is that change sets provide a simple mechanism for
exporting code from Pharo to the file system, from where it can be imported
into another Pharo image. Change set export is known as Filing-out, and
can be accomplished using the right-click menu on any change set, class or
method in either browser. Repeated file outs create new versions of the file,
but change sets are not a versioning tool like Monticello: they do not keep
track of dependencies.
Before the advent of Monticello, change sets were the main means for exchanging code between Pharoers. They have the advantage of simplicity (the
file out is just a text file, although we don’t recommend that you try to edit
them with a text editor), and a degree of portability.
The main drawback of change sets, compared to Monticello packages, is that
they do not support the notion of dependencies. In spite of these shortcomings, change sets still have their uses. In particular, you may find change sets
on the Internet that you want to look at and perhaps use. So, having filed out
a change set using the change sorter, we will now tell you how to file one in.
This requires the use of another tool, the file list browser. Note that dropping
a changeset on Pharo itself works brings a pop up to ask you whether you
want to load or browse the code contained in a changeset.
150
8.3 The File List Browser
Figure 8.10: A file list browser.
8.3 The File List Browser
The file list browser is in fact a general-purpose tool for browsing the file
system (as well as FTP servers) from Pharo. You can open it from the World
> Tools... > File Browser menu. What you see of course depends on the
contents of your local file system, but a typical view is shown in Figure 8.10.
When you first open a file list browser it will be focused on the current directory, that is, the one from which you started Pharo. The title bar shows the
directory name. The larger pane on the left-hand side can be used to navigate
the file system in the conventional way. When a directory is selected, the files
that it contains (but not the directories) are displayed on the right. This list
of files can be filtered by entering a Unix-style pattern in the small box at the
top-left of the window. Initially, this pattern is *, which matches all file names,
but you can type a different string there and accept it, changing the pattern.
(Note that a * is implicitly prepended and appended to the pattern that you
type.) The sort order of the files can be changed using the name, date and size
buttons. The rest of the buttons depend on the name of the file selected in
the browser. In Figure 8.10, the file name has the suffix .st, so the browser
assumes that it is pharo code, and provides the possible actions as buttons.
Because the choice of buttons to display depends on the file’s name, and not
on its contents, sometimes the button that you want won’t be on the screen.
151
Sharing code and source control
Figure 8.11: A File Contents Browser opened on a package .
However, the full set of options is always available from the right-click more...
menu, so you can easily work around this problem.
The code button is perhaps the most useful for working with change sets; it
opens a browser on the contents of the change set file or even an MC package;
an example is shown in Figure 8.11. The file contents browser is similar to the
browser except that the code is not loaded in memory.
8.4
In Pharo, you can’t lose code
It is quite possible to crash Pharo: as an experimental system, Pharo lets you
change anything, including things that are vital to make Pharo work!
The good news is that you will never lose any work, even if you crash and
go back to the last saved version of your image, which might be hours old.
This is because all of the code that you executed is saved in the .changes file.
This excludes Playground snippets , those are stored inside /pharo-local/playchaceand they can be searched using Spotter (the results are displayed under
Playground-cached pages section).
So here are the instructions on how to get your code back. There is no need to
read this until you need it. However, when you do need it, you’ll find it here
waiting for you.
In the worst case, you can use a text editor on the .changes file, but since it
is many megabytes in size, this can be slow and is not recommended. Pharo
152
8.5 Chapter summary
offers you better ways.
How to get your code back
Restart Pharo from the most recent snapshot, and select World > Tools... >
Recover lost changes.
Smalltalk recover: 10000.
ChangeList browseRecentLog.
ChangeList browseRecent: 2000.
This will give you the opportunity to decide how far back in history you
wish to browse. Normally, it’s sufficient to browse changes as far back as
the last snapshot. (You can get much the same effect by editing ChangeList
browseRecent: 2000 so that the number 2000 becomes something else, using
trial and error.)
Once you have a recent changes browser, showing, say, changes back as far as
your last snapshot, you will have a list of everything that you have done to
Pharo during that time. You can delete items from this list using the rightclick menu. When you are satisfied, you can file-in what is left, thus incorporating the changes into your new image. It’s a good idea to start a new change
set, using the ordinary change set browser, before you do the file in, so that all
of your recovered code will be in a new change set. You can then file out this
change set.
One useful thing to do in the recent changes browser is to remove doIts. Usually, you won’t want to file in (and thus re-execute) doIts. However, there is an
exception. Creating a class shows up as a doIt.
Before you can file in the methods for a class, the class must exist. So, if you have
created any new classes, first file-in the class creation doIts, then remove
doIts and file in the methods.
When I am finished with the recovery, I like to file out my new change set, quit
Pharo without saving the image, restart, and make sure that the new change
set files back in cleanly.
8.5 Chapter summary
• Monticello is a tool for exporting, importing, versioning and sharing
packages of classes and methods.
• A Monticello package consists of classes, and related methods in other
packages.
• Change sets are automatically generated logs of all changes to the source
code of your image. They have largely been superseded by Monticello
as a means to store and exchange versions of your source code, but are
153
Sharing code and source control
still useful, especially for recovering from catastrophic failures, however
rare these may be.
• The file list browser is a tool for browsing the file system. It also allows
you to File in source code from the file system.
• In case your image crashes before you could save it or backup your
source code with Monticello, you can always recover your most recent
changes using a change list browser. You can then select the changes you
want to replay and file them into the most recent copy of your image.
154
CHAPTER
9
SUnit
SUnit is a minimal yet powerful framework that supports the creation and
deployment of tests. As might be guessed from its name, the design of SUnit
focussed on Unit Tests, but in fact it can be used for integration tests and
functional tests as well. SUnit was originally developed by Kent Beck and
subsequently extended by Joseph Pelrine and others to incorporate the notion
of a resource (discussed below).
In this chapter we start by discussing why we test, and what makes a good
test. We then present a series of small examples showing how to use SUnit.
Finally, we look at the implementation of SUnit, so that you can understand
how Pharo uses the power of reflection in supporting its tools. Note that the
version documented in this chapter and used in Pharo is a modified version of
SUnit3.3.
9.1
Introduction
The interest in testing and Test Driven Development is not limited to Pharo.
Automated testing has become a hallmark of the Agile software development
movement, and any software developer concerned with improving software
quality would do well to adopt it. Indeed, developers in many languages have
come to appreciate the power of unit testing, and versions of xUnit now exist
for every programming language.
Neither testing, nor the building of test suites, is new. By now, everybody
knows that tests are a good way to catch errors. eXtreme Programming, by
making testing a core practice and by emphasizing automated tests, has helped
to make testing productive and fun, rather than a chore that programmers
dislike. The Smalltalk community has a long tradition of testing because of the
155
SUnit
incremental style of development supported by its programming environment.
In traditional Smalltalk development, the programmer would write tests in
a playground as soon as a method was finished. Sometimes a test would be
incorporated as a comment at the head of the method that it exercised, or
tests that needed some set up would be included as example methods in the
class. The problem with these practices is that tests in a playground are not
available to other programmers who modify the code. Comments and example
methods are better in this respect, but there is still no easy way to keep track
of them and to run them automatically. Tests that are not run do not help you
to find bugs! Moreover, an example method does not inform the reader of the
expected result: you can run the example and see the (perhaps surprising)
result, but you will not know if the observed behaviour is correct.
SUnit is valuable because it allows us to write tests that are self-checking:
the test itself defines what the correct result should be. It also helps us to
organize tests into groups, to describe the context in which the tests must
run, and to run a group of tests automatically. In less than two minutes you
can write tests using SUnit, so instead of writing small code snippets in a
playgound, we encourage you to use SUnit and get all the advantages of stored
and automatically executable tests.
9.2 Why testing is important
Unfortunately, many developers believe that tests are a waste of their time.
After all, they do not write bugs, only other programmers do that. Most of
us have said, at some time or other: I would write tests if I had more time. If
you never write a bug, and if your code will never be changed in the future,
then indeed tests are a waste of your time. However, this most likely also
means that your application is trivial, or that it is not used by you or anyone
else. Think of tests as an investment for the future: having a suite of tests is
quite useful now, but it will be extremely useful when your application, or the
environment in which it runs, changes in the future.
Tests play several roles. First, they provide documentation of the functionality
that they cover. This documentation is active: watching the tests pass tells
you that the documentation is up to date. Second, tests help developers to
confirm that some changes that they have just made to a package have not
broken anything else in the system, and to find the parts that break when that
confidence turns out to be misplaced. Finally, writing tests during, or even
before, programming forces you to think about the functionality that you
want to design, and how it should appear to the client code, rather than about how
to implement it.
By writing the tests first, i.e., before the code, you are compelled to state the
context in which your functionality will run, the way it will interact with the
client code, and the expected results. Your code will improve. Try it.
156
9.3 What makes a good test?
We cannot test all aspects of any realistic application. Covering a complete
application is simply impossible and should not be the goal of testing. Even
with a good test suite some bugs will still creep into the application, where
they can lay dormant waiting for an opportunity to damage your system. If
you find that this has happened, take advantage of it! As soon as you uncover
the bug, write a test that exposes it, run the test, and watch it fail. Now you
can start to fix the bug: the test will tell you when you are done.
9.3 What makes a good test?
Writing good tests is a skill that can be learned by practicing. Let us look at
the properties that tests should have to get the maximum benefit.
Tests should be repeatable. You should be able to run a test as often as you want,
and always get the same answer.
Tests should run without human intervention. You should be able to run them
unattended.
Tests should tell a story. Each test should cover one aspect of a piece of code. A
test should act as a scenario that you or someone else can read to understand
a piece of functionality.
Tests should have a change frequency lower than that of the functionality they cover.
You do not want to have to change all your tests every time you modify your
application. One way to achieve this is to write tests based on the public
interfaces of the class that you are testing. It is OK to write a test for a private
helper method if you feel that the method is complicated enough to need the
test, but you should be aware that such a test may have to be changed, or
thrown away entirely, when you think of a better implementation.
One consequence of such properties is that the number of tests should be
somewhat proportional to the number of functions to be tested: changing one
aspect of the system should not break all the tests but only a limited number.
This is important because having 100 tests fail should send a much stronger
message than having 10 tests fail. However, it is not always possible to achieve
this ideal: in particular, if a change breaks the initialization of an object, or
the set-up of a test, it is likely to cause all of the tests to fail.
Several software development methodologies such as eXtreme Programming and
Test-Driven Development (TDD) advocate writing tests before writing code.
This may seem to go against our deep instincts as software developers. All we
can say is: go ahead and try it. We have found that writing the tests before
the code helps us to know what we want to code, helps us know when we are
done, and helps us conceptualize the functionality of a class and to design its
interface. Moreover, test-first development gives us the courage to go fast,
because we are not afraid that we will forget something important.
157
SUnit
Writing tests is not difficult in itself. Choosing what to test is much more
difficult. The pragmatic programmers offer the ”right-BICEP” principle. It
stands for:
• Right: Are the results right?
• B: Are all the boundary conditions correct?
• I: Can you check inverse relationships?
• C: Can you cross-check results using other means?
• E: Can you force error conditions to happen?
• P: Are performance characteristics within bounds?
Now let’s write our first test, and show you the benefits of using SUnit.
9.4
SUnit by example
Before going into the details of SUnit, we will show a step by step example. We
use an example that tests the class Set. Try entering the code as we go along.
Step 1: Create the test class
First you should create a new subclass of TestCase called MyExampleSetTest.
Add two instance variables so that your new class looks like this:
TestCase subclass: #MyExampleSetTest
instanceVariableNames: 'full empty'
classVariableNames: ''
package: 'MySetTest'
We will use the class MyExampleSetTest to group all the tests related to the
class Set. It defines the context in which the tests will run. Here the context
is described by the two instance variables full and empty that we will use to
represent a full and an empty set.
The name of the class is not critical, but by convention it should end in Test.
If you define a class called Pattern and call the corresponding test class
PatternTest, the two classes will be alphabetized together in the browser
(assuming that they are in the same package). It is critical that your class is a
subclass of TestCase.
Step 2: Initialize the test context
The message TestCase >> setUp defines the context in which the tests will
run, a bit like an initialize method. setUp is invoked before the execution of
each test method defined in the test class.
158
9.4
SUnit by example
Define the setUp method as follows, to initialize the empty variable to refer to
an empty set and the full variable to refer to a set containing two elements.
MyExampleSetTest >> setUp
empty := Set new.
full := Set with: 5 with: 6
In testing jargon the context is called the fixture for the test.
Step 3: write some test methods
Let’s create some tests by defining some methods in the class MyExampleSetTest. Each method represents one test. The names of the methods should
start with the string 'test' so that SUnit will collect them into test suites.
Test methods take no arguments.
Define the following test methods. The first test, named testIncludes, tests
the includes: method of Set. The test says that sending the message includes: 5 to a set containing 5 should return true. Clearly, this test relies on
the fact that the setUp method has already run.
MyExampleSetTest >> testIncludes
self assert: (full includes: 5).
self assert: (full includes: 6)
The second test, named testOccurrences, verifies that the number of occurrences of 5 in full set is equal to one, even if we add another element 5 to the
set.
MyExampleSetTest >> testOccurrences
self assert: (empty occurrencesOf: 0) = 0.
self assert: (full occurrencesOf: 5) = 1.
full add: 5.
self assert: (full occurrencesOf: 5) = 1
Finally, we test that the set no longer contains the element 5 after we have
removed it.
MyExampleSetTest >> testRemove
full remove: 5.
self assert: (full includes: 6).
self deny: (full includes: 5)
Note the use of the method TestCase >> deny: to assert something that
should not be true. aTest deny: anExpression is equivalent to aTest assert: anExpression not, but is much more readable.
Step 4: Run the tests
The easiest way to run the tests is directly from the browser. Simply click on
the icon of the class name, or on an individual test method, and select Run tests
159
SUnit
Figure 9.1: Running SUnit tests from the System Browser.
(t) or press the icon. The test methods will be flagged green or red, depending
on whether they pass or not (as shown in 9.1).
You can also select sets of test suites to run, and obtain a more detailed log
of the results using the SUnit Test Runner, which you can open by selecting
World > Test Runner.
The Test Runner, shown in Figure 9.2, is designed to make it easy to execute
groups of tests.
The left-most pane lists all of the packages that contain test classes (i.e., subclasses of TestCase). When some of these packages are selected, the test
classes that they contain appear in the pane to the right. Abstract classes
are italicized, and the test class hierarchy is shown by indentation, so subclasses of ClassTestCase are indented more than subclasses of TestCase.
ClassTestCase is a class offering utilities methods to compute test coverage.
Open a Test Runner, select the package MySetTest, and click the Run Selected
button.
You can also run a single test (and print the usual pass/fail result summary) by executing a Print it on the following code: MyExampleSetTest run:
#testRemove.
Some people include an executable comment in their test methods that allows
running a test method with a Do it from the browser, as shown below.
MyExampleSetTest >> testRemove
"self run: #testRemove"
full remove: 5.
self assert: (full includes: 6).
self deny: (full includes: 5)
160
9.4
SUnit by example
Figure 9.2: Running SUnit tests using the TestRunner.
Introduce a bug in MyExampleSetTest >> testRemove and run the tests again.
For example, change 6 to 7, as in:
MyExampleSetTest >> testRemove
full remove: 5.
self assert: (full includes: 7).
self deny: (full includes: 5)
The tests that did not pass (if any) are listed in the right-hand panes of the Test
Runner. If you want to debug one, to see why it failed, just click on the name.
Alternatively, you can execute one of the following expressions:
(MyExampleSetTest selector: #testRemove) debug
MyExampleSetTest debug: #testRemove
Step 5: Interpret the results
The method assert: is defined in the class TestAsserter. This is a superclass
of TestCase and therefore all other TestCase subclasses and is responsible
for all kind of test result assertions. The assert: method expects a boolean
argument, usually the value of a tested expression. When the argument is true,
the test passes; when the argument is false, the test fails.
There are actually three possible outcomes of a test: passing, failing, and raising
an error.
• Passing. The outcome that we hope for is that all of the assertions in the
test are true, in which case the test passes. In the test runner, when all
of the tests pass, the bar at the top turns green. However, there are two
other ways that running a test can go wrong.
161
SUnit
• Failing. The obvious way is that one of the assertions can be false,
causing the test to fail.
• Error. The other possibility is that some kind of error occurs during
the execution of the test, such as a message not understood error or an
index out of bounds error. If an error occurs, the assertions in the test
method may not have been executed at all, so we can’t say that the test
has failed; nevertheless, something is clearly wrong!
In the test runner, failing tests cause the bar at the top to turn yellow, and are
listed in the middle pane on the right, whereas tests with errors cause the bar
to turn red, and are listed in the bottom pane on the right.
Modify your tests to provoke both errors and failures.
9.5 The SUnit cookbook
This section will give you more details on how to use SUnit. If you have used
another testing framework such as JUnit, much of this will be familiar, since
all these frameworks have their roots in SUnit. Normally you will use SUnit’s
GUI to run tests, but there are situations where you may not want to use it.
Other assertions
In addition to assert: and deny:, there are several other methods that can be
used to make assertions.
First, TestAsserter >> assert:description: and TestAsserter >> deny:description: take a second argument which is a message string that describes
the reason for the failure, if it is not obvious from the test itself. These methods are described in Section 9.7.
Next, SUnit provides two additional methods, TestAsserter >> should:raise:
and TestAsserter >> shouldnt:raise: for testing exception propagation.
For example, you would use self should: aBlock raise: anException to
test that a particular exception is raised during the execution of aBlock. The
method below illustrates the use of should:raise:.
MyExampleSetTest >> testIllegal
self should: [ empty at: 5 ] raise: Error.
self should: [ empty at: 5 put: #zork ] raise: Error
Try running this test. Note that the first argument of the should: and shouldnt:
methods is a block that contains the expression to be executed.
Running a single test
Normally, you will run your tests using the Test Runner or using your code
browser. If you don’t want to launch the Test Runner from the World menu,
162
9.6 The SUnit framework
Figure 9.3: The four classes representing the core of SUnit.
you can execute TestRunner open. You can also run a single test as follows:
MyExampleSetTest run: #testRemove
>>> 1 run, 1 passed, 0 failed, 0 errors
Running all the tests in a test class
Any subclass of TestCase responds to the message suite, which will build a
test suite that contains all the methods in the class whose names start with the
string test.
To run the tests in the suite, send it the message run. For example:
MyExampleSetTest suite run
>>> 5 run, 5 passed, 0 failed, 0 errors
Must I subclass TestCase?
In JUnit you can build a TestSuite from an arbitrary class containing test*
methods. In SUnit you can do the same but you will then have to create a
suite by hand and your class will have to implement all the essential TestCase
methods like assert:. We recommend, however, that you not try to do this.
The framework is there: use it.
9.6 The SUnit framework
SUnit consists of four main classes: TestCase, TestSuite, TestResult, and
TestResource, as shown in Figure 9.3. The notion of a test resource represents
a resource that is expensive to set-up but which can be used by a whole series
of tests. A TestResource specifies a setUp method that is executed just once
before a suite of tests; this is in distinction to the TestCase >> setUp method,
which is executed before each test.
163
SUnit
TestCase
TestCase is an abstract class that is designed to be subclassed. Each of its
subclasses represents a group of tests that share a common context (that
is, a test suite). Each test is run by creating a new instance of a subclass of
TestCase, running setUp, running the test method itself, and then sending
the tearDown.
The context is specified by instance variables of the subclass and by the specialization of the method setUp, which initializes those instance variables.
Subclasses of TestCase can also override method tearDown, which is invoked
after the execution of each test, and can be used to release any objects allocated during setUp.
TestSuite
Instances of the class TestSuite contain a collection of test cases. An instance
of TestSuite contains tests, and other test suites. That is, a test suite contains
sub-instances of TestCase and TestSuite.
Both individual test cases and test suites understand the same protocol, so
they can be treated in the same way (for example, both can be run). This is
in fact an application of the Composite pattern in which TestSuite is the
composite and the test cases are the leaves.
TestResult
The class TestResult represents the results of a TestSuite execution. It
records the number of tests passed, the number of tests failed, and the number
of errors signalled.
TestResource
One of the important features of a suite of tests is that they should be independent of each other. The failure of one test should not cause an avalanche
of failures of other tests that depend upon it, nor should the order in which
the tests are run matter. Performing setUp before each test and tearDown
afterwards helps to reinforce this independence.
However, there are occasions where setting up the necessary context is just
too time-consuming for it to be done before the execution of each test. Moreover, if it is known that the test cases do not disrupt the resources used by the
tests, then it is wasteful to set them up afresh for each test. It is sufficient to
set them up once for each suite of tests. Suppose, for example, that a suite of
tests needs to query a database, or do analysis on some compiled code. In such
cases, it may make sense to set up the database and open a connection to it, or
to compile some source code, before any of the tests start to run.
164
9.6 The SUnit framework
Where should we cache these resources, so that they can be shared by a suite
of tests? The instance variables of a particular TestCase subclass won’t do,
because a TestCase instance persists only for the duration of a single test (as
mentioned before, the instance is created anew for each test method). A global
variable would work, but using too many global variables pollutes the name
space, and the binding between the global and the tests that depend on it
will not be explicit. A better solution is to put the necessary resources in a
singleton object of some class. The class TestResource exists to be subclassed
by such resource classes. Each subclass of TestResource understands the
message current, which will answer a singleton instance of that subclass.
Methods setUp and tearDown should be overridden in the subclass to ensure
that the resource is initialized and finalized.
One thing remains: somehow, SUnit has to be told which resources are associated with which test suite. A resource is associated with a particular subclass
of TestCase by overriding the class method resources.
By default, the resources of a TestSuite are the union of the resources of the
TestCases that it contains.
Here is an example. We define a subclass of TestResource called MyTestResource. Then we associate it with MyTestCase by overriding the class method
MyTestCase class >> resources to return an array of the test resource
classes that MyTestCase will use.
TestResource subclass: #MyTestResource
instanceVariableNames: ''
...
MyTestCase class >> resources
"Associate the resource with this class of test cases"
^ { MyTestResource }
Exercise
The following trace (written to the Transcript) illustrates that a global set up
is run before and after each test in a sequence. Let’s see if you can obtain this
trace yourself.
MyTestResource >> setUp has run.
MyTestCase >> setUp has run.
MyTestCase >> testOne has run.
MyTestCase >> tearDown has run.
MyTestCase >> setUp has run.
MyTestCase >> testTwo has run.
MyTestCase >> tearDown has run.
MyTestResource >> tearDown has run.
165
SUnit
Create new classes MyTestResource and MyTestCase which are subclasses of
TestResource and TestCase respectively. Add the appropriate methods so
that the following messages are written to the Transcript when you run your
tests.
Solution.
You will need to write the following six methods.
MyTestCase >> setUp
Transcript show: 'MyTestCase>>setUp has run.'; cr
MyTestCase >> tearDown
Transcript show: 'MyTestCase>>tearDown has run.'; cr
MyTestCase >> testOne
Transcript show: 'MyTestCase>>testOne has run.'; cr
MyTestCase >> testTwo
Transcript show: 'MyTestCase>>testTwo has run.'; cr
MyTestCase class >> resources
^ Array with: MyTestResource
MyTestResource >> setUp
Transcript show: 'MyTestResource>>setUp has run'; cr
MyTestResource >> tearDown
Transcript show: 'MyTestResource>>tearDown has run.'; cr
9.7 Advanced features of SUnit
In addition to TestResource, SUnit contains assertion description strings,
logging support, the ability to skip tests, and resumable test failures.
Assertion description strings
The TestAsserter assertion protocol includes a number of methods that allow
the programmer to supply a description of the assertion. The description is a
String; if the test case fails, this string will be displayed by the test runner. Of
course, this string can be constructed dynamically.
...
e := 42.
self assert: e = 23 description: 'expected 23, got ', e printString
...
The relevant methods in TestAsserter are:
assert:description:
deny:description:
166
9.7 Advanced features of SUnit
should:description:
shouldnt:description:
Using assert:equals:
In addition to assert:, there is also assert:equals: that offers a better
report in case of error (incidentally, assert:equals: uses assert:description:).
For example, the two following tests are equivalent. However, the second
one will report the value that the test is expecting: this makes easier to understand the failure. In this example, we suppose that aDateAndTime is an
instance variable of the test class.
testAsDate
self assert: aDateAndTime asDate = ('February 29, 2004' asDate
translateTo: 2 hours).
testAsDate
self
assert: aDateAndTime asDate
equals: ('February 29, 2004' asDate translateTo: 2 hours).
Logging support
The description strings mentioned above may also be logged to a Stream,
such as the Transcript or a file stream. You can choose whether to log by
overriding isLogging in your test class; you must also choose where to log
by overriding failureLog to answer an appropriate stream. By default, the
Transcript is used to log.
Skipping tests
Sometimes in the middle of a development, you may want to skip a test instead of removing it or renaming it to prevent it from running. You can
simply invoke the TestAsserter message skip on your test case instance. For
example, the following test uses it to define a conditional test.
OCCompiledMethodIntegrityTest >> testPragmas
| newCompiledMethod originalCompiledMethod |
(Smalltalk globals hasClassNamed: #Compiler) ifFalse: [ ^ self skip
].
...
167
SUnit
9.8 Continuing after a failure
SUnit also allows us to specify whether or not a test should continue after a
failure. This is a really powerful feature that uses Pharo’s exception mechanisms. To see what this can be used for, let’s look at an example. Consider the
following test expression:
aCollection do: [ :each | self assert: each even ]
In this case, as soon as the test finds the first element of the collection that
isn’t even, the test stops. However, we would usually like to continue, and see
both how many elements, and which elements, aren’t even (and maybe also
log this information). You can do this as follows:
aCollection do: [ :each |
self
assert: each even
description: each printString, ' is not even'
resumable: true ]
This will print out a message on your logging stream for each element that
fails. It doesn’t accumulate failures, i.e, if the assertion fails 10 times in your
test method, you’ll still only see one failure. All the other assertion methods
that we have seen are not resumable by default; assert: p description: s
is equivalent to assert: p description: s resumable: false.
9.9 SUnit implementation
The implementation of SUnit makes an interesting case study of a Pharo
framework. Let’s look at some key aspects of the implementation by following
the execution of a test.
Running one test
To execute one test, we execute the expression (aTestClass selector:
aSymbol) run.
The method TestCase >> run creates an instance of TestResult that will
accumulate the results of the test, then it sends itself the message TestCase
>> run: (See Figure 9.4).
TestCase >> run
| result |
result := self classForTestResult new.
[ self run: result ]
ensure: [ self classForTestResource resetResources: self resources
].
^ result
168
9.9 SUnit implementation
Figure 9.4: Running one test.
The method TestCase >> run: sends the message runCase: to the test result:
TestCase >> run: aResult
aResult runCase: self
The method TestResult >> runCase: sends the message TestCase >> runCase to an individual test, to execute the test. TestResult >> runCase deals
with any exceptions that may be raised during the execution of a test, runs
a TestCase by sending it the runCase, and counts the errors, failures, and
passes.
TestResult >> runCase: aTestCase
[
aTestCase announce: TestCaseStarted withResult: self.
aTestCase runCase.
aTestCase announce: TestCaseEnded withResult: self.
self addPass: aTestCase ]
on: self class failure, self class skip, self class warning, self
class error
do: [ :ex | ex sunitAnnounce: aTestCase toResult: self ]
The method TestCase >> runCase sends the messages TestCase >> setUp
and TestCase >> tearDown as shown below.
TestCase >> runCase
self resources do: [ :each | each availableFor: self ].
[ self setUp.
self performTest ] ensure: [
self tearDown.
self cleanUpInstanceVariables ]
169
SUnit
Running a TestSuite
To run more than one test, we send the message run to a TestSuite that contains the relevant tests. TestCase class provides some functionality to build
a test suite from its methods. The expression MyTestCase buildSuiteFromSelectors returns a suite containing all the tests defined in the MyTestCase
class. The core of this process is:
TestCase class >> testSelectors
^ (self selectors select: [ :each | (each beginsWith: 'test') and: [
each numArgs isZero ]])
The method TestSuite >> run creates an instance of TestResult, verifies
that all the resources are available, and then sends itself the message TestSuite >> run:, which runs all the tests in the suite. All the resources are
then released.
TestSuite >> run: aResult
self setUp.
[ self tests
do: [ :each |
each run: aResult.
self announceTest: each.
self changed: each ] ]
ensure: [ self tearDown ]
TestSuite >> setUp
self resources do: [ :each |
each isAvailable ifFalse: [ each signalInitializationError ]
].
TestSuite >> tearDown
self resourceClass resetResources: self resources
The class TestResource and its subclasses keep track of their currently created singleton instances that can be accessed and created using the class
method TestResource class >> current. This instance is cleared when the
tests have finished running and the resources are reset.
The resource availability check makes it possible for the resource to be recreated if needed, as shown in the class method TestResource class >>
isAvailable. During the TestResource instance creation, it is initialized and
the message setUp is sent to a test resource.
TestResource class >> isAvailable
"This is (and must be) a lazy method. If my current has a value, an
attempt to make me available has already been made: trust its
result. If not, try to make me available."
current ifNil: [ self makeAvailable ].
^ self isAlreadyAvailable
170
9.10
A piece of advices on testing
TestResource class >> current
"This is a lazy accessor: the assert of self isAvailable does no
work unless current isNil. However this method should normally
be sent only to a resource that should already have been made
available, e.g. in a test whose test case class has the resource
class in its #resources, so should never be able to fail the
assert.
If the intent is indeed to access a possibly-unprepared or
reset-in-earlier-test resource lazily, then preface the call of
'MyResource current' with 'MyResource availableFor: self'."
self
assert: self isAvailable
description:
'Sent #current to unavailable resource ', self name,
'. Add it to test case'' class-side #resources (recommended)
or send #availableFor: beforehand'.
^ current
9.10
A piece of advices on testing
While the mechanics of testing are easy, writing good tests is not. Here is some
advice on how to design tests.
Self-contained tests
You do not want to have to change your tests each time you change your code,
so try to write the tests so that they are self-contained. This can be difficult,
but pays off in the long term. Writing tests against stable interfaces supports
this effort.
Do not over-test
Try to build your tests so that they do not overlap. It is annoying to have many
tests covering the same functionality, because one bug in the code will then
break many tests at the same time. This is covered by Black’s rule, below.
Feathers’ Rules for Unit tests
Michael Feathers, an agile process consultant and author, writes:
A test is not a unit test if: it talks to the database, it communicates across the network,
it touches the file system, it can’t run at the same time as any of your other unit
tests, or you have to do special things to your environment (such as editing config
files) to run it. Tests that do these things aren’t bad. Often they are worth writing,
and they can be written in a unit test harness. However, it is important to be able to
separate them from true unit tests so that we can keep a set of tests that we can run
171
SUnit
fast whenever we make our changes. Never get yourself into a situation where you
don’t want to run your unit test suite because it takes too long.
Unit Tests vs. Acceptance Tests
Unit tests capture one piece of functionality, and as such make it easier to
identify bugs in that functionality. As far as possible try to have unit tests
for each method that could possibly fail, and group them per class. However,
for certain deeply recursive or complex setup situations, it is easier to write
tests that represent a scenario in the larger application. These are called
acceptance tests (or integration tests, or functional tests). Tests that break
Feathers’ rules may make good acceptance tests. Group acceptance tests
according to the functionality that they test. For example, if you are writing
a compiler, you might write acceptance tests that make assertions about the
code generated for each possible source language statement. Such tests might
exercise many classes, and might take a long time to run because they touch
the file system. You can write them using SUnit, but you won’t want to run
them each time you make a small change, so they should be separated from
the true unit tests.
Black’s Rule of Testing
For every test in the system, you should be able to identify some property for
which the test increases your confidence. It’s obvious that there should be no
important property that you are not testing. This rule states the less obvious
fact that there should be no test that does not add value to the system by
increasing your confidence that a useful property holds. For example, several
tests of the same property do no good. In fact, they do harm in two ways. First,
they make it harder to infer the behaviour of the class by reading the tests.
Second, because one bug in the code might then break many tests, they make
it harder to estimate how many bugs remain in the code. So, have a property
in mind when you write a test.
9.11
Chapter summary
This chapter explained why tests are an important investment in the future of
your code. We explained in a step-by-step fashion how to define a few tests for
the class Set. Then we gave an overview of the core of the SUnit framework by
presenting the classes TestCase, TestResult, TestSuite and TestResources.
Finally we looked deep inside SUnit by following the execution of a test and a
test suite.
• To maximize their potential, unit tests should be fast, repeatable, independent of any direct human interaction and cover a single unit of
functionality.
172
9.11
Chapter summary
• Tests for a class called MyClass belong in a class named MyClassTest,
which should be introduced as a subclass of TestCase.
• Initialize your test data in a setUp method.
• Each test method should start with the word test.
• Use the TestCase methods assert:, deny: and others to make assertions.
• Run tests!
173
CHAPTER
10
Basic classes
Pharo is a really simple language but powerful language. Part of its power
is not in the language but in its class libraries. To program effectively in
it, you will need to learn how the class libraries support the language and
environment. The class libraries are entirely written in Pharo, and can easily
be extended. (Recall that a package may add new functionality to a class even
if it does not define this class.)
Our goal here is not to present in tedious detail the whole of the Pharo class
library, but rather to point out the key classes and methods that you will need
to use (or subclass/override) to program effectively. In this chapter, we will
cover the basic classes that you will need for nearly every application: Object,
Number and its subclasses, Character, String, Symbol, and Boolean.
10.1
Object
For all intents and purposes, Object is the root of the inheritance hierarchy.
Actually, in Pharo the true root of the hierarchy is ProtoObject, which is used
to define minimal entities that masquerade as objects, but we can ignore this
point for the time being.
Object defines almost 400 methods (in other words, every class that you
define will automatically provide all those methods). Note: You can count the
number of methods in a class like so:
Object selectors size "Count the instance methods in Object"
Object class selectors size "Count the class methods"
Class Object provides default behaviour common to all normal objects, such
as access, copying, comparison, error handling, message sending, and reflec175
Basic classes
tion. Also utility messages that all objects should respond to are defined here.
Object has no instance variables, nor should any be added. This is due to
several classes of objects that inherit from Object that have special implementations (SmallInteger and UndefinedObject for example) that the VM knows
about and depends on the structure and layout of certain standard classes.
If we begin to browse the method protocols on the instance side of Object we
will start to see some of the key behaviour it provides.
Printing
Every object can return a printed form of itself. You can select any expression
in a textpane and select the Print it menu item: this executes the expression
and asks the returned object to print itself. In fact this sends the message
printString to the returned object. The method printString, which is a
template method, at its core sends the message printOn: to its receiver. The
message printOn: is a hook that can be specialized.
Method Object>>printOn: is very likely one of the methods that you will
most frequently override. This method takes as its argument a Stream on
which a String representation of the object will be written. The default implementation simply writes the class name preceded by a or an. Object>>printString
returns the String that is written.
For example, the class OpalCompiler does not redefine the method printOn:
and sending the message printString to an instance executes the methods
defined in Object.
OpalCompiler new printString
>>> 'an OpalCompiler'
The class Color shows an example of printOn: specialization. It prints the
name of the class followed by the name of the class method used to generate
that color.
Color >> printOn: aStream
| name |
(name := self name).
name = #unnamed
ifFalse: [
^ aStream
nextPutAll: 'Color ';
nextPutAll: name ].
self storeOn: aStream]]]
[[[
Color red printString
>>> 'Color red'
Note that the message printOn: is not the same as storeOn:. The message
storeOn: writes to its argument stream an expression that can be used to
176
10.1
Object
recreate the receiver. This expression is executed when the stream is read
using the message readFrom:. On the other hand, the message printOn:
just returns a textual version of the receiver. Of course, it may happen that
this textual representation may represent the receiver as a self-evaluating
expression.
In functional programming, expressions return values when executed. In Pharo,
messages (expressions) return objects (values). Some objects have the nice
property that their value is themselves. For example, the value of the object
true is itself i.e., the object true. We call such objects self-evaluating objects.
You can see a printed version of an object value when you print the object in a
playground. Here are some examples of such self-evaluating expressions.
A word about representation and self-evaluating representation.
true
>>> true
3@4
>>> (3@4)
$a
>>> $a
#(1 2 3)
>>> #(1 2 3)
Color red
>>> Color red
Note that some objects such as arrays are self-evaluating or not depending on
the objects they contain. For example, an array of booleans is self-evaluating,
whereas an array of persons is not. The following example shows that a dynamic array is self-evaluating only if its elements are:
{10@10. 100@100}
>>> {(10@10). (100@100)}
{Nautilus new . 100@100}
>>> an Array(a Nautilus (100@100))
Remember that literal arrays can only contain literals. Hence the following
array does not contain two points but rather six literal elements.
#(10@10 100@100)
>>> #(10 #@ 10 100 #@ 100)
Lots of printOn: method specializations implement self-evaluating behavior.
The implementations of Point>>printOn: and Interval>>printOn: are selfevaluating.
Point >> printOn: aStream
"The receiver prints on aStream in terms of infix notation."
aStream nextPut: $(.
x printOn: aStream.
177
Basic classes
aStream nextPut: $@.
(y notNil and: [y negative])
ifTrue: [
"Avoid ambiguous @- construct"
aStream space ].
y printOn: aStream.
aStream nextPut: $).
Interval >> printOn: aStream
aStream nextPut: $(;
print: start;
nextPutAll: ' to: ';
print: stop.
step ~= 1 ifTrue: [aStream nextPutAll: ' by: '; print: step].
aStream nextPut: $)
1 to: 10
>>> (1 to: 10) "intervals are self-evaluating"
Identity and equality
In Pharo, the message = tests object equality (i.e., whether two objects represent the same value) whereas == tests object identity (whether two expressions
represent the same object).
The default implementation of object equality is to test for object identity:
Object >> = anObject
"Answer whether the receiver and the argument represent the same
object.
If = is redefined in any subclass, consider also redefining the
message hash."
^ self == anObject
This is a method that you will frequently want to override. Consider the
case of Complex numbers as defined in the SciSmalltalk/PolyMath packages
(PolyMath is a set of packages that offer support for numerical methods):
(1 + 2 i) = (1 + 2 i)
>>> true "same value"
(1 + 2 i) == (1 + 2 i)
>>> false "but different objects"
This works because Complex overrides = as follows:
Complex >> = anObject
^ anObject isComplex
ifTrue: [ (real = anObject real) & (imaginary = anObject
imaginary)]
ifFalse: [ anObject adaptToComplex: self andSend: #=]
178
10.1
Object
The default implementation of Object>>~= (a test for inequality) simply
negates Object>>=, and should not normally need to be changed.
(1 + 2 i) ~= (1 + 4 i)
>>> true
If you override =, you should consider overriding hash. If instances of your
class are ever used as keys in a Dictionary, then you should make sure that
instances that are considered to be equal have the same hash value:
Complex >> hash
"Hash is reimplemented because = is implemented."
^ real hash bitXor: imaginary hash.
Although you should override = and hash together, you should never override
==. The semantics of object identity is the same for all classes. Message == is a
primitive method of ProtoObject.
Note that Pharo has some strange equality behaviour compared to other
Smalltalks. For example a symbol and a string can be equal. (We consider this
to be a bug, not a feature.)
#'lulu' = 'lulu'
>>> true
'lulu' = #'lulu'
>>> true
Class membership
Several methods allow you to query the class of an object.
class.
You can ask any object about its class using the message class.
1 class
>>> SmallInteger
isMemberOf:.
class:
Conversely, you can ask if an object is an instance of a specific
1 isMemberOf: SmallInteger
>>> true "must be precisely this class"
1 isMemberOf: Integer
>>> false
1 isMemberOf: Number
>>> false
1 isMemberOf: Object
>>> false
Since Pharo is written in itself, you can really navigate through its structure
using the right combination of superclass and class messages (see Chapter :
Classes and Metaclasses).
179
Basic classes
isKindOf:. Object>>isKindOf: answers whether the receiver’s class is
either the same as, or a subclass of the argument class.
1 isKindOf:
>>> true
1 isKindOf:
>>> true
1 isKindOf:
>>> true
1 isKindOf:
>>> true
1 isKindOf:
>>> false
1/3
>>>
1/3
>>>
SmallInteger
Integer
Number
Object
String
isKindOf: Number
true
isKindOf: Integer
false
1/3 which is a Fraction is a kind of Number, since the class Number is a superclass of the class Fraction, but 1/3 is not an Integer.
respondsTo:. Object>>respondsTo: answers whether the receiver understands the message selector given as an argument.
1 respondsTo: #,
>>> false
A note on the usage of respondsTo:. Normally it is a bad idea to query an object for its class, or to ask it which messages it understands. Instead of making
decisions based on the class of object, you should simply send a message to the
object and let it decide (on the basis of its class) how it should behave. (This
concept is sometimes referred to as duck typing).
Copying
Copying objects introduces some subtle issues. Since instance variables are
accessed by reference, a shallow copy of an object would share its references to
instance variables with the original object:
a1 := { { 'harry' } }.
a1
>>> #(#('harry'))
a2 := a1 shallowCopy.
a2
>>> #(#('harry'))
(a1 at: 1) at: 1 put: 'sally'.
a1
>>> #(#('sally'))
a2
180
10.1
Object
>>> #(#('sally')) "the subarray is shared!"
Object>>shallowCopy is a primitive method that creates a shallow copy of an
object. Since a2 is only a shallow copy of a1, the two arrays share a reference
to the nested Array that they contain.
Object>>deepCopy makes an arbitrarily deep copy of an object.
a1 := { { { 'harry' } } } .
a2 := a1 deepCopy.
(a1 at: 1) at: 1 put: 'sally'.
a1
>>> #(#('sally'))
a2
>>> #(#(#('harry')))
The problem with deepCopy is that it will not terminate when applied to a
mutually recursive structure:
a1 := { 'harry' }.
a2 := { a1 }.
a1 at: 1 put: a2.
a1 deepCopy
>>> !''... does not terminate!''!
An alternate solution is to use message copy. It is implemented on Object as
follows:
Object >> copy
"Answer another instance just like the receiver.
Subclasses typically override postCopy;
they typically do not override shallowCopy."
^ self shallowCopy postCopy
Object >> postCopy
^ self
By default postCopy returns self. It means that by default copy is doing the
same as shallowCopy but each subclass can decide to customise the postCopy
method which acts as a hook. You should override postCopy to copy any
instance variables that should not be shared. In addition there is a good
chance that postCopy should always do a super postCopy to ensure that state
of the superclass is also copied.
Debugging
halt. The most important method here is halt. To set a breakpoint in a
method, simply insert the expression send self halt at some point in the
body of the method. (Note that since halt is defined on Object you can also
write 1 halt). When this message is sent, execution will be interrupted and
181
Basic classes
a debugger will open to this point in your program (See Chapter : The Pharo
Environment for more details about the debugger.).
You can also use Halt once or Halt if: aCondition. Have a look at the class
Halt which is an special exception.
assert:. The next most important message is assert:, which expects a
block as its argument. If the block evaluates to true, execution continues.
Otherwise an AssertionFailure exception will be raised. If this exception is
not otherwise caught, the debugger will open to this point in the execution.
assert: is especially useful to support design by contract. The most typical
usage is to check non-trivial pre-conditions to public methods of objects.
Stack>>pop could easily have been implemented as follows (Note that this
definition is anhypothetical example and not in the Pharo 50 system):
Stack >> pop
"Return the first element and remove it from the stack."
self assert: [ self isNotEmpty ].
^ self linkedList removeFirst element
Do not confuse Object>>assert: with TestCase>>assert:, which occurs in
the SUnit testing framework (see Chapter : SUnit). While the former expects
a block as its argument (actually, it will take any argument that understands
value, including a Boolean), the latter expects a Boolean. Although both are
useful for debugging, they each serve a very different purpose.
Error handling
This protocol contains several methods useful for signaling run-time errors.
deprecated:. Sending self deprecated: signals that the current method
should no longer be used, if deprecation has been turned on. You can turn it
on/off in the Debugging section using the Settings browser. The argument
should describe an alternative. Look for senders of the message deprecated:
to get an idea.
doesNotUnderstand:. doesNotUnderstand: (commonly abbreviated in
discussions as DNU or MNU) is sent whenever message lookup fails. The
default implementation, i.e., Object>>doesNotUnderstand: will trigger the
debugger at this point. It may be useful to override doesNotUnderstand: to
provide some other behaviour.
error. Object>>error and Object>>error: are generic methods that can
be used to raise exceptions. (Generally it is better to raise your own custom
exceptions, so you can distinguish errors arising from your code from those
coming from kernel classes.)
182
10.1
Object
subclassResponsibility. Abstract methods are implemented by convention with the body self subclassResponsibility. Should an abstract class
be instantiated by accident, then calls to abstract methods will result in Object>>subclassResponsibility being executed.
Object >> subclassResponsibility
"This message sets up a framework for the behavior of the class'
subclasses.
Announce that the subclass should have implemented this message."
self error: 'My subclass should have overridden ', thisContext
sender selector printString
Magnitude, Number, and Boolean are classical examples of abstract classes
that we shall see shortly in this chapter.
Number new + 1
>>> !''Error: My subclass should have overridden #+''!
shouldNotImplement. self shouldNotImplement is sent by convention to
signal that an inherited method is not appropriate for this subclass. This is
generally a sign that something is not quite right with the design of the class
hierarchy. Due to the limitations of single inheritance, however, sometimes it
is very hard to avoid such workarounds.
A typical example is Collection>>remove: which is inherited by Dictionary but flagged as not implemented. (A Dictionary provides removeKey:
instead.)
Testing
The testing methods have nothing to do with SUnit testing! A testing method
is one that lets you ask a question about the state of the receiver and returns a
Boolean.
Numerous testing methods are provided by Object. There are isArray, isBoolean, isBlock, isCollection and so on. Generally such methods are to be
avoided since querying an object for its class is a form of violation of encapsulation. Instead of testing an object for its class, one should simply send a
request and let the object decide how to handle it.
Nevertheless some of these testing methods are undeniably useful. The most
useful are probably ProtoObject>>isNil and Object>>notNil (though the
Null Object design pattern can obviate the need for even these methods).
Initialize
A final key method that occurs not in Object but in ProtoObject is initialize.
183
Basic classes
Figure 10.1: The number hierarchy.
ProtoObject >> initialize
"Subclasses should redefine this method to perform initializations
on instance creation"
The reason this is important is that in Pharo, the default new method defined
for every class in the system will send initialize to newly created instances.
Behavior >> new
"Answer a new initialized instance of the receiver (which is a
class) with no indexable
variables. Fail if the class is indexable."
^ self basicNew initialize
This means that simply by overriding the initialize hook method, new
instances of your class will automatically be initialized. The initialize
method should normally perform a super initialize to establish the class
invariant for any inherited instance variables.
10.2
Numbers
Numbers in Pharo are not primitive data values but true objects. Of course
numbers are implemented efficiently in the virtual machine, but the Number
hierarchy is as perfectly accessible and extensible as any other portion of the
class hierarchy.
The abstract root of this hierarchy is Magnitude, which represents all kinds of
classes supporting comparison operators. Number adds various arithmetic and
other operators as mostly abstract methods. Float and Fraction represent,
184
10.2
Numbers
respectively, floating point numbers and fractional values. Float subclasses
(BoxedFloat64 and SmallFloat64) represent Float on certain architectures.
For example BoxedFloat64 is only available for 64 bit systems. Integer is
also abstract, thus distinguishing between subclasses SmallInteger, LargePositiveInteger and LargeNegativeInteger. For the most part, users do
not need to be aware of the difference between the three Integer classes, as
values are automatically converted as needed.
Magnitude
Magnitude is the parent not only of the Number classes, but also of other
classes supporting comparison operations, such as Character, Duration
and Timespan.
Methods < and = are abstract. The remaining operators are generically defined.
For example:
Magnitude >> < aMagnitude
"Answer whether the receiver is less than the argument."
^ self subclassResponsibility
Magnitude >> > aMagnitude
"Answer whether the receiver is greater than the argument."
^ aMagnitude < self
Number
Similarly, Number defines +, -, * and / to be abstract, but all other arithmetic
operators are generically defined.
All Number objects support various converting operators, such as asFloat
and asInteger. There are also numerous shortcut constructor methods which
generate Durations, such as hour, day and week.
Numbers directly support common math functions such as sin, log, raiseTo:,
squared, sqrt and so on.
The method Number>>printOn: is implemented in terms of the abstract
method Number>>printOn:base:. (The default base is 10.)
Testing methods include even, odd, positive and negative. Unsurprisingly
Number overrides isNumber. More interestingly, isInfinite is defined to
return false.
Truncation methods include floor, ceiling, integerPart, fractionPart and
so on.
1 + 2.5
>>> 3.5 "Addition of two numbers"
185
Basic classes
3.4 * 5
>>> 17.0 "Multiplication of two numbers"
8 / 2
>>> 4 "Division of two numbers"
10 - 8.3
>>> 1.7 "Subtraction of two numbers"
12 = 11
>>> false "Equality between two numbers"
12 ~= 11
>>> true "Test if two numbers are different"
12 > 9
>>> true "Greater than"
12 >= 10
>>> true "Greater or equal than"
12 < 10
>>> false "Smaller than"
100@10
>>> 100@10 "Point creation"
The following example works surprisingly well in Pharo:
1000 factorial / 999 factorial
>>> 1000
Note that 1000 factorial is really calculated, which in many other languages
can be quite difficult to compute. This is an excellent example of automatic
coercion and exact handling of a number.
To do Try to display the result of 1000 factorial. It takes more time to
display it than to calculate it!
Float
Float implements the abstract Number methods for floating point numbers.
More interestingly, Float class (i.e., the class-side of Float) provides methods to return the following constants: e, infinity, nan and pi.
Float pi
>>> 3.141592653589793
Float infinity
>>> Infinity
Float infinity isInfinite
>>> true
186
10.2
Numbers
Fraction
Fractions are represented by instance variables for the numerator and denominator, which should be Integers. Fractions are normally created by
Integer division (rather than using the constructor method Fraction>>numerator:denominator:):
6/8
>>> (3/4)
(6/8) class
>>> Fraction
Multiplying a Fraction by an Integer or another Fraction may yield an
Integer:
6/8 * 4
>>> 3
Integer
Integer is the abstract parent of three concrete integer implementations.
In addition to providing concrete implementations of many abstract Number
methods, it also adds a few methods specific to integers, such as factorial,
atRandom, isPrime, gcd: and many others.
SmallInteger is special in that its instances are represented compactly —
instead of being stored as a reference, a SmallInteger is represented directly
using the bits that would otherwise be used to hold a reference. The first bit
of an object reference indicates whether the object is a SmallInteger or not.
Now the virtual machine abstracts that from you, therefore you cannot see
this directly when inspecting the object.
The class methods minVal and maxVal tell us the range of a SmallInteger:
SmallInteger maxVal = ((2 raisedTo: 30) - 1)
>>> true
SmallInteger minVal = (2 raisedTo: 30) negated
>>> true
When a SmallInteger goes out of this range, it is automatically converted to a
LargePositiveInteger or a LargeNegativeInteger, as needed:
(SmallInteger maxVal + 1) class
>>> LargePositiveInteger
(SmallInteger minVal - 1) class
>>> LargeNegativeInteger
Large integers are similarly converted back to small integers when appropriate.
187
Basic classes
As in most programming languages, integers can be useful for specifying
iterative behaviour. There is a dedicated method timesRepeat: for evaluating
a block repeatedly. We have already seen a similar example in Chapter :
Syntax in a Nutshell.
| n |
n := 2.
3 timesRepeat: [ n := n * n ].
n
>>> 256
10.3
Characters
Character is defined a subclass of Magnitude. Printable characters are represented in Pharo as $<char>. For example:
$a < $b
>>> true
Non-printing characters can be generated by various class methods. Character class>>value: takes the Unicode (or ASCII) integer value as argument
and returns the corresponding character. The protocol accessing untypeable characters contains a number of convenience constructor methods
such as backspace, cr, escape, euro, space, tab, and so on.
Character space = (Character value: Character space asciiValue)
>>> true
The printOn: method is clever enough to know which of the three ways to
generate characters offers the most appropriate representation:
Character value: 1
>>> Character home
Character value: 2
>>> Character value: 2
Character value: 32
>>> Character space
Character value: 97
>>> $a
Various convenient testing methods are built in: isAlphaNumeric, isCharacter, isDigit, isLowercase, isVowel, and so on.
To convert a Character to the string containing just that character, send
asString. In this case asString and printString yield different results:
$a asString
>>> 'a'
188
10.4
Strings
Figure 10.2: The String Hierarchy.
$a
>>> $a
$a printString
>>> '$a'
Like SmallInteger, a Character is a immediate value not a object reference.
Most of the time you won’t see any difference and can use objects of class
Character like any other too. But this means, equal value characters are
always identical:
(Character value: 97) == $a
>>> true
10.4
Strings
A String is an indexed Collection that holds only Characters.
In fact, String is abstract and Pharo strings are actually instances of the
concrete class ByteString.
'hello world' class
>>> ByteString
The other important subclass of String is Symbol. The key difference is that
there is only ever a single instance of Symbol with a given value. (This is
sometimes called the unique instance property). In contrast, two separately
constructed Strings that happen to contain the same sequence of characters
will often be different objects.
189
Basic classes
'hel','lo' == 'hello'
>>> false
('hel','lo') asSymbol == #hello
>>> true
Another important difference is that a String is mutable, whereas a Symbol is
immutable.
'hello' at: 2 put: $u; yourself
>>> 'hullo'
#hello at: 2 put: $u
>>> error!
It is easy to forget that since strings are collections, they understand the same
messages that other collections do:
#hello indexOf: $o
>>> 5
Although String does not inherit from Magnitude, it does support the usual
comparing methods, <, = and so on. In addition, String>>match: is useful for
some basic glob-style pattern-matching:
'*or*' match: 'zorro'
>>> true
Regular expressions will be discussed in more detail in Chapter : Regular
Expressions in Pharo.
Strings support a rather large number of conversion methods. Many of these
are shortcut constructor methods for other classes, such as asDate, asInteger
and so on. There are also a number of useful methods for converting a string
to another string, such as capitalized and translateToLowercase.
For more on strings and collections, see Chapter : Collections.
10.5
Booleans
The class Boolean offers a fascinating insight into how much of the Pharo
language has been pushed into the class library. Boolean is the abstract
superclass of the singleton classes True and False.
Most of the behaviour of Booleans can be understood by considering the
method ifTrue:ifFalse:, which takes two Blocks as arguments.
4 factorial > 20
ifTrue: [ 'bigger' ]
ifFalse: [ 'smaller' ]
>>> 'bigger'
190
10.5
Booleans
Figure 10.3: The Boolean Hierarchy.
The method ifTrue:ifFalse: is abstract in class Boolean. The implementations in its concrete subclasses are both trivial:
True >> ifTrue: trueAlternativeBlock ifFalse: falseAlternativeBlock
^ trueAlternativeBlock value
False >> ifTrue: trueAlternativeBlock ifFalse: falseAlternativeBlock
^ falseAlternativeBlock value
Each of them execute the correct block depending on the receiver of the
message. In fact, this is the essence of OOP: when a message is sent to an
object, the object itself determines which method will be used to respond.
In this case an instance of True simply executes the true alternative, while
an instance of False executes the false alternative. All the abstract Boolean
methods are implemented in this way for True and False. For example the
implementation of negation (message not) is defined the same way:
True >> not
"Negation--answer false since the receiver is true."
^ false
False >> not
"Negation--answer true since the receiver is false."
^ true
Booleans offer several useful convenience methods, such as ifTrue:, ifFalse:, and ifFalse:ifTrue. You also have the choice between eager and
lazy conjunctions and disjunctions.
( 1 > 2 ) & ( 3 < 4 )
>>> false "Eager, must evaluate both sides"
( 1 > 2 ) and: [ 3 < 4 ]
>>> false "Lazy, only evaluate receiver"
( 1 > 2 ) and: [ ( 1 / 0 ) > 0 ]
>>> false "argument block is never executed, so no exception"
191
Basic classes
In the first example, both Boolean subexpressions are executed, since & takes
a Boolean argument. In the second and third examples, only the first is executed, since and: expects a Block as its argument. The Block is executed only
if the first argument is true.
To do Try to imagine how and: and or: are implemented. Check the
implementations in Boolean, True and False.
10.6
Chapter summary
• If you override = then you should override hash as well.
• Override postCopy to correctly implement copying for your objects.
• Use self halt. to set a breakpoint.
• Return self subclassResponsibility to make a method abstract.
• To give an object a String representation you should override printOn:.
• Override the hook method initialize to properly initialize instances.
• Number methods automatically convert between Floats, Fractions and
Integers.
• Fractions truly represent rational numbers rather than floats.
• All Characters are like unique instances.
• Strings are mutable; Symbols are not. Take care not to mutate string
literals, however!
• Symbols are unique; Strings are not.
• Strings and Symbols are Collections and therefore support the usual
Collection methods.
192
CHAPTER
11
Collections
11.1
Introduction
The collection classes form a loosely-defined group of general-purpose subclasses of Collection and Stream. Many of these (like Bitmap, FileStream
and CompiledMethod) are special-purpose classes crafted for use in other parts
of the system or in applications, and hence not categorized as Collections by
the system organization. For the purposes of this chapter, we use the term
Collection Hierarchy to mean Collection and its subclasses that are also in the
packages labelled Collections-*. We use the term Stream Hierarchy to mean
Stream and its subclasses that are also in the Collections-Streams packages.
In this chapter we focus mainly on the subset of collection classes shown in
Figure 11.1. Streams will be discussed separately in Chapter : Streams.
11.2
The varieties of collections
To make good use of the collection classes, the reader needs at least a superficial knowledge of the wide variety of collections that they implement, and
their commonalities and differences.
Programming with collections using high-order functions rather than individual elements is an important way to raise the level of abstraction of a program.
The Lisp function map, which applies an argument function to every element
of a list and returns a new list containing the results is an early example of
this style. Following its Smalltalk root, Pharo adopts this collection-based
high-order programming as a central tenet. Modern functional programming
languages such as ML and Haskell have followed Smalltalk’s lead.
193
Collections
Figure 11.1: Some of the key collection classes in Pharo.
Why is this a good idea? Suppose you have a data structure containing a
collection of student records, and wish to perform some action on all of the
students that meet some criterion. Programmers raised to use an imperative
language will immediately reach for a loop. But the Pharo programmer will
write:
students
select: [ :each | each gpa < threshold ]
This expression returns a new collection containing precisely those elements
of students for which the block (the bracketed function) returns true. The
block can be thought of as a lambda-expression defining an anonymous
function x. x gpa < threshold. This code has the simplicity and elegance of
a domain-specific query language.
The message select: is understood by all collections in Pharo. There is no
need to find out if the student data structure is an array or a linked list: the
select: message is understood by both. Note that this is quite different from
using a loop, where one must know whether students is an array or a linked
list before the loop can be set up.
In Pharo, when one speaks of a collection without being more specific about
the kind of collection, one means an object that supports well-defined protocols for testing membership and enumerating the elements. All collections
understand the testing messages includes:, isEmpty and occurrencesOf:.
All collections understand the enumeration messages do:, select:, reject:
(which is the opposite of select:), collect: (which is like Lisp’s map), detect:ifNone:, inject:into: (which performs a left fold) and many more. It
194
11.2
The varieties of collections
is the ubiquity of this protocol, as well as its variety, that makes it so powerful.
The table below summarizes the standard protocols supported by most of
the classes in the collection hierarchy. These methods are defined, redefined,
optimized or occasionally even forbidden by subclasses of Collection.
Protocol
Methods
accessing
testing
adding
removing
enumerating
size, capacity, at:, at:put:
isEmpty, includes:, contains:, occurrencesOf:
add:, addAll:
remove:, remove:ifAbsent:, removeAll:
do:, collect:, select:, reject:
detect:, detect:ifNone:, inject:into:
asBag, asSet, asOrderedCollection, asSortedCollection,
asArray, asSortedCollection:
with:, with:with:, with:with:with:,
with:with:with:with:, withAll:
converting
creation
Beyond this basic uniformity, there are many different kinds of collection
either supporting different protocols, or providing different behaviour for the
same requests. Let us briefly survey some of the key differences:
Sequenceable: Instances of all subclasses of SequenceableCollection start
from a first element and proceed in a well-defined order to a last
element. Instances of Set, Bag and Dictionary, on the other hand, are
not sequenceable.
Sortable: A SortedCollection maintains its elements in sort order.
Indexable: Most sequenceable collections are also indexable, that is, elements can be retrieved with message at: anIndex. Array is the familiar
indexable data structure with a fixed size; anArray at: n retrieves
the nth element of anArray, and anArray at: n put: v changes the
nth element to v. LinkedLists and SkipLists are sequenceable but not
indexable, that is, they understand first and last, but not the message
at:.
Keyed: Instances of Dictionary and its subclasses are accessed by keys instead of indices.
Mutable: Most collections are mutable, but Intervals and Symbols are not.
An Interval is an immutable collection representing a range of Integers. For example, 5 to: 16 by: 2 is an interval that contains the
elements 5, 7, 9, 11, 13 and 15. It is indexable with message at: anIndex, but cannot be changed with message at: anIndex put: aValue.
Growable: Instances of Interval and Array are always of a fixed size. Other
kinds of collections (sorted collections, ordered collections, and linked
lists) can grow after creation. The class OrderedCollection is more
195
Collections
Figure 11.2: Some collection classes categorized by implementation technique.
general than Array; the size of an OrderedCollection grows on demand, and it defines messages addFirst: anElement and addLast:
anElement as well as messages at: anIndex and at: anIndex put:
aValue.
Accepts duplicates: A Set filters out duplicates, but a Bag does not. Classes
Dictionary, Set and Bag use the = method provided by the elements;
the Identity variants of these classes use the == method, which tests
whether the arguments are the same object, and the Pluggable variants
use an arbitrary equivalence relation supplied by the creator of the
collection.
Heterogeneous: Most collections will hold any kind of element. A String,
CharacterArray or Symbol, however, only holds Characters. An Array will hold any mix of objects, but a ByteArray only holds Bytes. A
LinkedList is constrained to hold elements that conform to the Link
accessing protocol.
11.3
Collection implementations
These categorizations by functionality are not our only concern; we must also
consider how the collection classes are implemented. As shown in Figure 11.2,
five main implementation techniques are employed.
• Arrays store their elements in the (indexable) instance variables of the
collection object itself; as a consequence, arrays must be of a fixed size,
but can be created with a single memory allocation.
• OrderedCollections and SortedCollections store their elements
in an array that is referenced by one of the instance variables of the
collection. Consequently, the internal array can be replaced with a
larger one if the collection grows beyond its storage capacity.
• The various kinds of set and dictionary also reference a subsidiary array
for storage, but use the array as a hash table. Bags use a subsidiary
196
11.4
Examples of key classes
Dictionary, with the elements of the bag as keys and the number of
occurrences as values.
• LinkedLists use a standard singly-linked representation.
• Intervals are represented by three integers that record the two endpoints and the step size.
In addition to these classes, there are also weak variants of Array, Set and of
the various kinds of dictionary. These collections hold onto their elements
weakly, i.e., in a way that does not prevent the elements from being garbage
collected. The Pharo virtual machine is aware of these classes and handles
them specially.
11.4
Examples of key classes
We present now the most common or important collection classes using
simple code examples. The main protocols of collections are:
• messages at:, at:put: — to access an element,
• messages add:, remove: — to add or remove an element,
• messages size, isEmpty, include: — to get some information about the
collection,
• messages do:, collect:, select: — to iterate over the collection.
Each collection may implement (or not) such protocols, and when they do,
they interpret them to fit with their semantics. We suggest you browse the
classes themselves to identify specific and more advanced protocols.
We will focus on the most common collection classes: OrderedCollection,
Set, SortedCollection, Dictionary, Interval, and Array.
There are several ways to create instances of
collections. The most generic ones use the message new: aSize and with:
anElement.
Common creation protocol.
• new: anInteger creates a collection of size anInteger whose elements
will all be nil.
• with: anObject creates a collection and adds anObject to the created
collection.
Different collections will realize this behaviour differently.
You can create collections with initial elements using the methods with:,
with:with: etc. for up to six elements.
Array with: 1
>>> #(1)
197
Collections
Array with: 1 with: 2
>>> #(1 2)
Array with: 1 with: 2 with: 3
>>> #(1 2 3)
Array with: 1 with: 2 with: 3 with: 4
>>> #(1 2 3 4)
Array with: 1 with: 2 with: 3 with: 4 with: 5
>>> #(1 2 3 4 5)
Array with: 1 with: 2 with: 3 with: 4 with: 5 with: 6
>>> #(1 2 3 4 5 6)
You can also use message addAll: aCollection to add all elements of one
kind of collection to another kind:
(1 to: 5) asOrderedCollection addAll: '678'; yourself
>>> an OrderedCollection(1 2 3 4 5 6 7 8)
Take care that addAll: returns its argument, and not the receiver!
You can also create many collections with withAll: aCollection.
Array withAll: #(7 3 1 3)
>>> #(7 3 1 3)
OrderedCollection withAll: #(7 3 1 3)
>>> an OrderedCollection(7 3 1 3)
SortedCollection withAll: #(7 3 1 3)
>>> a SortedCollection(1 3 3 7)
Set withAll: #(7 3 1 3)
>>> a Set(7 1 3)
Bag withAll: #(7 3 1 3)
>>> a Bag(7 1 3 3)
Array
An Array is a fixed-sized collection of elements accessed by integer indices.
Contrary to the C convention in Pharo, the first element of an array is at
position 1 and not 0. The main protocol to access array elements is the method
at: and at:put:. at: anInteger returns the element at index anInteger.
at: anInteger put: anObject puts anObject at index anInteger. Arrays
are fixed-size collections therefore we cannot add or remove elements at the
end of an array. The following code creates an array of size 5, puts values in
the first 3 locations and returns the first element.
| anArray |
anArray := Array new: 5.
anArray at: 1 put: 4.
198
11.4
Examples of key classes
anArray at: 2 put: 3/2.
anArray at: 3 put: 'ssss'.
anArray at: 1
>>> 4
There are several ways to create instances of the class Array. We can use new:,
with:, and the constructs #( ) (literal arrays) and { . } (dynamic compact
syntax).
Creation with new:
The message new: anInteger creates an array of size
anInteger. Array new: 5 creates an array of size 5. (Note: the value of each
element is initialized to nil).
Creation using with:
The with:* messages allow one to specify the value of
the elements. The following code creates an array of three elements consisting
of the number 4, the fraction 3/2 and the string 'lulu'.
Array with: 4 with: 3/2 with: 'lulu'
>>> {4. (3/2). 'lulu'}
Literal creation with #()
The expression #() creates literal arrays with
constants or literal elements that have to be known when the expression is
compiled, and not when it is executed. The following code creates an array
of size 2 where the first element is the (literal) number 1 and the second the
(literal) string 'here'.
#(1 'here') size
>>> 2
Now, if you execute the expression #(1+2), you do not get an array with a
single element 3 but instead you get the array #(1 #+ 2) i.e., with three
elements: 1, the symbol #+ and the number 2.
#(1+2)
>>> #(1 #+ 2)
This occurs because the construct #() does not execute the expressions it
contains. The elements are only objects that are created when parsing the
expression (called literal objects). The expression is scanned and the resulting
elements are fed to a new array. Literal arrays contain numbers, nil, true,
false, symbols, strings and other literal arrays. During the execution of #()
expressions, there are no messages sent.
Dynamic creation with { . }
Finally, you can create a dynamic array using
the construct { . }. The expression { a . b } is totally equivalent to Array
with: a with: b. This means in particular that the expressions enclosed by {
and } are executed (contrary to the case of #())
199
Collections
{ 1 + 2 }
>>> #(3)
{(1/2) asFloat} at: 1
>>> 0.5
{10 atRandom. 1/3} at: 2
>>> (1/3)
Element Access Elements of all sequenceable collections can be accessed
with messages at: anIndex and at: anIndex put: anObject.
| anArray |
anArray := #(1 2 3 4 5 6) copy.
anArray at: 3 >>> 3
anArray at: 3 put: 33.
anArray at: 3
>>> 33
Be careful: as a general principle do not modify literal arrays! Literal arrays
are kept in compiled method literal frames (a space where literals appearing
in a method are stored), therefore unless you copy the array, the second time
you execute the code your literal array may not have the value you expect. In
the example, without copying the array, the second time around, the literal
#(1 2 3 4 5 6) will actually be #(1 2 33 4 5 6)! Dynamic arrays do not
have this problem because they are not stored in literal frames.
OrderedCollection
OrderedCollection is one of the collections that can grow, and to which
elements can be added sequentially. It offers a variety of messages such as
add:, addFirst:, addLast:, and addAll:.
| ordCol |
ordCol := OrderedCollection new.
ordCol add: 'Seaside'; add: 'SmalltalkHub'; addFirst: 'Monticello'.
ordCol
>>> an OrderedCollection('Monticello' 'Seaside' 'SmalltalkHub')
The message remove: anObject removes the first occurrence of an object from the collection. If the collection does not include such
an object, it raises an error.
Removing Elements
ordCol add: 'Monticello'.
ordCol remove: 'Monticello'.
ordCol
>>> an OrderedCollection('Seaside' 'SmalltalkHub' 'Monticello')
200
11.4
Examples of key classes
There is a variant of remove: named remove:ifAbsent: that allows one to
specify as second argument a block that is executed in case the element to be
removed is not in the collection.
result := ordCol remove: 'zork' ifAbsent: [33].
result
>>> 33
It is possible to get an OrderedCollection from an Array (or
any other collection) by sending the message asOrderedCollection:
Conversion
#(1 2 3) asOrderedCollection
>>> an OrderedCollection(1 2 3)
'hello' asOrderedCollection
>>> an OrderedCollection($h $e $l $l $o)
Interval
The class Interval represents ranges of numbers. For example, the interval of
numbers from 1 to 100 is defined as follows:
Interval from: 1 to: 100
>>> (1 to: 100)
The result of printString reveals that the class Number provides us with a
convenience method called to: to generate intervals:
(Interval from: 1 to: 100) = (1 to: 100)
>>> true
We can use Interval class>>from:to:by: or Number>>to:by: to specify the
step between two numbers as follow:
(Interval from: 1 to: 100 by: 0.5) size
>>> 199
(1 to: 100 by: 0.5) at: 198
>>> 99.5
(1/2 to: 54/7 by: 1/3) last
>>> (15/2)
Dictionary
Dictionaries are important collections whose elements are accessed using
keys. Among the most commonly used messages of dictionary you will find
at: aKey, at: aKey put: aValue, at: aKey ifAbsent: aBlock, keys, and
values.
| colors |
colors := Dictionary new.
201
Collections
colors at: #yellow put: Color yellow.
colors at: #blue put: Color blue.
colors at: #red put: Color red.
colors at: #yellow
>>> Color yellow
colors keys
>>> #(#red #blue #yellow)
colors values
>>> {Color blue . Color yellow . Color red}
Dictionaries compare keys by equality. Two keys are considered to be the
same if they return true when compared using =. A common and difficult to
spot bug is to use as key an object whose = method has been redefined but not
its hash method. Both methods are used in the implementation of dictionary
and when comparing objects.
In its implementation, a Dictionary can be seen as consisting of a set of (key
value) associations created using the message ->. We can create a Dictionary
from a collection of associations, or we may convert a dictionary to an array of
associations.
| colors |
colors := Dictionary newFrom: { #blue->Color blue . #red->Color red .
#yellow->Color yellow }.
colors removeKey: #blue.
colors associations
>>> {#yellow->Color yellow. #red->Color red}
IdentityDictionary While a dictionary uses the result of the messages
= and hash to determine if two keys are the same, the class IdentityDictionary uses the identity (message ==) of the key instead of its values, i.e., it
considers two keys to be equal only if they are the same object.
Often Symbols are used as keys, in which case it is natural to use an IdentityDictionary, since a Symbol is guaranteed to be globally unique. If, on the
other hand, your keys are Strings, it is better to use a plain Dictionary, or
you may get into trouble:
a := 'foobar'.
b := a copy.
trouble := IdentityDictionary new.
trouble at: a put: 'a'; at: b put: 'b'.
trouble at: a
>>> 'a'
trouble at: b
>>> 'b'
trouble at: 'foobar'
>>> 'a'
202
11.4
Examples of key classes
Since a and b are different objects, they are treated as different objects. Interestingly, the literal 'foobar' is allocated just once, so is really the same
object as a. You don’t want your code to depend on behaviour like this! A
plain Dictionary would give the same value for any key equal to 'foobar'.
Use only globally unique objects (like Symbols or SmallIntegers) as keys for
an IdentityDictionary, and Strings (or other objects) as keys for a plain
Dictionary.
Note that the expression Smalltalk globals returns an instance of SystemDictionary, a subclass of IdentityDictionary, hence all its keys are Symbols
(actually, ByteSymbols, which contain only 8-bit characters).
Smalltalk globals keys collect: [ :each | each class ] as:Set
>>> a Set(ByteSymbol)
Here we are using collect:as: to specify the result collection to be of class
Set, that way we collect each kind of class used as a key only once.
Set
The class Set is a collection which behaves as a mathematical set, i.e., as
a collection with no duplicate elements and without any order. In a Set,
elements are added using the message add: and they cannot be accessed using
the message at:. Objects put in a set should implement the methods hash and
=.
s := Set new.
s add: 4/2; add: 4; add:2.
s size
>>> 2
You can also create sets using Set class>>newFrom: or the conversion message Collection>>asSet:
(Set newFrom: #( 1 2 3 1 4 )) = #(1 2 3 4 3 2 1) asSet
>>> true
asSet offers us a convenient way to eliminate duplicates from a collection:
{ Color black. Color white. (Color red + Color blue + Color green) }
asSet size
>>> 2
Note that red + blue + green = white.
A Bag is much like a Set except that it does allow duplicates:
{ Color black. Color white. (Color red + Color blue + Color green) }
asBag size
>>> 3
203
Collections
The set operations union, intersection and membership test are implemented
by the Collection messages union:, intersection:, and includes:. The
receiver is first converted to a Set, so these operations work for all kinds of
collections!
(1 to: 6) union: (4 to: 10)
>>> a Set(1 2 3 4 5 6 7 8 9 10)
'hello' intersection: 'there'
>>> 'he'
#Smalltalk includes: $k
>>> true
As we explain below, elements of a set are accessed using iterators (see Section
11.5).
SortedCollection
In contrast to an OrderedCollection, a SortedCollection maintains its
elements in sort order. By default, a sorted collection uses the message <=
to establish sort order, so it can sort instances of subclasses of the abstract
class Magnitude, which defines the protocol of comparable objects (<, =, >, >=,
between:and:...). (See Chapter : Basic Classes).
You can create a SortedCollection by creating a new instance and adding
elements to it:
SortedCollection new add: 5; add: 2; add: 50; add: -10; yourself.
>>> a SortedCollection(-10 2 5 50)
More usually, though, one will send the conversion message asSortedCollection to an existing collection:
#(5 2 50 -10) asSortedCollection
>>> a SortedCollection(-10 2 5 50)
'hello' asSortedCollection
>>> a SortedCollection($e $h $l $l $o)
How do you get a String back from this result? asString unfortunately
returns the printString representation, which is not what we want:
'hello' asSortedCollection asString
>>> 'a SortedCollection($e $h $l $l $o)'
The correct answer is to either use String class>>newFrom:, String class>>withAll: or Object>>as::
'hello' asSortedCollection as: String
>>> 'ehllo'
String newFrom: 'hello' asSortedCollection
>>> 'ehllo'
204
11.4
Examples of key classes
String withAll: 'hello' asSortedCollection
>>> 'ehllo'
It is possible to have different kinds of elements in a SortedCollection as
long as they are all comparable. For example, we can mix different kinds of
numbers such as integers, floats and fractions:
{ 5 . 2/ -3 . 5.21 } asSortedCollection
>>> a SortedCollection((-2/3) 5 5.21)
Imagine that you want to sort objects that do not define the method <= or
that you would like to have a different sorting criterion. You can do this by
supplying a two argument block, called a sortblock, to the sorted collection.
For example, the class Color is not a Magnitude and it does not implement the
method <=, but we can specify a block stating that the colors should be sorted
according to their luminance (a measure of brightness).
col := SortedCollection
sortBlock: [ :c1 :c2 | c1 luminance <= c2 luminance ].
col addAll: { Color red . Color yellow . Color white . Color black }.
col
>>> a SortedCollection(Color black Color red Color yellow Color white)
String
In Pharo, a String is a collection of Characters. It is sequenceable, indexable,
mutable and homogeneous, containing only Character instances. Like Arrays,
Strings have a dedicated syntax, and are normally created by directly specifying a String literal within single quotes, but the usual collection creation
methods will work as well.
'Hello'
>>> 'Hello'
String with: $A
>>> 'A'
String with: $h with: $i with: $!
>>> 'hi!'
String newFrom: #($h $e $l $l $o)
>>> 'hello'
In actual fact, String is abstract. When we instantiate a String we actually
get either an 8-bit ByteString or a 32-bit WideString. To keep things simple,
we usually ignore the difference and just talk about instances of String.
While strings are delimited by single quotes, a string can contain a single
quote: to define a string with a single quote we should type it twice. Note that
the string will contain only one element and not two as shown below:
205
Collections
'l''idiot' at: 2
>>> $'
'l''idiot' at: 3
>>> $i
Two instances of String can be concatenated with a comma.
s := 'no', ' ', 'worries'.
s
>>> 'no worries'
Since a string is a mutable collection we can also change it using the message
at:put:.
s at: 4 put: $h; at: 5 put: $u.
s
>>> 'no hurries'
Note that the comma method is defined by Collection, so it will work for any
kind of collection!
(1 to: 3), '45'
>>> #(1 2 3 $4 $5)
We can also modify an existing string using replaceAll:with: or replaceFrom:to:with: as shown below. Note that the number of characters and the
interval should have the same size.
s replaceAll: $n with: $N.
s
>>> 'No hurries'
s replaceFrom: 4 to: 5 with: 'wo'.
s
>>> 'No worries'
In contrast to the methods described above, the method copyReplaceAll:
creates a new string. (Curiously, here the arguments are substrings rather
than individual characters, and their sizes do not have to match.)
s copyReplaceAll: 'rries' with: 'mbats'
>>> 'No wombats'
A quick look at the implementation of these methods reveals that they are
defined not only for Strings, but for any kind of SequenceableCollection,
so the following also works:
(1 to: 6) copyReplaceAll: (3 to: 5) with: { 'three' . 'etc.' }
>>> #(1 2 'three' 'etc.' 6)
It is possible to ask whether a pattern matches a string by
sending the match: message. The pattern can use * to match an arbitrary
String matching
206
11.4
Examples of key classes
series of characters and # to match a single character. Note that match: is
sent to the pattern and not the string to be matched.
'Linux *' match: 'Linux mag'
>>> true
'GNU#Linux #ag' match: 'GNU/Linux tag'
>>> true
More advanced pattern matching facilities are also available in the Regex
package.
For substring manipulation we can use messages like first,
first:, allButFirst:, copyFrom:to: and others, defined in SequenceableCollection.
Substrings
'alphabet' at: 6
>>> $b
'alphabet' first
>>> $a
'alphabet' first: 5
>>> 'alpha'
'alphabet' allButFirst: 3
>>> 'habet'
'alphabet' copyFrom: 5 to: 7
>>> 'abe'
'alphabet' copyFrom: 3 to: 3
>>> 'p' (not $p)
Be aware that result type can be different, depending on the method used.
Most of the substring-related methods return String instances. But the
messages that always return one element of the String collection, return a
Character instance (for example, 'alphabet' at: 6 returns the character
$b). For a complete list of substring-related messages, browse the SequenceableCollection class (especially the accessing protocol).
The following examples illustrate the use of isEmpty,
includes: and anySatisfy: which are also messages defined not only on
Strings but more generally on collections.
Some tests on strings
'Hello' isEmpty
>>> false
'Hello' includes: $a
>>> false
'JOE' anySatisfy: [ :c | c isLowercase ]
>>> false
207
Collections
'Joe' anySatisfy: [ :c | c isLowercase ]
>>> true
There are three messages that are useful to manage string
templating: format:, expandMacros and expandMacrosWith:.
String templating
'{1} is {2}' format: {'Pharo' . 'cool'}
>>> 'Pharo is cool'
The messages of the expandMacros family offer variable substitution, using
<n> for carriage return, <t> for tabulation, <1s>, <2s>, <3s> for arguments
(<1p>, <2p>, surrounds the string with single quotes), and <1?value1:value2>
for conditional.
'look-<t>-here' expandMacros
>>> 'look- -here'
'<1s> is <2s>' expandMacrosWith: 'Pharo' with: 'cool'
>>> 'Pharo is cool'
'<2s> is <1s>' expandMacrosWith: 'Pharo' with: 'cool'
>>> 'cool is Pharo'
'<1p> or <1s>' expandMacrosWith: 'Pharo' with: 'cool'
>>> '''Pharo'' or Pharo'
'<1?Quentin:Thibaut> plays' expandMacrosWith: true
>>> 'Quentin plays'
'<1?Quentin:Thibaut> plays' expandMacrosWith: false
>>> 'Thibaut plays'
The class String offers numerous other utilities
including the messages asLowercase, asUppercase and capitalized.
Some other utility methods
'XYZ' asLowercase
>>> 'xyz'
'xyz' asUppercase
>>> 'XYZ'
'hilaire' capitalized
>>> 'Hilaire'
'Hilaire' uncapitalized
>>> 'hilaire'
'1.54' asNumber
>>> 1.54
'this sentence is without a doubt far too long' contractTo: 20
>>> 'this sent...too long'
208
11.5
Collection iterators
Note that there is generally a difference between asking an object its string
representation by sending the message printString and converting it to a
string by sending the message asString. Here is an example of the difference.
#ASymbol printString
>>> '#ASymbol'
#ASymbol asString
>>> 'ASymbol'
A symbol is similar to a string but is guaranteed to be globally unique. For this
reason symbols are preferred to strings as keys for dictionaries, in particular
for instances of IdentityDictionary. See also Chapter : Basic Classes for
more about String and Symbol.
11.5
Collection iterators
In Pharo loops and conditionals are simply messages sent to collections or
other objects such as integers or blocks (see also Chapter : Syntax in a Nutshell). In addition to low-level messages such as to:do: which evaluates a
block with an argument ranging from an initial to a final number, the collection hierarchy offers various high-level iterators. Using such iterators will
make your code more robust and compact.
Iterating (do:)
The method do: is the basic collection iterator. It applies its argument (a
block taking a single argument) to each element of the receiver. The following
example prints all the strings contained in the receiver to the transcript.
#('bob' 'joe' 'toto') do: [:each | Transcript show: each; cr].
There are a lot of variants of do:, such as do:without:, doWithIndex: and reverseDo:.
Variants.
For the indexed collections (Array, OrderedCollection, SortedCollection)
the message doWithIndex: also gives access to the current index. This message is related to to:do: which is defined in class Number.
#('bob' 'joe' 'toto')
doWithIndex: [ :each :i | (each = 'joe') ifTrue: [ ^ i ] ]
>>> 2
For ordered collections, the message reverseDo: walks the collection in the
reverse order.
The following code shows an interesting message: do:separatedBy: which
executes the second block only in between two elements.
209
Collections
| res |
res := ''.
#('bob' 'joe' 'toto')
do: [ :e | res := res, e ]
separatedBy: [ res := res, '.' ].
res
>>> 'bob.joe.toto'
Note that this code is not especially efficient since it creates intermediate
strings and it would be better to use a write stream to buffer the result (see
Chapter : Streams):
String streamContents: [ :stream |
#('bob' 'joe' 'toto') asStringOn: stream delimiter: '.' ]
>>> 'bob.joe.toto'
Dictionaries
When the message do: is sent to a dictionary, the elements taken into account
are the values, not the associations. The proper messages to use are keysDo:,
valuesDo:, and associationsDo:, which iterate respectively on keys, values
or associations.
colors := Dictionary newFrom: { #yellow -> Color yellow. #blue -> Color
blue. #red -> Color red }.
colors keysDo: [ :key | Transcript show: key; cr ].
colors valuesDo: [ :value | Transcript show: value; cr ].
colors associationsDo: [:value | Transcript show: value; cr].
Collecting results (collect:)
If you want to apply a function to the elements of a collection and get a new
collection with the results, rather than using do:, you are probably better off
using collect:, or one of the other iterator methods. Most of these can be
found in the enumerating protocol of Collection and its subclasses.
Imagine that we want a collection containing the doubles of the elements in
another collection. Using the method do: we must write the following:
| double |
double := OrderedCollection new.
#(1 2 3 4 5 6) do: [ :e | double add: 2 * e ].
double
>>> an OrderedCollection(2 4 6 8 10 12)
The message collect: executes its argument block for each element and
returns a new collection containing the results. Using collect: instead, the
code is much simpler:
210
11.5
Collection iterators
#(1 2 3 4 5 6) collect: [ :e | 2 * e ]
>>> #(2 4 6 8 10 12)
The advantages of collect: over do: are even more important in the following example, where we take a collection of integers and generate as a result a
collection of absolute values of these integers:
aCol := #( 2 -3 4 -35 4 -11).
result := aCol species new: aCol size.
1 to: aCol size do: [ :each | result at: each put: (aCol at: each) abs
].
result
>>> #(2 3 4 35 4 11)
Contrast the above with the much simpler following expression:
#( 2 -3 4 -35 4 -11) collect: [ :each | each abs ]
>>> #(2 3 4 35 4 11)
A further advantage of the second solution is that it will also work for sets and
bags. Generally you should avoid using do:, unless you want to send messages
to each of the elements of a collection.
Note that sending the message collect: returns the same kind of collection
as the receiver. For this reason the following code fails. (A String cannot hold
integer values.)
'abc' collect: [ :ea | ea asciiValue ]
>>> "error!"
Instead we must first convert the string to an Array or an OrderedCollection:
'abc' asArray collect: [ :ea | ea asciiValue ]
>>> #(97 98 99)
Actually collect: is not guaranteed to return a collection of exactly the same
class as the receiver, but only the same species. In the case of an Interval, the
species is an Array!
(1 to: 5) collect: [ :ea | ea * 2 ]
>>> #(2 4 6 8 10)
Selecting and rejecting elements
The message select: returns the elements of the receiver that satisfy a
particular condition:
(2 to: 20) select: [ :each | each isPrime ]
>>> #(2 3 5 7 11 13 17 19)
The message reject: does the opposite:
211
Collections
(2 to: 20) reject: [ :each | each isPrime ]
>>> #(4 6 8 9 10 12 14 15 16 18 20)
Identifying an element with detect:
The message detect: returns the first element of the receiver that matches
block argument.
'through' detect: [ :each | each isVowel ]
>>> $o
The message detect:ifNone: is a variant of the method detect:. Its second
block is evaluated when there is no element matching the block.
Smalltalk globals allClasses
detect: [:each | '*cobol*' match: each asString]
ifNone: [ nil ]
>>> nil
Accumulating results with inject:into:
Functional programming languages often provide a higher-order function
called fold or reduce to accumulate a result by applying some binary operator
iteratively over all elements of a collection. In Pharo this is done by Collection>>inject:into:.
The first argument is an initial value, and the second argument is a twoargument block which is applied to the result this far, and each element in
turn.
A trivial application of inject:into: is to produce the sum of a collection
of numbers. In Pharo we could write this expression to sum the first 100
integers:
(1 to: 100) inject: 0 into: [ :sum :each | sum + each ]
>>> 5050
Another example is the following one-argument block which computes factorials:
factorial := [ :n | (1 to: n) inject: 1 into: [ :product :each |
product * each ] ].
factorial value: 10
>>> 3628800
Other messages
There are many other iterator messages. Just check the Collection class.
212
11.6
Some hints for using collections
count: The message count: returns the number of elements satisfying a
condition. The condition is represented as a boolean block.
Smalltalk globals allClasses
count: [ :each | 'Collection*' match: each asString ]
>>> 6
includes: The message includes: checks whether the argument is contained in the collection.
| colors |
colors := {Color white . Color yellow . Color blue . Color orange}.
colors includes: Color blue.
>>> true
anySatisfy: The message anySatisfy: answers true if at least one element
of the collection satisfies the condition represented by the argument.
colors anySatisfy: [ :c | c red > 0.5 ]
>>> true
11.6
Some hints for using collections
A common mistake with add:
The following error is one of the most fre-
quent Smalltalk mistakes.
| collection |
collection := OrderedCollection new add: 1; add: 2.
collection
>>> 2
Here the variable collection does not hold the newly created collection but
rather the last number added. This is because the method add: returns the
element added and not the receiver.
The following code yields the expected result:
| collection |
collection := OrderedCollection new.
collection add: 1; add: 2.
collection
>>> an OrderedCollection(1 2)
You can also use the message yourself to return the receiver of a cascade of
messages:
| collection |
collection := OrderedCollection new add: 1; add: 2; yourself
>>> an OrderedCollection(1 2)
213
Collections
Another mistake
you may make is to remove an element from a collection you are currently
iterating over.
Removing an element of the collection you are iterating on
| range |
range := (2 to: 20) asOrderedCollection.
range do: [ :aNumber | aNumber isPrime
ifFalse: [ range remove: aNumber ] ].
range
>>> an OrderedCollection(2 3 5 7 9 11 13 15 17 19)
This result is clearly incorrect since 9 and 15 should have been filtered out!
The solution is to copy the collection before going over it.
| range |
range := (2 to: 20) asOrderedCollection.
range copy do: [ :aNumber | aNumber isPrime
ifFalse: [ range remove: aNumber ] ].
range
>>> an OrderedCollection(2 3 5 7 11 13 17 19)
Redefining = but not hash
A difficult error to spot is when you redefine = but
not hash. The symptoms are that you will lose elements that you put in sets or
other strange behaviour. One solution proposed by Kent Beck is to use bitXor:
to redefine hash. Suppose that we want two books to be considered equal if
their titles and authors are the same. Then we would redefine not only = but
also hash as follows:
Book >> = aBook
self class = aBook class ifFalse: [ ^ false ].
^ title = aBook title and: [ authors = aBook authors ]
Book >> hash
^ title hash bitXor: authors hash
Another nasty problem arises if you use a mutable object, i.e., an object that
can change its hash value over time, as an element of a Set or as a key to a
Dictionary. Don’t do this unless you love debugging!
11.7
Chapter summary
The collection hierarchy provides a common vocabulary for uniformly manipulating a variety of different kinds of collections.
• A key distinction is between SequenceableCollections, which maintain their elements in a given order, Dictionary and its subclasses,
which maintain key-to-value associations, and Sets and Bags, which are
unordered.
214
11.7
Chapter summary
• You can convert most collections to another kind of collection by sending them the messages asArray, asOrderedCollection etc..
• To sort a collection, send it the message asSortedCollection.
• #( ... ) creates arrays containing only literal objects (i.e., objects
created without sending messages). { ... } creates dynamic arrays
using a compact form.
• A Dictionary compares keys by equality. It is most useful when keys
are instances of String. An IdentityDictionary instead uses object
identity to compare keys. It is more suitable when Symbols are used as
keys, or when mapping object references to values.
• Strings also understand the usual collection messages. In addition, a
String supports a simple form of pattern-matching. For more advanced
application, look instead at the RegEx package.
• The basic iteration message is do:. It is useful for imperative code, such
as modifying each element of a collection, or sending each element a
message.
• Instead of using do:, it is more common to use collect:, select:,
reject:, includes:, inject:into: and other higher-level messages to
process collections in a uniform way.
• Never remove an element from a collection you are iterating over. If you
must modify it, iterate over a copy instead.
• If you override =, remember to override hash as well!
215
CHAPTER
12
Streams
Streams are used to iterate over sequences of elements such as sequenced
collections, files, and network streams. Streams may be either readable,
or writeable, or both. Reading or writing is always relative to the current
position in the stream. Streams can easily be converted to collections, and vice
versa.
12.1
Two sequences of elements
A good metaphor to understand a stream is the following. A stream can be
represented as two sequences of elements: a past element sequence and a
future element sequence. The stream is positioned between the two sequences.
Understanding this model is important, since all stream operations in Pharo
rely on it. For this reason, most of the Stream classes are subclasses of PositionableStream. Figure 12.1 presents a stream which contains five characters.
This stream is in its original position, i.e., there is no element in the past. You
can go back to this position using the message reset defined in PositionableStream.
Reading an element conceptually means removing the first element of the
future element sequence and putting it after the last element in the past
Figure 12.1: A stream positioned at its beginning.
217
Streams
Figure 12.2: The same stream after the execution of the method next: the
character a is in the past whereas b, c, d and e are in the future.
Figure 12.3: The same stream after having written an x.
element sequence. After having read one element using the message next, the
state of your stream is that shown in Figure 12.2.
Writing an element means replacing the first element of the future sequence
by the new one and moving it to the past. Figure 12.3 shows the state of the
same stream after having written an x using the message nextPut: anElement defined in Stream.
12.2
Streams vs. collections
The collection protocol supports the storage, removal and enumeration
of the elements of a collection, but does not allow these operations to be
intermingled. For example, if the elements of an OrderedCollection are
processed by a do: method, it is not possible to add or remove elements from
inside the do: block. Nor does the collection protocol offer ways to iterate
over two collections at the same time, choosing which collection goes forward
and which does not. Procedures like these require that a traversal index or
position reference is maintained outside of the collection itself: this is exactly
the role of ReadStream, WriteStream and ReadWriteStream.
These three classes are defined to stream over some collection. For example, the following snippet creates a stream on an interval, then it reads two
elements.
| r |
r := ReadStream on: (1 to: 1000).
r next.
>>> 1
r next.
>>> 2
r atEnd.
>>> false
218
12.3 Streaming over collections
WriteStreams can write data to the collection:
| w |
w := WriteStream on: (String new: 5).
w nextPut: $a.
w nextPut: $b.
w contents.
>>> 'ab'
It is also possible to create ReadWriteStreams that support both the reading
and writing protocols.
Streams are not only meant for collections, they can be used for files or sockets too. The following example creates a file named test.txt, writes two
strings to it, separated by a carriage return, and closes the file.
StandardFileStream
fileNamed: 'test.txt'
do: [:str | str
nextPutAll: '123';
cr;
nextPutAll: 'abcd' ].
The following sections present the protocols in more depth.
12.3
Streaming over collections
Streams are really useful when dealing with collections of elements, and can
be used for reading and writing those elements. We will now explore the
stream features for collections.
Reading collections
Using a stream to read a collection essentially provides you a pointer into the
collection. That pointer will move forward on reading, and you can place it
wherever you want. The class ReadStream should be used to read elements
from collections.
Messages next and next: defined in ReadStream are used to retrieve one or
more elements from the collection.
| stream |
stream := ReadStream on: #(1 (a b c) false).
stream next.
>>> 1
stream next.
>>> #(#a #b #c)
stream next.
>>> false
219
Streams
| stream |
stream := ReadStream on: 'abcdef'.
stream next: 0.
>>> ''
stream next: 1.
>>> 'a'
stream next: 3.
>>> 'bcd'
stream next: 2.
>>> 'ef'
The message peek defined in PositionableStream is used when you want to
know what is the next element in the stream without going forward.
stream := ReadStream on: '-143'.
"look at the first element without consuming it."
negative := (stream peek = $-).
negative.
>>> true
"ignores the minus character"
negative ifTrue: [ stream next ].
number := stream upToEnd.
number.
>>> '143'
This code sets the boolean variable negative according to the sign of the
number in the stream, and number to its absolute value. The message upToEnd
defined in ReadStream returns everything from the current position to the
end of the stream and sets the stream to its end. This code can be simplified
using the message peekFor: defined in PositionableStream, which moves
forward if the following element equals the parameter and doesn’t move
otherwise.
| stream |
stream := '-143' readStream.
(stream peekFor: $-)
>>> true
stream upToEnd
>>> '143'
peekFor: also returns a boolean indicating if the parameter equals the element.
You might have noticed a new way of constructing a stream in the above
example: one can simply send the message readStream to a sequenceable collection (such as a String) to get a reading stream on that particular collection.
Positioning
There are messages to position the stream pointer. If you have the index, you
can go directly to it using position: defined in PositionableStream. You
220
12.3 Streaming over collections
Figure 12.4: A stream at position 2.
can request the current position using position. Please remember that a
stream is not positioned on an element, but between two elements. The index
corresponding to the beginning of the stream is 0.
You can obtain the state of the stream depicted in 12.4 with the following
code:
| stream |
stream := 'abcde' readStream.
stream position: 2.
stream peek
>>> $c
To position the stream at the beginning or the end, you can use the message
reset or setToEnd. The messages skip: and skipTo: are used to go forward
to a location relative to the current position: skip: accepts a number as
argument and skips that number of elements whereas skipTo: skips all
elements in the stream until it finds an element equal to its parameter. Note
that it positions the stream after the matched element.
| stream |
stream := 'abcdef' readStream.
stream next.
>>> $a "stream is now positioned just after the a"
stream skip: 3. "stream is now after the d"
stream position.
>>> 4
stream skip: -2. "stream is after the b"
stream position.
>>> 2
stream reset.
stream position.
>>> 0
stream skipTo: $e. "stream is just after the e now"
stream next.
>>> $f
stream contents.
>>> 'abcdef'
As you can see, the letter e has been skipped.
The message contents always returns a copy of the entire stream.
221
Streams
Testing
Some messages allow you to test the state of the current stream: atEnd returns true if and only if no more elements can be read, whereas isEmpty
returns true if and only if there are no elements at all in the collection.
Here is a possible implementation of an algorithm using atEnd that takes two
sorted collections as parameters and merges those collections into another
sorted collection:
| stream1 stream2 result |
stream1 := #(1 4 9 11 12 13) readStream.
stream2 := #(1 2 3 4 5 10 13 14 15) readStream.
"The variable result will contain the sorted collection."
result := OrderedCollection new.
[stream1 atEnd not & stream2 atEnd not ]
whileTrue: [
stream1 peek < stream2 peek
"Remove the smallest element from either stream and add it to
the result."
ifTrue: [result add: stream1 next ]
ifFalse: [result add: stream2 next ] ].
"One of the two streams might not be at its end. Copy whatever remains."
result
addAll: stream1 upToEnd;
addAll: stream2 upToEnd.
result.
>>> an OrderedCollection(1 1 2 3 4 4 5 9 10 11 12 13 13 14 15)
Writing to collections
We have already seen how to read a collection by iterating over its elements using a ReadStream. We’ll now learn how to create collections using
WriteStreams.
WriteStreams are useful for appending a lot of data to a collection at various
locations. They are often used to construct strings that are based on static and
dynamic parts, as in this example:
| stream |
stream := String new writeStream.
stream
nextPutAll: 'This Smalltalk image contains: ';
print: Smalltalk allClasses size;
nextPutAll: ' classes.';
cr;
nextPutAll: 'This is really a lot.'.
222
12.3 Streaming over collections
stream contents.
>>> 'This Smalltalk image contains: 2322 classes.
This is really a lot.'
This technique is used in the different implementations of the method
printOn:, for example. There is a simpler and more efficient way of creating
strings if you are only interested in the content of the stream:
| string |
string := String streamContents:
[ :stream |
stream
print: #(1 2 3);
space;
nextPutAll: 'size';
space;
nextPut: $=;
space;
print: 3. ].
string.
>>> '#(1 2 3) size = 3'
The message streamContents: defined SequenceableCollection creates
a collection and a stream on that collection for you. It then executes the
block you gave passing the stream as a parameter. When the block ends,
streamContents: returns the contents of the collection.
The following WriteStream methods are especially useful in this context:
nextPut: adds the parameter to the stream;
nextPutAll: adds each element of the collection, passed as a parameter, to
the stream;
print: adds the textual representation of the parameter to the stream.
There are also convenient messages for printing useful characters to a stream,
such as space, tab and cr (carriage return). Another useful method is ensureASpace which ensures that the last character in the stream is a space; if
the last character isn’t a space it adds one.
About String Concatenation
Using nextPut: and nextPutAll: on a WriteStream is often the best way to
concatenate characters. Using the comma concatenation operator (,) is far
less efficient:
[| temp |
temp := String new.
(1 to: 100000)
do: [:i | temp := temp, i asString, ' ' ] ] timeToRun
>>> 115176 "(milliseconds)"
223
Streams
Figure 12.5: A new history is empty. Nothing is displayed in the web browser.
Figure 12.6: The user opens to page 1.
[| temp |
temp := WriteStream on: String new.
(1 to: 100000)
do: [:i | temp nextPutAll: i asString; space ].
temp contents ] timeToRun
>>> 1262 "(milliseconds)"
The reason that using a stream can be much more efficient is that using a
comma creates a new string containing the concatenation of the receiver and
the argument, so it must copy both of them. When you repeatedly concatenate onto the same receiver, it gets longer and longer each time, so that the
number of characters that must be copied goes up exponentially. This also
creates a lot of garbage, which must be collected. Using a stream instead of
string concatenation is a well-known optimization.
In fact, you can use the message streamContents: defined in SequenceableCollection class (mentioned earlier) to help you do this:
String streamContents: [ :tempStream |
(1 to: 100000)
do: [:i | tempStream nextPutAll: i asString; space ] ]
Reading and writing at the same time
It’s possible to use a stream to access a collection for reading and writing at
the same time. Imagine you want to create a History class which will manage
backward and forward buttons in a web browser. A history would react as in
figures 12.5 to 12.11.
This behaviour can be implemented using a ReadWriteStream.
Object subclass: #History
instanceVariableNames: 'stream'
224
Figure 12.7: The user clicks on a link to page 2.
Figure 12.8: The user clicks on a link to page 3.
Figure 12.9: The user clicks on the Back button. They are now viewing page 2 again.
Figure 12.10: The user clicks again the back button. Page 1 is now displayed.
Figure 12.11: From page 1, the user clicks on a link to page 4. The history forgets
pages 2 and 3.
Streams
classVariableNames: ''
package: 'PBE-Streams'
History >> initialize
super initialize.
stream := ReadWriteStream on: Array new.
Nothing really difficult here, we define a new class which contains a stream.
The stream is created during the initialize method.
We need methods to go backward and forward:
History >> goBackward
self canGoBackward
ifFalse: [ self error: 'Already on the first element' ].
stream skip: -2.
^ stream next.
History >> goForward
self canGoForward
ifFalse: [ self error: 'Already on the last element' ].
^ stream next
Up to this point, the code is pretty straightforward. Next, we have to deal with
the goTo: method which should be activated when the user clicks on a link. A
possible implementation is:
History >> goTo: aPage
stream nextPut: aPage.
This version is incomplete however. This is because when the user clicks on
the link, there should be no more future pages to go to, i.e., the forward button
must be deactivated. To do this, the simplest solution is to write nil just after,
to indicate that history is at the end:
History >> goTo: anObject
stream nextPut: anObject.
stream nextPut: nil.
stream back.
Now, only methods canGoBackward and canGoForward remain to be implemented.
A stream is always positioned between two elements. To go backward, there
must be two pages before the current position: one page is the current page,
and the other one is the page we want to go to.
History >> canGoBackward
^ stream position > 1
History >> canGoForward
^ stream atEnd not and: [stream peek notNil ]
226
12.4
Using streams for file access
Let us add a method to peek at the contents of the stream:
History >> contents
^ stream contents
And the history works as advertised:
History new
goTo: #page1;
goTo: #page2;
goTo: #page3;
goBackward;
goBackward;
goTo: #page4;
contents
>>> #(#page1 #page4 nil nil)
12.4
Using streams for file access
You have already seen how to stream over collections of elements. It’s also
possible to stream over files on your hard disk. Once created, a stream on
a file is really like a stream on a collection; you will be able to use the same
protocol to read, write or position the stream. The main difference appears
in the creation of the stream. There are several different ways to create file
streams, as we shall now see.
Creating file streams
To create file streams, you will have to use one of the following instance
creation messages offered by the class FileStream:
fileNamed: Open a file with the given name for reading and writing. If the
file already exists, its prior contents may be modified or replaced, but
the file will not be truncated on close. If the name has no directory part,
then the file will be created in the default directory.
newFileNamed: Create a new file with the given name, and answer a stream
opened for writing on that file. If the file already exists, ask the user
what to do.
forceNewFileNamed: Create a new file with the given name, and answer a
stream opened for writing on that file. If the file already exists, delete it
without asking before creating the new file.
oldFileNamed: Open an existing file with the given name for reading and
writing. If the file already exists, its prior contents may be modified or
replaced, but the file will not be truncated on close. If the name has no
directory part, then the file will be created in the default directory.
readOnlyFileNamed: Open an existing file with the given name for reading.
227
Streams
You have to remember that each time you open a stream on a file, you have to
close it too. This is done through the =close message defined in FileStream.
| stream |
stream := FileStream forceNewFileNamed: 'test.txt'.
stream
nextPutAll: 'This text is written in a file named ';
print: stream localName.
stream close.
stream := FileStream readOnlyFileNamed: 'test.txt'.
stream contents.
>>> 'This text is written in a file named ''test.txt'''
stream close.
The message localName defined in class FileStream answers the last component of the name of the file. You can also access the full path name using the
message fullName.
You will soon notice that manually closing the file stream is painful and
error-prone. That’s why FileStream offers a message called forceNewFileNamed:do: to automatically close a new stream after evaluating a block that
sets its contents.
| string |
FileStream
forceNewFileNamed: 'test.txt'
do: [ :stream |
stream
nextPutAll: 'This text is written in a file named ';
print: stream localName ].
string := FileStream
readOnlyFileNamed: 'test.txt'
do: [ :stream | stream contents ].
string
>>> 'This text is written in a file named ''test.txt'''
The stream creation methods that take a block as an argument first create a
stream on a file, then execute the block with the stream as an argument, and
finally close the stream. These methods return what is returned by the block,
which is to say, the value of the last expression in the block. This is used in
the previous example to get the content of the file and put it in the variable
string.
Binary streams
By default, created streams are text-based which means you will read and
write characters. If your stream must be binary, you have to send the message
binary to your stream.
228
12.5 Chapter summary
When your stream is in binary mode, you can only write numbers from 0 to
255 (1 Byte). If you want to use nextPutAll: to write more than one number
at a time, you have to pass a ByteArray as argument.
FileStream
forceNewFileNamed: 'test.bin'
do: [ :stream |
stream
binary;
nextPutAll: #(145 250 139 98) asByteArray ].
FileStream
readOnlyFileNamed: 'test.bin'
do: [ :stream |
stream binary.
stream size.
>>> 4
stream next.
>>> 145
stream upToEnd.
>>> #[250 139 98 ]
].
Here is another example which creates a picture in a file named test.pgm
(portable graymap file format). You can open this file with your favorite
drawing program.
FileStream
forceNewFileNamed: 'test.pgm'
do: [ :stream |
stream
nextPutAll: 'P5'; cr;
nextPutAll: '4 4'; cr;
nextPutAll: '255'; cr;
binary;
nextPutAll: #(255 0 255 0)
nextPutAll: #(0 255 0 255)
nextPutAll: #(255 0 255 0)
nextPutAll: #(0 255 0 255)
asByteArray;
asByteArray;
asByteArray;
asByteArray ]
This creates a 4x4 checkerboard as shown in 12.12.
12.5
Chapter summary
Streams offer a better way (compared to collections) to incrementally read
and write a sequence of elements. There are easy ways to convert back and
forth between streams and collections.
• Streams may be either readable, writeable or both readable and writeable.
229
Streams
Figure 12.12: A 4x4 checkerboard you can draw using binary streams.
• To convert a collection to a stream, define a stream on a collection, e.g.,
ReadStream on: (1 to: 1000), or send the messages readStream, etc.
to the collection.
• To convert a stream to a collection, send the message contents.
• To concatenate large collections, instead of using the comma operator, it
is more efficient to create a stream, append the collections to the stream
with nextPutAll:, and extract the result by sending contents.
• File streams are by default character-based. Send binary to explicitly
make them binary.
230
CHAPTER
13
Morphic
Morphic is the name given to Pharo’s graphical interface. Morphic is written
in Pharo, so it is fully portable between operating systems. As a consequence,
Pharo looks exactly the same on Unix, MacOS and Windows. What distinguishes Morphic from most other user interface toolkits is that it does not
have separate modes for composing and running the interface: all the graphical
elements can be assembled and disassembled by the user, at any time. (We
thank Hilaire Fernandes for permission to base this chapter on his original
article in French.)
13.1
The history of Morphic
Morphic was developed by John Maloney and Randy Smith for the Self programming language, starting around 1993. Maloney later wrote a new version
of Morphic for Squeak, but the basic ideas behind the Self version are still
alive and well in Pharo Morphic: directness and liveness. Directness means
that the shapes on the screen are objects that can be examined or changed
directly, that is, by clicking on them using a mouse. Liveness means that
the user interface is always able to respond to user actions: information on
the screen is continuously updated as the world that it describes changes. A
simple example of this is that you can detach a menu item and keep it as a
button.
Bring up the World Menu and meta-click once on it to bring up its morphic
halo, then meta-click again on a menu item you want to detach, to bring up
that item’s halo. (Recall that you should set halosEnabled in the Preferences
browser.) Now drag that item elsewhere on the screen by grabbing the black
handle (see Figure 13.1), as shown in Figure 13.2.
231
Morphic
Figure 13.1: The grab handle.
Figure 13.2: Detaching a morph, here the Playground menu item, to make it an
independent button.
All of the objects that you see on the screen when you run Pharo are Morphs,
that is, they are instances of subclasses of class Morph. Morph itself is a large
class with many methods; this makes it possible for subclasses to implement
interesting behaviour with little code. You can create a morph to represent
any object, although how good a representation you get depends on the
object!
To create a morph to represent a string object, execute the following code in a
Playground.
'Morph' asMorph openInWorld
This creates a Morph to represent the string 'Morph', and then opens it (that
is, displays it) in the world, which is the name that Pharo gives to the screen.
You should obtain a graphical element (a Morph), which you can manipulate by
meta-clicking.
Of course, it is possible to define morphs that are more interesting graphical
representations than the one that you have just seen. The method asMorph
has a default implementation in class Object class that just creates a StringMorph. So, for example, Color tan asMorph returns a StringMorph labeled
with the result of Color tan printString. Let’s change this so that we get a
coloured rectangle instead.
Open a browser on the Color class and add the following method to it:
Color >> asMorph
^ Morph new color: self
Now execute Color orange asMorph openInWorld in a Playground. Instead
of the string-like morph, you get an orange rectangle (see Figure 13.3)!
232
13.2 Manipulating morphs
Figure 13.3: Color orange asMorph openInWorld with our new method.
13.2
Manipulating morphs
Morphs are objects, so we can manipulate them like any other object in Pharo:
by sending messages, we can change their properties, create new subclasses of
Morph, and so on.
Every morph, even if it is not currently open on the screen, has a position and
a size. For convenience, all morphs are considered to occupy a rectangular
region of the screen; if they are irregularly shaped, their position and size are
those of the smallest rectangular box that surrounds them, which is known as
the morph’s bounding box, or just its bounds. The position method returns
a Point that describes the location of the morph’s upper left corner (or the
upper left corner of its bounding box). The origin of the coordinate system is
the screen’s upper left corner, with y coordinates increasing down the screen
and x coordinates increasing to the right. The extent method also returns a
point, but this point specifies the width and height of the morph rather than a
location.
Type the following code into a playground and Do it:
joe := Morph new color: Color blue.
joe openInWorld.
bill := Morph new color: Color red.
bill openInWorld.
Then type joe position and then Print it. To move joe, execute joe position: (joe position + (10@3)) repeatedly (see Figure 13.4).
It is possible to do a similar thing with size. joe extent answers joe’s size;
to have joe grow, execute joe extent: (joe extent * 1.1). To change the
color of a morph, send it the color: message with the desired Color object as
233
Morphic
Figure 13.4: Bill and Joe after 10 moves.
argument, for instance, joe color: Color orange. To add transparency, try
joe color: (Color orange alpha: 0.5).
To make bill follow joe, you can repeatedly execute this code:
bill position: (joe position + (100@0))
If you move joe using the mouse and then execute this code, bill will move so
that it is 100 pixels to the right of joe.
You can see the result on Figure 13.5
13.3
Composing morphs
One way of creating new graphical representations is by placing one morph
inside another. This is called composition; morphs can be composed to any
depth.
You can place a morph inside another by sending the message addMorph: to
the container morph.
Try adding a morph to another one:
balloon := BalloonMorph new color: Color yellow.
joe addMorph: balloon.
234
13.3
Composing morphs
Figure 13.5: Bill follows Joe.
Figure 13.6: The balloon is contained inside joe, the translucent orange morph.
balloon position: joe position.
The last line positions the balloon at the same coordinates as joe. Notice that
the coordinates of the contained morph are still relative to the screen, not to
the containing morph. There are many methods available to position a morph;
browse the geometry protocol of class Morph to see for yourself. For example,
to center the balloon inside joe, execute balloon center: joe center.
If you now try to grab the balloon with the mouse, you will find that you
actually grab joe, and the two morphs move together: the balloon is embedded
inside joe. It is possible to embed more morphs inside joe. In addition to doing
this programmatically, you can also embed morphs by direct manipulation.
235
Morphic
Figure 13.7: A CrossMorph with its halo; you can resize it as you wish.
13.4
Creating and drawing your own morphs
While it is possible to make many interesting and useful graphical representations by composing morphs, sometimes you will need to create something
completely different.
To do this you define a subclass of Morph and override the drawOn: method to
change its appearance.
The morphic framework sends the message drawOn: to a morph when it needs
to redisplay the morph on the screen. The parameter to drawOn: is a kind
of Canvas; the expected behaviour is that the morph will draw itself on that
canvas, inside its bounds. Let’s use this knowledge to create a cross-shaped
morph.
Using the browser, define a new class CrossMorph inheriting from Morph:
Morph subclass: #CrossMorph
instanceVariableNames: ''
classVariableNames: ''
package: 'PBE-Morphic'
We can define the drawOn: method like this:
CrossMorph >> drawOn: aCanvas
| crossHeight crossWidth horizontalBar verticalBar |
crossHeight := self height / 3.0.
crossWidth := self width / 3.0.
horizontalBar := self bounds insetBy: 0 @ crossHeight.
verticalBar := self bounds insetBy: crossWidth @ 0.
aCanvas fillRectangle: horizontalBar color: self color.
aCanvas fillRectangle: verticalBar color: self color
Sending the bounds message to a morph answers its bounding box, which is
an instance of Rectangle. Rectangles understand many messages that create
other rectangles of related geometry. Here, we use the insetBy: message
with a point as its argument to create first a rectangle with reduced height,
and then another rectangle with reduced width.
To test your new morph, execute CrossMorph new openInWorld.
236
13.4
Creating and drawing your own morphs
The result should look something like Figure 13.7. However, you will notice
that the sensitive zone — where you can click to grab the morph — is still the
whole bounding box. Let’s fix this.
When the Morphic framework needs to find out which Morphs lie under
the cursor, it sends the message containsPoint: to all the morphs whose
bounding boxes lie under the mouse pointer. So, to limit the sensitive zone
of the morph to the cross shape, we need to override the containsPoint:
method.
Define the following method in class CrossMorph:
CrossMorph >> containsPoint: aPoint
| crossHeight crossWidth horizontalBar verticalBar |
crossHeight := self height / 3.0.
crossWidth := self width / 3.0.
horizontalBar := self bounds insetBy: 0 @ crossHeight.
verticalBar := self bounds insetBy: crossWidth @ 0.
^ (horizontalBar containsPoint: aPoint) or: [ verticalBar
containsPoint: aPoint ]
This method uses the same logic as drawOn:, so we can be confident that the
points for which containsPoint: answers true are the same ones that will be
colored in by drawOn. Notice how we leverage the containsPoint: method in
class Rectangle to do the hard work.
There are two problems with the code in the two methods above.
The most obvious is that we have duplicated code. This is a cardinal error: if
we find that we need to change the way that horizontalBar or verticalBar
are calculated, we are quite likely to forget to change one of the two occurrences. The solution is to factor out these calculations into two new methods,
which we put in the private protocol:
CrossMorph >> horizontalBar
| crossHeight |
crossHeight := self height / 3.0.
^ self bounds insetBy: 0 @ crossHeight
CrossMorph >> verticalBar
| crossWidth |
crossWidth := self width / 3.0.
^ self bounds insetBy: crossWidth @ 0
We can then define both drawOn: and containsPoint: using these methods:
CrossMorph >> drawOn: aCanvas
aCanvas fillRectangle: self horizontalBar color: self color.
aCanvas fillRectangle: self verticalBar color: self color
CrossMorph >> containsPoint: aPoint
^ (self horizontalBar containsPoint: aPoint) or: [ self verticalBar
containsPoint: aPoint ]
237
Morphic
Figure 13.8: The center of the cross is filled twice with the color.
Figure 13.9: The cross-shaped morph, showing a row of unfilled pixels.
This code is much simpler to understand, largely because we have given
meaningful names to the private methods. In fact, it is so simple that you may
have noticed the second problem: the area in the center of the cross, which is
under both the horizontal and the vertical bars, is drawn twice. This doesn’t
matter when we fill the cross with an opaque colour, but the bug becomes
apparent immediately if we draw a semi-transparent cross, as shown in Figure
13.8.
Execute the following code in a playground, line by line:
CrossMorph new openInWorld;
bounds: (0@0 corner: 200@200);
color: (Color blue alpha: 0.4)
The fix is to divide the vertical bar into three pieces, and to fill only the top
238
13.5
Interaction and animation
and bottom. Once again we find a method in class Rectangle that does the
hard work for us: r1 areasOutside: r2 answers an array of rectangles comprising the parts of r1 outside r2. Here is the revised code:
CrossMorph >> drawOn: aCanvas
| topAndBottom |
aCanvas fillRectangle: self horizontalBar color: self color.
topAndBottom := self verticalBar areasOutside: self horizontalBar.
topAndBottom do: [ :each | aCanvas fillRectangle: each color: self
color ]
This code seems to work, but if you try it on some crosses and resize them, you
may notice that at some sizes, a one-pixel wide line separates the bottom of
the cross from the remainder, as shown in Figure 13.9. This is due to rounding:
when the size of the rectangle to be filled is not an integer, fillRectangle:
color: seems to round inconsistently, leaving one row of pixels unfilled. We
can work around this by rounding explicitly when we calculate the sizes of the
bars.
CrossMorph >> horizontalBar
| crossHeight |
crossHeight := (self height / 3.0) rounded.
^ self bounds insetBy: 0 @ crossHeight
CrossMorph >> verticalBar
| crossWidth |
crossWidth := (self width / 3.0) rounded.
^ self bounds insetBy: crossWidth @ 0
13.5
Interaction and animation
To build live user interfaces using morphs, we need to be able to interact with
them using the mouse and keyboard. Moreover, the morphs need to be able
respond to user input by changing their appearance and position — that is, by
animating themselves.
Mouse events
When a mouse button is pressed, Morphic sends each morph under the mouse
pointer the message handlesMouseDown:. If a morph answers true, then
Morphic immediately sends it the mouseDown: message; it also sends the
mouseUp: message when the user releases the mouse button. If all morphs
answer false, then Morphic initiates a drag-and-drop operation. As we
will discuss below, the mouseDown: and mouseUp: messages are sent with an
argument — a MouseEvent object — that encodes the details of the mouse
action.
Let’s extend CrossMorph to handle mouse events. We start by ensuring that all
crossMorphs answer true to the handlesMouseDown: message.
239
Morphic
Figure 13.10: Move Handle button.
Figure 13.11: Grab Handle button.
Add this method to CrossMorph:
CrossMorph >> handlesMouseDown: anEvent
^ true
Suppose that when we click on the cross, we want to change the color of the
cross to red, and when we action-click on it, we want to change the color to
yellow. This can be accomplished by the mouseDown: method as follows:
CrossMorph >> mouseDown: anEvent
anEvent redButtonPressed
ifTrue: [ self color: Color red ]. "click"
anEvent yellowButtonPressed
ifTrue: [ self color: Color yellow ]. "action-click"
self changed
Notice that in addition to changing the color of the morph, this method also
sends self changed. This makes sure that morphic sends drawOn: in a timely
fashion.
Note also that once the morph handles mouse events, you can no longer grab
it with the mouse and move it. Instead you have to use the halo: meta-click on
the morph to make the halo appear and grab either the brown move handle
(see Figure 13.10) or the black pickup handle (see Figure 13.11) at the top of
the morph.
The anEvent argument of mouseDown: is an instance of MouseEvent, which is
a subclass of MorphicEvent. MouseEvent defines the redButtonPressed and
yellowButtonPressed methods. Browse this class to see what other methods
it provides to interrogate the mouse event.
Keyboard events
To catch keyboard events, we need to take three steps.
1. Give the keyboard focus to a specific morph. For instance, we can give
focus to our morph when the mouse is over it.
2. Handle the keyboard event itself with the handleKeystroke: method.
This message is sent to the morph that has keyboard focus when the
user presses a key.
240
13.5
Interaction and animation
3. Release the keyboard focus when the mouse is no longer over our
morph.
Let’s extend CrossMorph so that it reacts to keystrokes. First, we need to
arrange to be notified when the mouse is over the morph. This will happen if
our morph answers true to the handlesMouseOver: message
Declare that CrossMorph will react when it is under the mouse pointer.
CrossMorph >> handlesMouseOver: anEvent
^true
This message is the equivalent of handlesMouseDown: for the mouse position.
When the mouse pointer enters or leaves the morph, the mouseEnter: and
mouseLeave: messages are sent to it.
Define two methods so that CrossMorph catches and releases the keyboard
focus, and a third method to actually handle the keystrokes.
CrossMorph >> mouseEnter: anEvent
anEvent hand newKeyboardFocus: self
CrossMorph >> mouseLeave: anEvent
anEvent hand newKeyboardFocus: nil
CrossMorph >> handleKeystroke: anEvent
| keyValue |
keyValue := anEvent keyValue.
keyValue = 30 "up arrow"
ifTrue: [self position: self position
keyValue = 31 "down arrow"
ifTrue: [self position: self position
keyValue = 29 "right arrow"
ifTrue: [self position: self position
keyValue = 28 "left arrow"
ifTrue: [self position: self position
- (0 @ 1)].
+ (0 @ 1)].
+ (1 @ 0)].
- (1 @ 0)]
We have written this method so that you can move the morph using the
arrow keys. Note that when the mouse is no longer over the morph, the
handleKeystroke: message is not sent, so the morph stops responding to keyboard commands. To discover the key values, you can open a Transcript window and add Transcript show: anEvent keyValue to the handleKeystroke:
method.
The anEvent argument of handleKeystroke: is an instance of KeyboardEvent,
another subclass of MorphicEvent. Browse this class to learn more about
keyboard events.
Morphic animations
Morphic provides a simple animation system with two main methods: step is
sent to a morph at regular intervals of time, while stepTime specifies the time
241
Morphic
Figure 13.12: The debug handle button.
in milliseconds between steps. stepTime is actually the minimum time between steps. If you ask for a stepTime of 1 ms, don’t be surprised if Pharo is
too busy to step your morph that often. In addition, startStepping turns on
the stepping mechanism, while stopStepping turns it off again. isStepping
can be used to find out whether a morph is currently being stepped.
Make CrossMorph blink by defining these methods as follows:
CrossMorph >> stepTime
^ 100
CrossMorph >> step
(self color diff: Color black) < 0.1
ifTrue: [ self color: Color red ]
ifFalse: [ self color: self color darker ]
To start things off, you can open an inspector on a CrossMorph using the debug handle (see Figure 13.12) in the morphic halo, type self startStepping
in the small playground pane at the bottom, and Do it.
Alternatively, you can modify the handleKeystroke: method so that you can
use the + and - keys to start and stop stepping. Add the following code to the
handleKeystroke: method:
keyValue =
ifTrue:
keyValue =
ifTrue:
13.6
$+ asciiValue
[ self startStepping ].
$- asciiValue
[ self stopStepping ].
Interactors
To prompt the user for input, the UIManager class provides a large number of
ready to use dialog boxes. For instance, the request:initialAnswer: method
returns the string entered by the user (Figure 13.13).
UIManager default request: 'What''s your name?' initialAnswer: 'no name'
To display a popup menu, use one of the various chooseFrom: methods (Figure 13.14):
UIManager default
chooseFrom: #('circle' 'oval' 'square' 'rectangle' 'triangle')
lines: #(2 4) message: 'Choose a shape'
Browse the UIManager class and try out some of the interaction methods
offered.
242
13.7
Drag-and-drop
Figure 13.13: An input dialog.
Figure 13.14: Pop-up menu.
13.7
Drag-and-drop
Morphic also supports drag-and-drop. Let’s examine a simple example with
two morphs, a receiver morph and a dropped morph. The receiver will accept
a morph only if the dropped morph matches a given condition: in our example, the morph should be blue. If it is rejected, the dropped morph decides
what to do.
Let’s first define the receiver morph:
Morph subclass: #ReceiverMorph
instanceVariableNames: ''
classVariableNames: ''
package: 'PBE-Morphic'
Now define the initialization method in the usual way:
ReceiverMorph >> initialize
super initialize.
color := Color red.
bounds := 0 @ 0 extent: 200 @ 200
How do we decide if the receiver morph will accept or reject the dropped
morph? In general, both of the morphs will have to agree to the interaction.
243
Morphic
Figure 13.15: A ReceiverMorph and an EllipseMorph.
The receiver does this by responding to wantsDroppedMorph:event:. Its first
argument is the dropped morph, and the second the mouse event, so that the
receiver can, for example, see if any modifier keys were held down at the time
of the drop. The dropped morph is also given the opportunity to check and
see if it likes the morph onto which it is being dropped, by responding to the
message wantsToBeDroppedInto:. The default implementation of this method
(in class Morph) answers true.
ReceiverMorph >> wantsDroppedMorph: aMorph event: anEvent
^ aMorph color = Color blue
What happens to the dropped morph if the receiving morph doesn’t want
it? The default behaviour is for it to do nothing, that is, to sit on top of the
receiving morph, but without interacting with it. A more intuitive behavior is
for the dropped morph to go back to its original position. This can be achieved
by the receiver answering true to the message repelsMorph:event: when it
doesn’t want the dropped morph:
ReceiverMorph >> repelsMorph: aMorph event: anEvent
^ (self wantsDroppedMorph: aMorph event: anEvent) not
That’s all we need as far as the receiver is concerned.
Create instances of ReceiverMorph and EllipseMorph in a playground:
ReceiverMorph new openInWorld;
bounds: (100@100 corner: 200@200).
EllipseMorph new openInWorld.
Try to drag and drop the yellow EllipseMorph onto the receiver. It will be
rejected and sent back to its initial position.
To change this behaviour, change the color of the ellipse morph to the color
blue (by sending it the message color: Color blue; right after new). Blue
morphs should be accepted by the ReceiverMorph.
244
13.8
A complete example
Let’s create a specific subclass of Morph, named DroppedMorph, so we can
experiment a bit more:
Morph subclass: #DroppedMorph
instanceVariableNames: ''
classVariableNames: ''
package: 'PBE-Morphic'
DroppedMorph >> initialize
super initialize.
color := Color blue.
self position: 250 @ 100
Now we can specify what the dropped morph should do when it is rejected by
the receiver; here it will stay attached to the mouse pointer:
DroppedMorph >> rejectDropMorphEvent: anEvent
| h |
h := anEvent hand.
WorldState addDeferredUIMessage: [ h grabMorph: self ].
anEvent wasHandled: true
Sending the hand message to an event answers the hand, an instance of HandMorph that represents the mouse pointer and whatever it holds. Here we tell
the World that the hand should grab self, the rejected morph.
Create two instances of DroppedMorph, and then drag and drop them onto the
receiver.
ReceiverMorph new openInWorld.
morph := (DroppedMorph new color: Color blue) openInWorld.
morph position: (morph position + (70@0)).
(DroppedMorph new color: Color green) openInWorld.
The green morph is rejected and therefore stays attached to the mouse
pointer.
13.8
A complete example
Let’s design a morph to roll a die. Clicking on it will display the values of all
sides of the die in a quick loop, and another click will stop the animation.
Define the die as a subclass of BorderedMorph instead of Morph, because we
will make use of the border.
BorderedMorph subclass: #DieMorph
instanceVariableNames: 'faces dieValue isStopped'
classVariableNames: ''
package: 'PBE-Morphic'
The instance variable faces records the number of faces on the die; we allow dice with up to 9 faces! dieValue records the value of the face that is
245
Morphic
Figure 13.16: Creation of DroppedMorph and ReceiverMorph.
Figure 13.17: The die in Morphic
currently displayed, and isStopped is true if the die animation has stopped
running. To create a die instance, we define the faces: n method on the class
side of DieMorph to create a new die with n faces.
DieMorph class >> faces: aNumber
^ self new faces: aNumber
The initialize method is defined on the instance side in the usual way;
remember that new automatically sends initialize to the newly-created
instance.
DieMorph >> initialize
super initialize.
self extent: 50 @ 50.
self
246
13.8
A complete example
useGradientFill;
borderWidth: 2;
useRoundedCorners.
self setBorderStyle: #complexRaised.
self fillStyle direction: self extent.
self color: Color green.
dieValue := 1.
faces := 6.
isStopped := false
We use a few methods of BorderedMorph to give a nice appearance to the die:
a thick border with a raised effect, rounded corners, and a color gradient on
the visible face. We define the instance method faces: to check for a valid
parameter as follows:
DieMorph >> faces: aNumber
"Set the number of faces"
((aNumber isInteger and: [ aNumber > 0 ]) and: [ aNumber <= 9 ])
ifTrue: [ faces := aNumber ]
It may be good to review the order in which the messages are sent when a die
is created. For instance, if we start by evaluating DieMorph faces: 9:
• The class method DieMorph class >> faces: sends new to DieMorph
class.
• The method for new (inherited by DieMorph class from Behavior)
creates the new instance and sends it the initialize message.
• The initialize method in DieMorph sets faces to an initial value of 6.
• DieMorph class >> new returns to the class method DieMorph class
>> faces:, which then sends the message faces: 9 to the new instance.
• The instance method DieMorph >> faces: now executes, setting the
faces instance variable to 9.
Before defining drawOn:, we need a few methods to place the dots on the
displayed face:
DieMorph >> face1
^ {(0.5 @ 0.5)}
DieMorph >> face2
^{0.25@0.25 . 0.75@0.75}
DieMorph >> face3
^{0.25@0.25 . 0.75@0.75 . 0.5@0.5}
DieMorph >> face4
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75}
247
Morphic
DieMorph >> face5
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75 . 0.5@0.5}
DieMorph >> face6
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75 . 0.25@0.5 .
0.75@0.5}
DieMorph >> face7
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75 . 0.25@0.5 .
0.75@0.5 . 0.5@0.5}
DieMorph >> face8
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75 . 0.25@0.5 .
0.75@0.5 . 0.5@0.5 . 0.5@0.25}
DieMorph >> face9
^{0.25@0.25 . 0.75@0.25 . 0.75@0.75 . 0.25@0.75 . 0.25@0.5 .
0.75@0.5 . 0.5@0.5 . 0.5@0.25 . 0.5@0.75}
These methods define collections of the coordinates of dots for each face. The
coordinates are in a square of size 1x1; we will simply need to scale them to
place the actual dots.
The drawOn: method does two things: it draws the die background with the
super-send, and then draws the dots.
DieMorph >> drawOn: aCanvas
super drawOn: aCanvas.
(self perform: ('face', dieValue asString) asSymbol)
do: [:aPoint | self drawDotOn: aCanvas at: aPoint]
The second part of this method uses the reflective capacities of Pharo. Drawing the dots of a face is a simple matter of iterating over the collection given
by the faceX method for that face, sending the drawDotOn:at: message
for each coordinate. To call the correct faceX method, we use the perform:
method which sends a message built from a string, ('face', dieValue asString) asSymbol. You will encounter this use of perform: quite regularly.
DieMorph >> drawDotOn: aCanvas at: aPoint
aCanvas
fillOval: (Rectangle
center: self position + (self extent * aPoint)
extent: self extent / 6)
color: Color black
Since the coordinates are normalized to the [0:1] interval, we scale them to
the dimensions of our die: self extent * aPoint.
We can already create a die instance from a playground (see result on Figure
13.18):
(DieMorph faces: 6) openInWorld.
248
13.8
A complete example
Figure 13.18: A new die 6 with (DieMorph faces: 6) openInWorld
Figure 13.19: Result of (DieMorph faces: 6) openInWorld; dieValue: 5.
To change the displayed face, we create an accessor that we can use as myDie
dieValue: 5:
DieMorph >> dieValue: aNumber
((aNumber isInteger and: [ aNumber > 0 ]) and: [ aNumber <= faces ])
ifTrue: [
dieValue := aNumber.
self changed ]
Now we will use the animation system to show quickly all the faces:
DieMorph >> stepTime
^ 100
DieMorph >> step
isStopped ifFalse: [self dieValue: (1 to: faces) atRandom]
Now the die is rolling!
To start or stop the animation by clicking, we will use what we learned previously about mouse events. First, activate the reception of mouse events:
DieMorph >> handlesMouseDown: anEvent
^ true
249
Morphic
Figure 13.20: The die displayed with alpha-transparency
DieMorph >> mouseDown: anEvent
anEvent redButtonPressed
ifTrue: [isStopped := isStopped not]
Now the die will roll or stop rolling when we click on it.
13.9
More about the canvas
The drawOn: method has an instance of Canvas as its sole argument; the
canvas is the area on which the morph draws itself. By using the graphics
methods of the canvas you are free to give the appearance you want to a
morph. If you browse the inheritance hierarchy of the Canvas class, you will
see that it has several variants. The default variant of Canvas is FormCanvas,
and you will find the key graphics methods in Canvas and FormCanvas. These
methods can draw points, lines, polygons, rectangles, ellipses, text, and images
with rotation and scaling.
It is also possible to use other kinds of canvas, for example to obtain transparent morphs, more graphics methods, antialiasing, and so on. To use these
features you will need an AlphaBlendingCanvas or a BalloonCanvas. But how
can you obtain such a canvas in a drawOn: method, when drawOn: receives an
instance of FormCanvas as its argument? Fortunately, you can transform one
kind of canvas into another.
To use a canvas with a 0.5 alpha-transparency in DieMorph, redefine drawOn:
like this:
DieMorph >> drawOn: aCanvas
| theCanvas |
theCanvas := aCanvas asAlphaBlendingCanvas: 0.5.
super drawOn: theCanvas.
(self perform: ('face', dieValue asString) asSymbol)
do: [:aPoint | self drawDotOn: theCanvas at: aPoint]
That’s all you need to do!
250
13.10
13.10
Chapter summary
Chapter summary
Morphic is a graphical framework in which graphical interface elements can
be dynamically composed.
• You can convert an object into a morph and display that morph on the
screen by sending it the messages asMorph openInWorld.
• You can manipulate a morph by meta-clicking on it and using the handles that appear. (Handles have help balloons that explain what they
do.)
• You can compose morphs by embedding one onto another, either by
drag and drop or by sending the message addMorph:.
• You can subclass an existing morph class and redefine key methods, like
initialize and drawOn:.
• You can control how a morph reacts to mouse and keyboard events by
redefining the methods handlesMouseDown:, handlesMouseOver:, etc.
• You can animate a morph by defining the methods step (what to do)
and stepTime (the number of milliseconds between steps).
251
CHAPTER
14
Seaside by example
Seaside is a framework for building web applications in Smalltalk, originally
developed by Avi Bryant and Julian Fitzell in 2002. Once mastered, Seaside
makes web applications almost as easy to write as desktop applications. Seaside is really interesting for developing fast complex applications. For example http://allstocker.com is a Seaside commercial applications or the Quuve,
FinWorks and CableExpertise application from http://www.pharo.org/success
have been developed in Seaside.
Seaside is unusual in that it is thoroughly object-oriented: there are no HTML
templates, no complicated control flows through web pages, and no encoding
of state in URLs. Instead, you just send messages to objects. What a nice idea!
14.1
Why do we need Seaside?
Modern web applications try to interact with the user in the same way as desktop applications: they ask the user questions and the user responds, usually by
filling in a form or clicking a button. But the web works the other way around:
the user’s browser makes a request of the server, and the server responds
with a new web page. So web application development frameworks have to
cope with a host of problems, chief among them being the management of
this inverted control flow. Because of this, many web applications try to forbid
the use of the browser’s Back button due to the difficulty of keeping track of
the state of a session. Expressing non-trivial control flows across multiple
web pages is often cumbersome, and multiple control flows can be difficult or
impossible to express.
Seaside is a component-based framework that makes web development easier
in several ways. First, control flow can be expressed naturally using message
253
Seaside by example
sends. Seaside keeps track of which web page corresponds to which point
in the execution of the web application. This means that the browser’s Back
button works correctly.
Second, state is managed for you. As the developer, you have the choice of
enabling backtracking of state, so that navigation Back in time will undo sideeffects. Alternatively, you can use the transaction support built into Seaside to
prevent users from undoing permanent side-effects when they use the back
button. You do not have to encode state information in the URL — this too is
managed automatically for you.
Third, web pages are built up from nested components, each of which can
support its own, independent control flow. There are no HTML templates —
instead valid HTML is generated programmatically using a simple Smalltalkbased protocol. Seaside supports Cascading Style Sheets (CSS), so content and
layout are cleanly separated.
Finally, Seaside provides a convenient web-based development interface, making it easy to develop applications iteratively, debug applications interactively,
and recompile and extend applications while the server is running.
14.2
Getting started
The Seaside community
The Seaside.st website (http://www.seaside.st) contains many Seaside-related
resources, including downloads, documentation and tutorials. (Keep in mind
that Seaside has evolved considerably over the years, and not all available
material refers to the latest version.) There are also several active mailing
lists, which you can find at http://www.seaside.st/community/mailinglist.
In addition, the Wiki on the Seaside GitHub repository (https://github.com/
seasidest/seaside/wiki) provides some crucial documentation, including the
Release Notes for each version of Seaside.
Installing Seaside using the one-click experience image
The easiest way to get started is to download the Seaside One-Click Experience 3.1 from the Pharo Downloads section of the Seaside web site
(http://www.seaside.st/download/pharo). This is a prepackaged version of the
latest stable version of Seaside for Mac OSX, Linux and Windows, built on
Pharo 4. The current latest Seaside version is 3.2. These are full Pharo distributions, which include the Pharo VM as well as a development image with many
Seaside-related packages preloaded, and a helpful Seaside control panel that
starts when you open the image.
Download and launch the Seaside One-Click image (feel free to refer to the
Getting Started section of Chapter : A Quick Tour of Pharo for an in-depth
254
14.2
Getting started
Figure 14.1: Start up the Seaside One-Click Experience image.
discussion of launching Pharo images). Once started, you should see something similar to Figure 14.1 - a familiar Pharo development environment with
some windows pre-opened, such as welcome messages as well as a Seaside
Control Panel.
Starting the Seaside server
The easiest way to start and stop the Seaside web server is through the Seaside Control Panel. If you’re using the Seaside One-Click Experience image,
it should already be open (and running a server adapter) when you open the
image for the first time. You can manually open it at any time, however, by
evaluating WAPharoServerAdapterSpecBrowser open.
Right-clicking on the top pane of the Seaside Control Panel lets you add new
server adapters. For example, you can add a ZnZincServerAdaptor, specify a
port number, then start it using the Start button on the control panel, which
launches a new Seaside server that listens on that port.
ZnZincServerAdaptor startOn: 8080. "start on port 8080"
ZnZincServerAdaptor stop.
You can also start and stop the Seaside web server from a playground, by
sendning the startOn: and stop messages to your server adapter of choice.
255
Seaside by example
The Seaside welcome page
Once the Seaside server is running, navigate to http://localhost:8080/ in your
web browser. You should see a web page that looks like Figure 14.2.
The Welcome page contains links to some sample Seaside applications. It also
links to various documents and resources, and (in the sidebar) links to the
Configuration and Browse applications that allow you to interact with Seaside
applications registered in your image.
Let’s look at one of the example applications that demonstrates the Counter
component: click on the Counter link.
This page is a small Seaside application: it displays a counter component that
can be incremented or decremented by clicking on the ++ and –– links (see
Figure 14.3).
Play with the counter by clicking on these links. Use your browser’s Back
button to go back to a previous state, and then click on ++ again. Notice how
the counter is correctly incremented with respect to the currently displayed
state, rather than the state that the counter was in when you started using the
Back button.
Notice the toolbar at the bottom of the web page in Figure 14.2. Seaside supports a notion of sessions to keep track of the state of the application for
different users. New Session will start a new session on the counter application. Configure allows you to configure the settings of your application
through a web interface. (To close the Configure view, click on the x in the
top right corner.) Halos provides a way to explore the state of the application
running on the Seaside server. Profile and Memory provide detailed information about the run-time performance of the application. XHTML can be used to
validate the generated web page, but works only when the web page is publicly
accessible from the Internet, because it uses the W3C validation service.
Single components
Seaside applications are built up from pluggable components. In fact, components are ordinary Smalltalk objects. The only thing that is special about them
is that they are instances of classes that inherit from the Seaside framework
class WAComponent. We can explore components and their classes from the
Pharo image, or directly from the web interface using halos.
Click on the Halos link at the bottom of the page to toggle on halo functionality. You’ll see a number of nested components each with their own halo icons,
including the Counter component (you may have to scroll down a bit to reach
it), like the one seen in Figure 14.4.
At the top left of the component, the text WACounter tells us the class of the
Seaside component that implements its behavior. Next to this are three
clickable icons. The first is the Class Browser (the notepad with pencil icon),
256
Figure 14.2: The Seaside Welcome application at http://localhost:8080/.
Figure 14.3: The counter.
Seaside by example
Figure 14.4: Halos.
Figure 14.5: Debugging the running Counter application in the Pharo image.
which opens up a (web-based) Seaside class browser on this component’s
class. The second is the Object Inspector (notepad with magnifying glass
icon), which opens a web-based object inspector on the actual WACounter
instance. The third is the CSS Style Manager (the coloured circles icon),
opens a Seaside view to display the CSS style sheet for this component.
At the top right of the component, the Render / Source links let you toggle
between the Rendered and (formatted) Source views of the component’s HTML
source code. Experiment with all of these links. Note that the ++ and – links
are also active in the source view.
The Seaside Class Browser and Object Inspector can be very convenient when
the server is running on another computer, especially when the server does
not have a display, or if it is in remote place. However, when you are first
developing a Seaside application, the server will be running locally, and it is
easy to use the ordinary Pharo development tools in the server image.
Using the Object Inspector link in the web browser, open an inspector on the
underlying Pharo counter object, type self halt and click the doIt button.
The form submits, and the browser will hang, spinning. Now switch to the
Pharo Seaside image. You should see a pre-debugger window (Figure 14.5)
showing a WACounter object executing a halt. Examine this execution in the
debugger, and then Proceed. Go back to the web browser and notice that the
258
14.2
Getting started
Figure 14.6: Independent subcomponents.
Counter application is running again.
Multiple components
Seaside components can be instantiated multiple times and in different contexts.
Now, point your web browser to http://localhost:8080/examples/multicounter.
You will see a simple application composed of a number of independent
instances of the WACounter component, nested inside a WAMultiCounter.
Increment and decrement several of the counters. Verify that they behave
correctly even if you use the Back button. Toggle the halos to see how the
application is built out of nested components. Use the Seaside class browser to
view the implementation of WAMultiCounter. You should see two methods on
the class side (description, and initialize) and three on the instance side
(children, initialize, and renderContentOn:). Note that an application
is simply a component that is willing to be at the root of the component
containment hierarchy; this willingness is indicated by defining a class-side
method canBeRoot to answer true.
You can use the Seaside web interface to configure, copy or remove individual
applications (which are root-level components). Try making the following
configuration change.
Point your web browser to http://localhost:8080/config.
If you are asked for a login, supply the default login and password (admin
and seaside). You can see a list of a registered applications and dispatchers.
259
Seaside by example
Figure 14.7: Add a new application.
Change into the examples dispatcher list by clicking on the link on the left
side. With the buttons at the topbar, we can add new request handlers, open
the current, copy, remove or set it as the default page. We want to add a new
example application. Select Add, enter the name counter2 and choose the
type Application, click OK (see Figure 14.7).
On the next screen, set the Root Component to WACounter, apply the changes
(Figure 14.8).
Now we have a new counter installed at http://localhost:8080/examples/counter2.
Use the same Remove-button in the configuration interface to remove this
entry point.
Seaside operates in two modes: development mode, which is what we have seen
so far, and deployment mode, in which the toolbar is not available.
You can put a Seaside Application into deployment mode by removing the
inherited Root Decoration Class (again, see Figure 14.8).
Alternatively, you can disable the development toolbar for all new aplications
by evaluating the code:
AAdmin applicationDefaults
removeParent: WADevelopmentConfiguration instance
If you want to enable the password protection, you need to add the WAAuthConfiguration in the Inherited Configuration. Choose WAAuthConfiguration from the list of possible parents, after adding this configuration, you can
define the Login and Password setting.
14.3
Seaside components
As we mentioned in the previous section, Seaside applications are built out of
components. Let’s take a closer look at how Seaside works by implementing
260
Figure 14.8: Configure the new application.
Seaside by example
Figure 14.9: Hello World in Seaside.
the Hello World component.
Every Seaside component should inherit directly or indirectly from WAComponent, as shown in Figure 14.10. (Incidentally, the ’WA’ prefix often used in
Seaside code stands for ’Web Application’.)
Define a subclass of WAComponent called WAHelloWorld.
Components must know how to render themselves. Usually this is done by
implementing the method renderContentOn:, which gets as its argument an
instance of WAHtmlCanvas, which knows how to render HTML.
Implement the following method, and put it in a protocol called rendering:
WAHelloWorld >> renderContentOn: html
html text: 'hello world'
Now we must inform Seaside that this component is willing to be a standalone
application.
Implement the following method on the class side of WAHelloWorld.
WAHelloWorld class >> canBeRoot
^ true
We are almost done!
Point your web browser at http://localhost:8080/config, add a new entry point
called hello, and set its root component to be WAHelloWorld.
Now navigate to http://localhost:8080/hello. That’s it! You should see a web
page similar to Figure 14.9.
262
14.3
Seaside components
Figure 14.10: The WACounter class, which implements the counter application.
Methods with underlined names are on the class-side; those with plain-text names
are on the instance side.
State backtracking and the Counter application
The counter application is only slightly more complex than the hello world
application.
The class WACounter is a standalone application, so WACounter class must
answer true to the canBeRoot message. It must also register itself as an application; this is done in its class-side initialize method, as shown in Figure
14.10.
WACounter defines two methods, increase and decrease, which will be triggered from the ++ and –– links on the web page. It also defines an instance
variable count to record the state of the counter. However, we also want Seaside to synchronize the counter with the browser page: when the user clicks
on the browser’s Back button, we want seaside to backtrack the state of the
WACounter object. Seaside includes a general mechanism for backtracking, but
each application has to tell Seaside which parts of its state to track.
A component enables backtracking by implementing the states method on
the instance side: states should answer an array containing all the objects to
be tracked. In this case, the WACounter object adds itself to Seaside’s table of
backtrackable objects by returning Array with: self.
There is a subtle but important point to watch for when
declaring objects for backtracking. Seaside tracks state by making a copy of
Backtracking caveat
263
Seaside by example
all the objects declared in the states array. It does this using a WASnapshot
object; WASnapshot is a subclass of IdentityDictionary that records the
objects to be tracked as keys and shallow copies of their state as values. If the
state of an application is backtracked to a particular snapshot, the state of
each object entered into the snapshot dictionary is overwritten by the copy
saved in the snapshot.
Here is the point to watch out for: In the case of WACounter, you might think
that the state to be tracked is a number — the value of the count instance variable. However, having the states method answer Array with: count won’t
work. This is because the object named by count is an integer, and integers
are immutable. The increase and decrease methods don’t change the state
of the object 0 into 1 or the object 3 into 2. Instead, they make count hold a
different integer: every time the count is incremented or decremented, the
object named by count is replaced by another. This is why WACounter>>states
must return Array with: self. When the state of a WACounter object is replaced by a previous state, the value of each of the instance variable in the
object is replaced by a previous value; this correctly replaces the current value
of count by a prior value.
14.4
Rendering HTML
The purpose of a web application is to create, or render, web pages. As we mentioned in Section 14.3, each Seaside component is responsible for rendering
itself. So, let’s start our exploration of rendering by seeing how the counter
component renders itself.
Rendering the Counter
The rendering of the counter is relatively straightforward; the code is shown
in Figure 14.10. The current value of the counter is displayed as an HTML
heading, and the increment and decrement operations are implemented as
HTML anchors (that is, links) with callbacks to blocks that will send increase
and decrease to the counter object.
We will have a closer look at the rendering protocol in a moment. But before
we do, let’s have a quick look at the multi-counter.
From Counter to MultiCounter
WAMultiCounter, shown in Figure 14.11 is also a standalone application, so it
overrides canBeRoot to answer true. In addition, it is a composite component,
so Seaside requires it to declare its children by implementing a method children that answers an array of all the components it contains. It renders itself
by rendering each of its subcomponents, separated by a horizontal rule. Aside
264
14.4
Rendering HTML
Figure 14.11: WAMultiCounter.
from instance and class-side initialization methods, there is nothing else to
the multi-counter!
More about rendering HTML
As you can see from these examples, Seaside does not use templates to generate web pages. Instead it generates HTML programmatically. The basic idea is
that every Seaside component should override the method renderContentOn:;
this message will be sent by the framework to each component that needs to
be rendered. This renderContentOn: message will have an argument that is
an HTML canvas onto which the component should render itself. By convention, the HTML canvas parameter is called html. An HTML canvas is analogous
to the graphics canvas used by Morphic (and most other drawing frameworks)
to abstract away from the device-dependent details of drawing.
Here are some of the most basic rendering methods:
html text: 'hello world'. "render a plain text string"
html html: '&ndash;'. "render an HTML incantation"
html render: 1. "render any object"
The message render: anyObject can be sent to an HTML canvas to render
anyObject; it is normally used to render subcomponents. anyObject will
itself be sent the message renderContentOn: this is what happens in the
multi-counter (see Figure 14.11).
Using brushes
A canvas provides a number of brushes that can be used to render (i.e., paint)
content on the canvas. There are brushes for every kind of HTML element
— paragraphs, tables, lists, and so on. To see the full protocol of brushes
and convenience methods, you should browse the class WACanvas and its
subclasses. The argument to renderContentOn: is actually an instance of the
subclass WARender.
265
Seaside by example
Figure 14.12: RenderingDemo.
We have already seen the following brush used in the counter and multicounter examples:
html horizontalRule.
In Figure 14.12 we can see the output of many of the basic brushes offered
by Seaside. (The source code for method SeasideDemo >> renderContentOn:
defined below is in the package PBE-SeasideDemo in the project http://www.
squeaksource.com/PharoByExample.) The root component SeasideDemo simply renders its subcomponents, which are instances of SeasideHtmlDemo,
SeasideFormDemo, SeasideEditCallDemo and SeasideDialogDemo, as shown
below.
266
14.4
Rendering HTML
SeasideDemo >> renderContentOn: html
html heading: 'Rendering Demo'.
html heading
level: 2;
with: 'Rendering basic HTML: '.
html div
class: 'subcomponent';
with: htmlDemo.
"render the remaining components ..."
Recall that a root component must always declare its children, or Seaside will
refuse to render them.
SeasideDemo >> children
^ { htmlDemo . formDemo . editDemo . dialogDemo }
Notice that there are two different ways of instantiating the heading brush.
The first way is to set the text directly by sending the message heading:. The
second way is to instantiate the brush by sending heading, and then to send a
cascade of messages to the brush to set its properties and render it. Many of
the available brushes can be used in these two ways.
Important If you send a cascade of messages to a brush including the
message with:, then with: should be the final message. with: both sets
the content and renders the result.
In method renderContentOn: above, the first heading is at level 1, since this
is the default. We explicitly set the level of the second heading to 2. The subcomponent is rendered as an HTML div with the CSS class subcomponent. (More
on CSS in Section 14.5.) Also note that the argument to the with: keyword
message need not be a literal string: it can be another component, or even —
as in the next example — a block containing further rendering actions.
The SeasideHtmlDemo component demonstrates many of the most basic
brushes. Most of the code should be self-explanatory.
SeasideHtmlDemo >> renderContentOn: html
self renderParagraphsOn: html.
self renderListsAndTablesOn: html.
self renderDivsAndSpansOn: html.
self renderLinkWithCallbackOn: html
It is common practice to break up long rendering methods into many helper
methods, as we have done here.
A word of advice: Don’t put all your rendering code into a single method.
Instead, split it into helper methods named using the pattern render*On:.
All rendering methods go in the rendering protocol. Don’t send renderContentOn: from your own code, use render: instead.
267
Seaside by example
Look at the following code. The first helper method, SeasideHtmlDemo>>renderParagraphsOn:, shows you how to generate HTML paragraphs, plain and
emphasized text, and images. Note that in Seaside simple elements are rendered by specifying the text they contain directly, whereas complex elements
are specified using blocks. This is a simple convention to help you structure
your rendering code.
SeasideHtmlDemo >> renderParagraphsOn: html
html paragraph: 'A plain text paragraph.'.
html paragraph: [
html
text: 'A paragraph with plain text followed by a line break. ';
break;
emphasis: 'Emphasized text ';
text: 'followed by a horizontal rule.';
horizontalRule;
text: 'An image URI: '.
html image
url: self squeakImageUrl;
width: '50']
The next helper method, SeasideHtmlDemo>>renderListsAndTablesOn:,
shows you how to generate lists and tables. A table uses two levels of blocks to
display each of its rows and the cells within the rows.
SeasideHtmlDemo >> renderListsAndTablesOn: html
html orderedList: [
html listItem: 'An ordered list item'].
html unorderedList: [
html listItem: 'An unordered list item'].
html table: [
html tableRow: [
html tableData: 'A table with one data cell.']]
The next example shows how we can specify class or id attributes for div
and span elements (for use with CSS). Of course, the messages class: and id:
can also be sent to the other brushes, not just to div and span. The method
SeasideDemoWidget>>style defines how these HTML elements should be
displayed (see Section 14.5).
SeasideHtmlDemo >> renderDivsAndSpansOn: html
html div
id: 'author';
with: [
html text: 'Raw text within a div with id ''author''. '.
html span
class: 'highlight';
with: 'A span with class ''highlight''.']
Finally we see a simple example of a link, created by binding a simple callback
to an anchor (i.e., a link). Clicking on the link will cause the subsequent text
268
14.4
Rendering HTML
to toggle between true and false by toggling the instance variable toggleValue.
SeasideHtmlDemo >> renderLinkWithCallbackOn: html
html paragraph: [
html text: 'An anchor with a local action: '.
html span with: [
html anchor
callback: [toggleValue := toggleValue not];
with: 'toggle boolean:'].
html space.
html span
class: 'boolean';
with: toggleValue ]
Note that actions should appear only in callbacks.
The code executed while rendering should not change the state of the application!
Forms
Forms are rendered just like the other examples that we have already seen.
Here is the code for the SeasideFormDemo component in Figure 14.12.
SeasideFormDemo >> renderContentOn: html
| radioGroup |
html heading: heading.
html form: [
html span: 'Heading: '.
html textInput on: #heading of: self.
html select
list: self colors;
on: #color of: self.
radioGroup := html radioGroup.
html text: 'Radio on:'.
radioGroup radioButton
selected: radioOn;
callback: [radioOn := true].
html text: 'off:'.
radioGroup radioButton
selected: radioOn not;
callback: [radioOn := false].
html checkbox on: #checked of: self.
html submitButton
text: 'done' ]
Since a form is a complex entity, it is rendered using a block. Note that all the
state changes happen in the callbacks, not as part of the rendering.
269
Seaside by example
There is one Seaside feature used here that is worth special mention, namely
the message on:of:. In the example, this message is used to bind a text input
field to the variable heading. Anchors and buttons also support this message.
The first argument is the name of an instance variable for which accessors
have been defined; the second argument is the object to which this instance
variable belongs. Both observer (heading) and mutator (heading:) accessor
messages must be understood by the object, with the usual naming convention.
In the case of a text input field, this saves us the trouble of having to define a
callback that updates the field as well as having to bind the default contents of
the HTML input field to the current value of the instance variable. Using on:
#heading of: self, the heading variable is updated automatically whenever
the user updates the text input field.
The same message is used twice more in this example, to cause the selection
of a colour on the HTML form to update the color variable, and to bind the
result of the checkbox to the checked variable. Many other examples can be
found in the functional tests for Seaside. Have a look at the package SeasideTests-Functional, or just point your browser to http://localhost:8080/tests/
functional. Select WAInputPostFunctionalTest and click on the Restart
button to see most of the features of forms.
Don’t forget, if you toggle Halos, you can browse the source code of the examples directly using the Seaside class browser.
14.5
CSS: Cascading style sheets
Cascading Style Sheets (http://www.w3.org/Style/CSS/), or CSS for short, are a
standard way for web applications to separate style from content. Seaside relies on CSS to avoid cluttering your rendering code with layout considerations.
You can set the CSS style sheet for your web components by defining the
method style, which should return a string containing the CSS rules for that
component. The styles of all the components displayed on a web page are
joined together, so each component can have its own style. A better approach
can be to define an abstract class for your web application that defines a
common style for all its subclasses.
Actually, for deployed applications, it is more common to define style sheets as
external files. This way the look and feel of the component is completely separate from its functionality. (Have a look at WAFileLibrary, which provides a
way to serve static files without the need for a standalone server.)
If you already are familiar with CSS, then that’s all you need to know. Otherwise, read on for a very brief introduction to CSS.
Instead of directly encoding display attributes in the paragraph and text
elements of your web pages, CSS lets you define different classes of elements,
and place all display considerations in a separate style sheet.
270
14.5
CSS: Cascading style sheets
To put it another way, a CSS style sheet consists of a set of rules that specify
how to format given HTML elements. Each rule consists of two parts. There is
a selector that specifies which HTML elements the rule applies to, and there is a
declaration which sets a number of attributes for that element.
SeasideDemoWidget >> style
^ '
body {
font: 10pt Arial, Helvetica, sans-serif, Times New Roman;
}
h2 {
font-size: 12pt;
font-weight: normal;
font-style: italic;
}
table { border-collapse: collapse; }
td {
border: 2px solid #CCCCCC;
padding: 4px;
}
#author {
border: 1px solid black;
padding: 2px;
margin: 2px;
}
.subcomponent {
border: 2px solid lightblue;
padding: 2px;
margin: 2px;
}
.highlight { background-color: yellow; }
.boolean { background-color: lightgrey; }
.field { background-color: lightgrey; }
'
The previous method illustrates a simple style sheet for the rendering demo
shown earlier in Figure 14.12. The first rule specifies a preference for the fonts
to use for the body of the web page. The next few rules specify properties of
second-level headings (h2), tables (table), and table data (td).
The remaining rules have selectors that will match HTML elements that have
the given class or id attributes. CSS selectors for class attributes start with
a . and those for id attributes with #. The main difference between class and
id attributes is that many elements may have the same class, but only one
element may have a given id (i.e., an identifier). So, whereas a class attribute,
such as highlight, may occur multiple times on any page, an id must identify
a unique element on the page, such as a particular menu, the modified date, or
author. Note that a particular HTML element may have multiple classes, in
which case all the applicable display attributes will be applied in sequence.
271
Seaside by example
Selector conditions may be combined, so the selector div.subcomponent will
only match an HTML element if it is both a div and it has a class attribute
equal to subcomponent.
It is also possible to specify nested elements, though this is seldom necessary.
For example, the selector p span will match a span within a paragraph but not
within a div.
There are numerous books and web sites to help you learn CSS. For a dramatic
demonstration of the power of CSS, we recommend you to have a look at the
CSS Zen Garden (http://www.csszengarden.com/), which shows how the same
content can be rendered in radically different ways simply by changing the
CSS style sheet.
14.6
Managing control flow
Seaside makes it particularly easy to design web applications with non-trivial
control flow. There are basically two mechanisms that you can use:
1. A component can call another component by sending caller call: callee.
The caller is temporarily replaced by the callee, until the callee returns control by sending answer:. The caller is usually self, but could also be any other
currently visible component.
2. A workflow can be defined as a task. This is a special kind of component that
subclasses WATask (instead of WAComponent). Instead of defining renderContentOn:, it defines no content of its own, but rather defines a go method that
sends a series of call: messages to activate various subcomponents in turn.
Call and answer
Call and answer are used to realize simple dialogues.
There is a trivial example of call: and answer: in the rendering demo of
Figure 14.12. The component SeasideEditCallDemo displays a text field and
an edit link. The callback for the edit link calls a new instance of SeasideEditAnswerDemo initialized to the value of the text field. The callback also updates this text field to the result which is sent as an answer.
(We underline the call: and answer: sends to draw attention to them.)
SeasideEditCallDemo >> renderContentOn: html
html span
class: 'field';
with: self text.
html space.
html anchor
callback: [self text: (self call: (SeasideEditAnswerDemo new text:
self text))];
with: 'edit'
272
14.6
Managing control flow
What is particularly elegant is that the code makes absolutely no reference to
the new web page that must be created. At run-time, a new page is created in
which the SeasideEditCallDemo component is replaced by a SeasideEditAnswerDemo component; the parent component and the other peer components
are untouched.
It is important to keep in mind that call: and answer: should never be used
while rendering. They may safely be sent from within a callback, or from
within the go method of a task.
The SeasideEditAnswerDemo component is also remarkably simple. It just
renders a form with a text field. The submit button is bound to a callback that
will answer the final value of the text field.
SeasideEditAnswerDemo >> renderContentOn: html
html form: [
html textInput
on: #text of: self.
html submitButton
callback: [ self answer: self text ];
text: 'ok'.
]
That’s it.
Seaside takes care of the control flow and the correct rendering of all the
components. Interestingly, the Back button of the browser will also work just
fine (though side effects are not rolled back unless we take additional steps).
Convenience methods
Since certain call–answer dialogues are very common, Seaside provides some
convenience methods to save you the trouble of writing components like
SeasideEditAnswerDemo. The generated dialogues are shown in Figure
14.13. We can see these convenience methods being used within SeasideDialogDemo>>renderContentOn:
The message request: performs a call to a component that will let you edit
a text field. The component answers the edited string. An optional label and
default value may also be specified.
SeasideDialogDemo >> renderContentOn: html
html anchor
callback: [ self request: 'edit this' label: 'done' default: 'some
text' ];
with: 'self request:'.
...
The message inform: calls a component that simply displays the argument
message and waits for the user to click Ok. The called component just returns
self.
273
Seaside by example
Figure 14.13: Some standard dialogs.
...
html space.
html anchor
callback: [ self inform: 'yes!' ];
with: 'self inform:'.
...
The message confirm: asks a question and waits for the user to select either
Yes or No. The component answers a boolean, which can be used to perform
further actions.
...
html space.
html anchor
callback: [
(self confirm: 'Are you happy?')
ifTrue: [ self inform: ':-)' ]
ifFalse: [ self inform: ':-(' ]
];
with: 'self confirm:'.
A few further convenience methods, such as chooseFrom:caption:, are defined in the convenience protocol of WAComponent.
Tasks
A task is a component that subclasses WATask. It does not render anything
itself, but simply calls other components in a control flow defined by implementing the method go.
WAConvenienceTest is a simple example of a task defined in the package
Seaside-Tests-Functional. To see its effect, just point your browser to http:
274
14.6
Managing control flow
Figure 14.14: A simple task.
//localhost:8080/tests/functional, select WAFlowConvenienceFunctionalTest
and click Restart.
WAFlowConvenienceFunctionalTest >> go
[ self chooseCheese.
self confirmCheese ] whileFalse.
self informCheese
This task calls in turn three components. The first, generated by the convenience method chooseFrom: caption:, is a WAChoiceDialog that asks the
user to choose a cheese.
WAFlowConvenienceFunctionalTest >> chooseCheese
cheese := self
chooseFrom: #('Greyerzer' 'Tilsiter' 'Sbrinz')
caption: 'What''s your favorite Cheese?'.
cheese isNil ifTrue: [ self chooseCheese ]
The second is a WAYesOrNoDialog to confirm the choice (generated by the
convenience method confirm:).
WAFlowConvenienceFunctionalTest >> confirmCheese
^self confirm: 'Is ', cheese, ' your favorite cheese?'
Finally a WAFormDialog is called (via the convenience method inform:).
WAFlowConvenienceFunctionalTest >> informCheese
self inform: 'Your favorite cheese is ', cheese, '.'
The generated dialogues are shown in Figure 14.14.
Transactions
We saw in Section 14.3 that Seaside can keep track of the correspondence
between the state of components and individual web pages by having components register their state for backtracking: all that a component need do
275
Seaside by example
Figure 14.15: The Sushi Store.
is implement the method states to answer an array of all the objects whose
state must be tracked.
Sometimes, however, we do not want to backtrack state: instead we want to
prevent the user from accidentally undoing effects that should be permanent.
This is often referred to as ”the shopping cart problem”. Once you have
checked out your shopping cart and paid for the items you have purchased, it
should not be possible to go Back with the browser and add more items to the
shopping cart!
Seaside allows you to enforce restriction this by defining a task within which
certain actions are grouped together as transactions. You can backtrack within
a transaction, but once a transaction is complete, you can no longer go back
to it. The corresponding pages are invalidated, and any attempt to go back to
them will cause Seaside to generate a warning and redirect the user to the
most recent valid page.
The Seaside Sushi Store is sample application that illustrates many of the
features of Seaside, including transactions. This application is bundled with
your installation of Seaside, so you can try it out by pointing your browser
at http://localhost:8080/seaside/examples/store. (If you cannot find it in your
image, there is a version of the sushi store available on SqueakSource from
http://www.squeaksource.com/SeasideExamples/.)
The sushi store supports the following workflow:
• Visit the store.
• Browse or search for sushi.
• Add sushi to your shopping cart.
• Checkout.
• Verify your order.
276
14.6
Managing control flow
• Enter shipping address.
• Verify shipping address.
• Enter payment information.
• Your fish is on its way!
If you toggle the halos, you will see that the top-level component of the sushi
store is an instance of WAStore. It does nothing but render the title bar, and
then it renders task, an instance of WAStoreTask.
WAStore >> renderContentOn: html
"... render the title bar ..."
html div id: 'body'; with: task
WAStoreTask captures this workflow sequence. At a couple of points it is critical that the user not be able to go back and change the submitted information.
Purchase some sushi and then use the Back button to try to put more sushi
into your cart.
You will get the message That page has expired.
Seaside lets the programmer say that a certain part of a workflow acts like a
transaction: once the transaction is complete, the user cannot go back and
undo it. This is done by sending isolate: to a task with the transactional
block as its argument. We can see this in the sushi store workflow as follows:
WAStoreTask >> go
| shipping billing creditCard |
cart := WAStoreCart new.
self isolate:
[[ self fillCart.
self confirmContentsOfCart ]
whileFalse ].
self isolate:
[ shipping := self getShippingAddress.
billing := (self useAsBillingAddress: shipping)
ifFalse: [ self getBillingAddress ]
ifTrue: [ shipping ].
creditCard := self getPaymentInfo.
self shipTo: shipping billTo: billing payWith: creditCard ].
self displayConfirmation.
Here we see quite clearly that there are two transactions. The first fills the
cart and closes the shopping phase. (The helper methods such as fillCart
take care of instantiating and calling the right subcomponents.) Once you
have confirmed the contents of the cart you cannot go back without starting
a new session. The second transaction completes the shipping and payment
data. You can navigate back and forth within the second transaction until you
277
Seaside by example
confirm payment. However, once both transactions are complete, any attempt
to navigate back will fail.
Transactions may also be nested. A simple demonstration of this is found in
the class WANestedTransaction. The first isolate: takes as argument a block
that contains another, nested isolate:
WANestedTransaction >> go
self inform: 'Before parent txn'.
self isolate:
[self inform: 'Inside parent txn'.
self isolate: [self inform: 'Inside child txn'].
self inform: 'Outside child txn'].
self inform: 'Outside parent txn'
Go to http://localhost:8080/tests/functionals, select WATransactionTest and
click on Restart. Try to navigate back and forth within the parent and child
transaction by clicking the Back button and then clicking Ok. Note that as soon
as a transaction is complete, you can no longer go back inside the transaction
without generating an error upon clicking Ok.
14.7
A complete tutorial example
Let’s see how we can build a complete Seaside application from scratch. (The
exercise should take at most a couple of hours. If you prefer to just look at
the completed source code, you can grab it from the SqueakSource project
http://www.squeaksource.com/PharoByExample. The package to load is PBESeasideRPN. The tutorial that follows uses slightly different class names so
that you can compare your implementation with ours.) We will build a RPN
(Reverse Polish Notation) calculator as a Seaside application that uses a simple
stack machine as its underlying model. Furthermore, the Seaside interface
will let us toggle between two displays — one which just shows us the current
value on top of the stack, and the other which shows us the complete state of
the stack. The calculator with the two display options is shown in Figure 14.16.
We begin by implementing the stack machine and its tests.
First, Define a new class called MyStackMachine with an instance variable
contents initialized to a new OrderedCollection.
MyStackMachine >> initialize
super initialize.
contents := OrderedCollection new.
The stack machine should provide operations to push: and pop values, view
the top of the stack, and perform various arithmetic operations to add, subtract, multiply and divide the top values on the stack.
Write some tests for the stack operations and then implement these operations. Here is a sample test:
278
14.7
A complete tutorial example
Figure 14.16: RPN calculator and its stack machine.
MyStackMachineTest >> testDiv
stack
push: 3;
push: 4;
div.
self assert: stack size = 1.
self assert: stack top = (4/3).
You might consider using some helper methods for the arithmetic operations
to check that there are two numbers on the stack before doing anything, and
raising an error if this precondition is not fulfilled. (It’s a good idea to use
Object>>assert: to specify the preconditions for an operation. This method
will raise an AssertionFailure if the user tries to use the stack machine in an
invalid state.) If you do this, most of your methods will just be one or two lines
long.
You might also consider implementing MyStackMachine>>printOn: to make
it easier to debug your stack machine implementation with the help of an
object inspector. (Hint: just delegate printing to the contents variable.)
Complete the MyStackMachine by writing operations dup (push a duplicate of
the top value onto the stack), exch (exchange the top two values), and rotUp
(rotate the entire stack contents up — the top value will move to the bottom).
Now we have a simple stack machine implementation. We can start to implement the Seaside RPN Calculator.
We will make use of 5 classes:
1. MyRPNWidget — this should be an abstract class that defines the common
CSS style sheet for the application, and other common behavior for the components of the RPN calculator. It is a subclass of WAComponent and the direct
superclass of the following four classes.
279
Seaside by example
2. MyCalculator — this is the root component. It should register the application (on the class side), it should instantiate and render its subcomponents,
and it should register any state for backtracking.
3. MyKeypad - this displays the keys that we use to interact with the calculator.
4. MyDisplay — this component displays the top of the stack and provides a
button to call another component to display the detailed view.
5. MyDisplayStack — this component shows the detailed view of the stack and
provides a button to answer back. It is a subclass of MyDisplay.
Create a package MyCalculator, and define MyRPNWidget.
Define the common style for the application.
Here is a minimal CSS for the application. You can make it more fancy if you
like.
MyRPNWidget >> style
^ 'table.keypad { float: left; }
td.key {
border: 1px solid grey;
background: lightgrey;
padding: 4px;
text-align: center;
}
table.stack { float: left; }
td.stackcell {
border: 2px solid white;
border-left-color: grey;
border-right-color: grey;
border-bottom-color: grey;
padding: 4px;
text-align: right;
}
td.small { font-size: 8pt; }'
Define MyCalculator to be a root component and register itself as an application (i.e., implement canBeRoot and initialize on the class side).
Implement MyCalculator>>renderContentOn: to render something trivial
(such as its name), and verify that the application runs in a browser.
MyCalculator is responsible for instantiating MyStackMachine, MyKeypad and
MyDisplay.
Define MyKeypad and MyDisplay as subclasses of MyRPNWidget.
All three components will need access to a common instance of the stack
machine, so define the instance variable stackMachine and an initialization
method setMyStackMachine: in the common parent, MyRPNWidget. Add
instance variables keypad and display to MyCalculator and initialize them in
MyCalculator>>initialize. (Don’t forget to send super initialize!)
280
14.7
A complete tutorial example
Figure 14.17: Displaying the top of the stack.
Pass the shared instance of the stack machine to the keypad and the display in
the same initialize method.
Implement MyCalculator>>renderContentOn: to simply render in turn the
keypad and the display. To correctly display the subcomponents, you must
implement MyCalculator>>children to return an array with the keypad and
the display. Implement placeholder rendering methods for the keypad and the
display and verify that the calculator now displays its two subcomponents.
Now we will change the implementation of the display to show the top value
of the stack.
Use a table with class keypad containing a row with a single table data cell
with class stackcell.
Change the rendering method of the keypad to ensure that the number 0
is pushed on the stack in case it is empty. (Define and use MyKeypad>>ensureStackMachineNotEmpty.) Also make it display an empty table with class
keypad. Now the calculator should display a single cell containing the value 0.
If you toggle the halos, you should see something like this: Figure 14.17
Now let’s implement an interface to interact with the stack.
First, define the following helper methods, which will make it easier to script
the interface:
MyKeypad >> renderStackButton: text callback: aBlock colSpan: anInteger
on: html
html tableData
class: 'key';
colSpan: anInteger;
with:
[ html anchor
callback: aBlock;
with: [ html html: text ]]
MyKeypad >> renderStackButton: text callback: aBlock on: html
self
renderStackButton: text
281
Seaside by example
callback: aBlock
colSpan: 1
on: html
We will use these two methods to define the buttons on the keypad with
appropriate callbacks. Certain buttons may span multiple columns, but the
default is to occupy just one column.
Use the two helper methods to script the keypad as follows.
(Hint: start by getting the digit and Enter keys working, then the arithmetic
operators.)
MyKeypad >> renderContentOn: html
self ensureStackMachineNotEmpty.
html table
class: 'keypad';
with: [
html tableRow: [
self renderStackButton: '+' callback: [self stackOp: #add] on:
html.
self renderStackButton: '&ndash;' callback: [self stackOp:
#min] on: html.
self renderStackButton: '&times;' callback: [self stackOp:
#mul] on: html.
self renderStackButton: '&divide;' callback: [self stackOp:
#div] on: html.
self renderStackButton: '&plusmn;' callback: [self stackOp:
#neg] on: html ].
html tableRow: [
self renderStackButton: '1' callback: [self type: '1'] on:
html.
self renderStackButton: '2' callback: [self type: '2'] on:
html.
self renderStackButton: '3' callback: [self type: '3'] on:
html.
self renderStackButton: 'Drop' callback: [self
stackPopIfNotEmpty]
colSpan: 2 on: html ].
" and so on ... "
html tableRow: [
self renderStackButton: '0' callback: [self type: '0']
colSpan: 2 on: html.
self renderStackButton: 'C' callback: [self stackClearTop] on:
html.
self renderStackButton: 'Enter'
callback: [self stackOp: #dup. self setClearMode]
colSpan: 2 on: html ]]
Check that the keypad displays properly. If you try to click on the keys, however, you will find that the calculator does not work yet.
282
14.7
A complete tutorial example
Implement MyKeypad>>type: to update the top of the stack by appending the
typed digit. You will need to convert the top value to a string, update it, and
convert it back to an integer, something like this:
MyKeypad >> type: aString
stackMachine push: (self stackPopTopOrZero asString, aString)
asNumber.
The two methods stackPopTopOrZero and stackPopIfNotEmpty are used to
guard against operating on an empty stack.
MyKeypad >> stackPopTopOrZero
^ stackMachine isEmpty
ifTrue: [ 0 ]
ifFalse: [ stackMachine pop ]
MyKeypad >> stackPopIfNotEmpty
stackMachine isEmpty
ifFalse: [ stackMachine pop ]
Now when you click on the digit keys the display should be updated. (Be sure
that MyStackMachine>>pop returns the value popped, or this will not work!)
Next, we must implement MyKeypad>>stackOp:. Something like this will do
the trick:
MyKeypad >> stackOp: op
[ stackMachine perform: op ] on: AssertionFailure do: [ ].
The point is that we are not sure that all operations will succeed. For example,
addition will fail if we do not have two numbers on the stack. For the moment
we can just ignore such errors. If we are feeling more ambitious later on, we
can provide some user feedback in the error handler block.
The first version of the calculator should be working now. Try to enter some
numbers by pressing the digit keys, hitting Enter to push a copy of the current value, and entering + to sum the top two values.
You will notice that typing digits does not behave the way you might expect.
Actually the calculator should be aware of whether you are typing a new
number, or appending to an existing number.
Adapt MyKeypad>>type: to behave differently depending on the current
typing mode.
Introduce an instance variable mode which takes on one of the three values:
the symbol #typing (when you are typing), #push (after you have performed a
calculator operation and typing should force the top value to be pushed), or
#clear (after you have performed Enter and the top value should be cleared
before typing). The new type: method might look like this:
MyKeypad >> type: aString
self inPushMode ifTrue: [
283
Seaside by example
stackMachine push: stackMachine top.
self stackClearTop ].
self inClearMode ifTrue: [ self stackClearTop ].
stackMachine push: (self stackPopTopOrZero asString, aString)
asNumber.
Typing might work better now, but it is still frustrating not to be able to see
what is on the stack.
Define MyDisplayStack as a subclass of MyDisplay.
Add a button to the rendering method of MyDisplay which will call a new
instance of MyDisplayStack. You will need an HTML anchor that looks something like this:
html anchor
callback: [ self call: (MyDisplayStack new setMyStackMachine:
stackMachine)];
with: 'open'
The callback will cause the current instance of MyDisplay to be temporarily
replaced by a new instance of MyDisplayStack whose job it is to display the
complete stack. When this component signals that it is done (i.e., by sending
self answer), then control will return to the original instance of MyDisplay.
Define the rendering method of MyDisplayStack to display all of the values on
the stack.
(You will either need to define an accessor for the stack machine’s contents
or you can define MyStackMachine>>do: to iterate over the stack values.) The
stack display should also have a button labelled close whose callback will
simply perform self answer.
html anchor
callback: [ self answer];
with: 'close'
Now you should be able to open and close the stack while you are using the
calculator.
There is, however, one thing we have forgotten. Try to perform some operations on the stack. Now use the Back button of your browser and try to
perform some more stack operations. For example, open the stack, type 1,
Enter twice and +. The stack should display 2 and 1. Now hit the Back button.
The stack now shows three times 1 again. Now if you type +, the stack shows 3.
Backtracking is not yet working.
Implement MyCalculator>>states to return an array with the contents of
the stack machine.
Check that backtracking now works correctly.
284
14.8
14.8
A quick look at AJAX
A quick look at AJAX
AJAX (Asynchronous JavaScript and XML) is a technique to make web applications more interactive by exploiting JavaScript functionality on the client
side.
Two well-known JavaScript libraries are Prototype (http://www.prototypejs.org)
and script.aculo.us (http://script.aculo.us). Prototype provides a framework
to ease writing JavaScript. script.aculo.us provides some additional features to support animations and drag-and-drop on top of Prototype. Both
frameworks are supported in Seaside through the package Scriptaculous.
All ready made images have the Scriptaculous package extensions already
loaded. The latest version is available from http://www.squeaksource.com/
Seaside. An online demo is available at http://scriptaculous.seasidehosting.
st. Alternatively, if you have a enabled image running, simply go to http:
//localhost:8080/javascript/scriptaculous.
The Scriptaculous extensions follow the same approach as Seaside itself —
simply configure Pharo objects to model your application, and the needed
Javascript code will be generated for you.
Let us look at a simple example of how client-side Javascript support can
make our RPN calculator behave more naturally. Currently every keystroke to
enter a digit generates a request to refresh the page. We would like instead to
handle editing of the display on the client-side by updating the display in the
existing page.
To address the display from JavaScript code, we must first give it a unique id.
Update the calculator’s rendering method as follows. (If you have not implemented the tutorial example yourself, you can simply load the complete example (PBE-SeasideRPN) from http://www.squeaksource.com/PharoByExample and
apply the suggested changes to the classes RPN* instead of My*.
MyCalculator >> renderContentOn: html
html div id: 'keypad'; with: keypad.
html div id: 'display'; with: display.
To be able to re-render the display when a keyboard button is pressed, the
keyboard needs to know the display component.
Add a display instance variable to MyKeypad, an initialize method MyKeypad>>setDisplay:, and call this from MyCalculator>>initialize. Now we
are able to assign some JavaScript code to the buttons by updating MyKeypad>>renderStackButton:callback:colSpan:on: as follows:
MyKeypad >> renderStackButton: text callback: aBlock colSpan: anInteger
on: html
html tableData
class: 'key';
colSpan: anInteger;
285
Seaside by example
with: [
html anchor
callback: aBlock;
onClick:
"handle Javascript event"
(html scriptaculous updater
id: 'display';
callback: [ :r |
aBlock value.
r render: display ];
return: false);
with: [ html html: text ] ]
onClick: specifies a JavaScript event handler. html updater returns an instance of PTUpdater, a Smalltalk object representing the JavaScript AJAX
Updater object (http://www.prototypejs.org/api/ajax/updater). This object performs AJAX requests and updates a container’s contents based on the response
text. id: tells the updater what HTML DOM element to update, in this case the
contents of the div element with the id ’display’. callback: specifies a block
that is triggered when the user presses the button. The block argument is a
new renderer r, which we can use to render the display component. (Note:
Even though HTML is still accessible, it is not valid anymore at the time this
callback block is evaluated). Before rendering the display component we
evaluate aBlock to perform the desired action.
return: false tells the JavaScript engine to not trigger the original link
callback, which would cause a full refresh. We could instead remove the
original anchor callback:, but like this the calculator will still work even if
JavaScript is disabled.
Try the calculator again, and notice how a full page refresh is triggered every time you press a digit key. (The URL of the web page is updated at each
keystroke.)
Although we have implemented the client-side behavior, we have not yet
activated it. Now we will enable the Javascript event handling.
Click on the Configure link in the toolbar of the calculator.
Configure the Libraries attribute under the General section. (You may
need to enable the modification of this attribute, by first selecting Modify).
From the list of available libraries, select PTDevelopmentLibrary and apply
the changes.
Instead of manually adding the library, you may also do it programmatically
when you register the application:
MyCalculator class >> initialize
(WAAdmin register: self asApplicationAt: self applicationName)
addLibrary: PTDevelopmentLibrary
286
14.8
A quick look at AJAX
Figure 14.18: Seaside AJAX processing (simplified).
For this example the PTDevelopmentLibrary is sufficient, but for the full set
of the scriptaculous extensions you need to add the SUDevelopmentLibrary,
too.
Try the revised application. Note that the feedback is much more natural. In
particular, a new URL is not generated with each keystroke.
You may well ask, yes, but how does this work? Figure 14.18 shows how the RPN
applications would both without and with AJAX. Basically AJAX short-circuits
the rendering to only update the display component. Javascript is responsible
both for triggering the request and updating the corresponding DOM element.
Have a look at the generated source-code, especially the JavaScript code:
new Ajax.Updater(
'display',
'http://localhost/seaside/RPN+Calculator',
{'evalScripts': true,
'parameters': ['UNDERSCOREs=zcdqfonqwbeYzkza',
'UNDERSCOREk=jMORHtqr','9'].join('&')});
287
Seaside by example
return false
For more advanced examples, have a further look at http://localhost:8080/
javascript/scriptaculous.
Hints
In case of server side problems use the Debugger. In case of client side problems use FireFox (http://www.mozilla.com) with the JavaScript debugger FireBug (http://www.getfirebug.com/) plugin enabled.
14.9
Chapter summary
• The easiest way to get started is to download the Seaside One-Click Experience from http://seaside.st
• Turn the server on and off by evaluating ZnZincServerAdaptor startOn: 8080 and ZnZincServerAdaptor stop.
• Toggle Halos to directly view application source code, run-time objects, CSS and HTML.
• Remove the root decoration class WAToolDecoration in the application
configuration, to disable the toolbar
• Send WAAdmin applicationDefaults removeParent: WADevelopmentConfiguration instance, to disable toolbar for new Components
• Seaside web applications are composed of components, each of which is
an instance of a subclass of WAComponent.
• Only a root component may be registered as an application. It should
implement canBeRoot on the class side. Alternatively it may register
itself as an application in its class-side initialize method by sending
WAAdmin register: self asApplicationAt: application path. If you
override description it is possible to return a descriptive application
name that will be displayed in the configuration editor.
• To backtrack state, a component must implement the states method to
answer an array of objects whose state will be restored if the user clicks
the browser’s Back button.
• A component renders itself by implementing renderContentOn:. The
argument to this method is an HTML rendering canvas (usually called
html).
• A component can render a subcomponent by sending self render:
subcomponent.
288
14.9
Chapter summary
• HTML is generated programmatically by sending messages to brushes.
A brush is obtained by sending a message, such as paragraph or div, to
the HTML canvas.
• If you send a cascade of messages to a brush that includes the message
with:, then with: should be the last message sent. The with: message
sets the contents and renders the result.
• Actions should appear only in callbacks. You should not change the
state of the application while you are rendering it.
• You can bind various form widgets and anchors to instance variables
with accessors by sending the message on: instance variable of: object to
the brush.
• You can define the CSS for a component hierarchy by defining the
method style, which should return a string containing the style sheet.
(For deployed applications, it is more usual to refer to a style sheet
located at a static URL.)
• Control flows can be programmed by sending x call: y, in which case
component x will be replaced by y until y answers by sending answer:
with a result in a callback. The receiver of call: is usually self, but
may in general be any visible component.
• A control flow can also be specified as a task — a instance of a subclass
of WATask. It should implement the method go, which should call: a
series of components in a workflow.
• Use WAComponents’s convenience methods request:, inform:, confirm:
and chooseFrom:caption: for basic interactions.
• To prevent the user from using the browser’s Back button to access
a previous execution state of the web application, you can declare
portions of the workflow to be a transaction by enclosing them in an
isolate: block.
289
CHAPTER
15
Classes and metaclasses
As we saw in preceding chapters, in Pharo, everything is an object, and every
object is an instance of a class. Classes are no exception: classes are objects,
and class objects are instances of other classes. This object model captures
the essence of object-oriented programming, and is lean, simple, elegant
and uniform. However, the implications of this uniformity may confuse
newcomers.
Note that you do not need to fully understand the implications of this uniformity to program fluently in Pharo. Nevertheless, the goal of this chapter is
twofold: (1) go as deep as possible and (2) show that there is nothing complex,
magic or special here: just simple rules applied uniformly. By following these
rules you can always understand why the situation is the way that it is.
15.1
Rules for classes and metaclasses
The Pharo object model is based on a limited number of concepts applied
uniformly. To refresh your memory, here are the rules of the object model
that we explored in Chapter : The Pharo Object Model.
Rule 1 Everything is an object.
Rule 2 Every object is an instance of a class.
Rule 3 Every class has a superclass.
Rule 4 Everything happens by sending messages.
Rule 5 Method lookup follows the inheritance chain.
As we mentioned in the introduction to this chapter, a consequence of Rule
1 is that classes are objects too, so Rule 2 tells us that classes must also be in291
Classes and metaclasses
Figure 15.1: Sending the message class to a sorted collection
stances of classes. The class of a class is called a metaclass. A metaclass is
created automatically for you whenever you create a class. Most of the time
you do not need to care or think about metaclasses. However, every time that
you use the browser to browse the class side of a class, it is helpful to recall that
you are actually browsing a different class. A class and its metaclass are two
separate classes, even though the former is an instance of the latter.
To properly explain classes and metaclasses, we need to extend the rules from
Chapter : The Pharo Object Model with the following additional rules.
Rule 6 Every class is an instance of a metaclass.
Rule 7 The metaclass hierarchy parallels the class hierarchy.
Rule 8 Every metaclass inherits from Class and Behavior.
Rule 9 Every metaclass is an instance of Metaclass.
Rule 10 The metaclass of Metaclass is an instance of Metaclass.
Together, these 10 rules complete Pharo’s object model.
We will first briefly revisit the 5 rules from Chapter : The Pharo Object Model
with a small example. Then we will take a closer look at the new rules, using
the same example.
15.2
Revisiting the Pharo object model
Rule 1. Since everything is an object, an ordered collection in Pharo is also an
object.
292
15.3
Every class is an instance of a metaclass
OrderedCollection withAll: #(4 5 6 1 2 3)
>>> an OrderedCollection(4 5 6 1 2 3)
Rule 2. Every object is an instance of a class. The class of an ordered collection
is the class OrderedCollection:
(OrderedCollection withAll: #(4 5 6 1 2 3)) class
>>> OrderedCollection
Rule 3. Every class has a superclass. The superclass of OrderedCollection is
SequenceableCollection and the superclass of SequenceableCollection is
Collection:
OrderedCollection superclass
>>> SequenceableCollection
SequenceableCollection superclass
>>> Collection
Collection superclass
>>> Object
Let us take an example. When we send the message asSortedCollection, we
convert the ordered collection into a sorted collection. We verify simply as
follows:
(OrderedCollection withAll: #(4 5 6 1 2 3)) asSortedCollection class
>>> SortedCollection
Rule 4. Everything happens by sending messagess, so we can deduce that
withAll: is a message to OrderedCollection and asSortedCollection are
messages sent to the ordered collection instance, and superclass is a message
to OrderedCollection and SequenceableCollection, and Collection. The
receiver in each case is an object, since everything is an object, but some of
these objects are also classes.
Rule 5. Method lookup follows the inheritance chain, so when we send the
message class to the result of (OrderedCollection withAll: #(4 5 6 1 2
3)) asSortedCollection, the message is handled when the corresponding
method is found in the class Object, as shown in Figure 15.1.
15.3
Every class is an instance of a metaclass
As we mentioned earlier in Section 15.1, classes whose instances are themselves classes are called metaclasses.
Metaclasses are implicit
Metaclasses are automatically created when you define a class. We say that
they are implicit since as a programmer you never have to worry about them.
293
Classes and metaclasses
Figure 15.2: The metaclasses of SortedCollection and its superclasses (elided).
An implicit metaclass is created for each class you create, so each metaclass
has only a single instance.
Whereas ordinary classes are named, metaclasses are anonymous. However,
we can always refer to them through the class that is their instance. The class
of SortedCollection, for instance, is SortedCollection class, and the class
of Object is Object class:
SortedCollection class
>>> SortedCollection class
Object class
>>> Object class
In fact metaclasses are not truly anonymous, their name is deduced from the
one of their single instance
SortedCollection class name
>>> 'SortedCollection class'
Figure 15.2 shows how each class is an instance of its metaclass. Note that
we only skip SequenceableCollection and Collection from the figures and
explanation due to space constraints. Their absence does not change the
overall meaning.
Querying Metaclasses
The fact that classes are also objects makes it easy for us to query them by
sending messages. Let’s have a look:
OrderedCollection subclasses
>>> {SortedCollection . ObjectFinalizerCollection .
WeakOrderedCollection . OCLiteralList . GLMMultiValue}
SortedCollection subclasses
>>> #()
294
15.4
The metaclass hierarchy parallels the class hierarchy
Figure 15.3: The metaclass hierarchy parallels the class hierarchy (elided).
SortedCollection allSuperclasses
>>> an OrderedCollection(OrderedCollection SequenceableCollection
Collection Object ProtoObject)
SortedCollection instVarNames
>>> #('sortBlock')
SortedCollection allInstVarNames
>>> #('array' 'firstIndex' 'lastIndex' 'sortBlock')
SortedCollection selectors
>>> #(#indexForInserting: #sort:to: #addAll: #reSort #sortBlock:
#copyEmpty #addFirst: #insert:before: #defaultSort:to: #median
#at:put: #add: #= #collect: #flatCollect: #sort: #join: #sortBlock)
15.4
The metaclass hierarchy parallels the class hierarchy
Rule 7 says that the superclass of a metaclass cannot be an arbitrary class: it
is constrained to be the metaclass of the superclass of the metaclass’s unique
instance.
SortedCollection class superclass
>>> OrderedCollection class
SortedCollection superclass class
>>> OrderedCollection class
This is what we mean by the metaclass hierarchy being parallel to the class
hierarchy. Figure 15.3 shows how this works in the SortedCollection hierarchy.
SortedCollection class
>>> SortedCollection class
SortedCollection class superclass
>>> OrderedCollection class
295
Classes and metaclasses
Figure 15.4: Message lookup for classes is the same as for ordinary objects.
SortedCollection class superclass superclass
>>> SequenceableCollection class
SortedCollection class superclass superclass superclass superclass
>>> Object class
Uniformity between Classes and Objects
It is interesting to step back a moment and realize that there is no difference
between sending a message to an object and to a class. In both cases the
search for the corresponding method starts in the class of the receiver, and
proceeds up the inheritance chain.
Thus, messages sent to classes must follow the metaclass inheritance chain.
Consider, for example, the method withAll:, which is implemented on the
class side of Collection. When we send the message withAll: to the class
OrderedCollection, then it is looked up the same way as any other message.
The lookup starts in OrderedCollection class (since it starts in the class
of the receiver and the receiver is OrderedCollection), and proceeds up the
metaclass hierarchy until it is found in Collection class (see Figure 15.4). It
returns a new instance of OrderedCollection.
OrderedCollection withAll: #(4 5 6 1 2 3)
>>> an OrderedCollection (4 5 6 1 2 3)
Only one method lookup
Thus we see that there is one uniform kind of method lookup in Pharo. Classes
are just objects, and behave like any other objects. Classes have the power to
create new instances only because classes happen to respond to the message
new, and because the method for new knows how to create new instances.
296
15.5
Every metaclass inherits from Class and Behavior
Figure 15.5: Classes are objects too.
Normally, non-class objects do not understand this message, but if you have a
good reason to do so, there is nothing stopping you from adding a new method
to a non-metaclass.
Inspecting objects and classes
Since classes are objects, we can also inspect them.
Inspect OrderedCollection withAll: #(4 5 6 1 2 3) and OrderedCollection.
Notice that in one case you are inspecting an instance of OrderedCollection
and in the other case the OrderedCollection class itself. This can be a bit
confusing, because the title bar of the inspector names the class of the object
being inspected.
The inspector on OrderedCollection allows you to see the superclass, instance variables, method dictionary, and so on, of the OrderedCollection
class, as shown in Figure 15.5.
15.5
Every metaclass inherits from Class and Behavior
Every metaclass is a kind of a class (a class with a single instance), hence
inherits from Class. Class in turn inherits from its superclasses, ClassDescription and Behavior. Since everything in Pharo is an object, these classes
all inherit eventually from Object. We can see the complete picture in Figure
15.6.
Where is new defined?
To understand the importance of the fact that metaclasses inherit from Class
and Behavior, it helps to ask where new is defined and how it is found. When
297
Classes and metaclasses
Figure 15.6: Metaclasses inherit from Class and Behavior.
the message new is sent to a class, it is looked up in its metaclass chain and
ultimately in its superclasses Class, ClassDescription and Behavior as
shown in Figure 15.7.
The question Where is new defined? is crucial. new is first defined in the class
Behavior, and it can be redefined in its subclasses, including any of the metaclass of the classes we define, when this is necessary. Now when a message
new is sent to a class it is looked up, as usual, in the metaclass of this class,
continuing up the superclass chain right up to the class Behavior, if it has not
been redefined along the way.
Note that the result of sending SortedCollection new is an instance of SortedCollection and not of Behavior, even though the method is looked up in
the class Behavior! new always returns an instance of self, the class that
receives the message, even if it is implemented in another class.
SortedCollection new class
>>> SortedCollection "not Behavior!"
A common mistake is to look for new in the superclass
of the receiving class. The same holds for new:, the standard message to
create an object of a given size. For example, Array new: 4 creates an array
of 4 elements. You will not find this method defined in Array or any of its
superclasses. Instead you should look in Array class and its superclasses,
since that is where the lookup will start (See Figure 15.7).
Common mistake.
298
15.5
Every metaclass inherits from Class and Behavior
Figure 15.7: new is an ordinary message looked up in the metaclass chain.
Responsibilities of Behavior, ClassDescription, and Class
Behavior provides the minimum state necessary for objects that have instances, which includes a superclass link, a method dictionary and the class
format. The class format is an integer that encodes the pointer/non-pointer
distinction, compact/non-compact class distinction, and basic size of instances. Behavior inherits from Object, so it, and all of its subclasses, can
behave like objects.
Behavior is also the basic interface to the compiler. It provides methods for
creating a method dictionary, compiling methods, creating instances (i.e., new,
basicNew, new:, and basicNew:), manipulating the class hierarchy (i.e., superclass:, addSubclass:), accessing methods (i.e., selectors, allSelectors,
compiledMethodAt:), accessing instances and variables (i.e., allInstances,
instVarNames...), accessing the class hierarchy (i.e., superclass, subclasses)
and querying (i.e., hasMethods, includesSelector, canUnderstand:, inheritsFrom:, isVariable).
ClassDescription is an abstract class that provides facilities needed by its
two direct subclasses, Class and Metaclass. ClassDescription adds a number of facilities to the base provided by Behavior: named instance variables,
the categorization of methods into protocols, the maintenance of change sets
and the logging of changes, and most of the mechanisms needed for filing out
changes.
Class represents the common behaviour of all classes. It provides a class
name, compilation methods, method storage, and instance variables. It pro299
Classes and metaclasses
Figure 15.8: Every metaclass is a Metaclass.
vides a concrete representation for class variable names and shared pool
variables (addClassVarName:, addSharedPool:, initialize). Since a metaclass is a class for its sole instance (i.e., the non-meta class), all metaclasses
ultimately inherit from Class (as shown by Figure 15.9).
15.6
Every metaclass is an instance of Metaclass
One question left is since metaclasses are objects too, they should be instances
of another class, but which one? Metaclasses are objects too; they are instances of the class Metaclass as shown in Figure 15.8. The instances of class
Metaclass are the anonymous metaclasses, each of which has exactly one
instance, which is a class.
Metaclass represents common metaclass behaviour. It provides methods
for instance creation (subclassOf:), creating initialized instances of the
metaclass’s sole instance, initialization of class variables, metaclass instance,
method compilation, and class information (inheritance links, instance variables, etc.).
15.7
The metaclass of Metaclass is an instance of Meta-
class
The final question to be answered is: what is the class of Metaclass class?
The answer is simple: it is a metaclass, so it must be an instance of Metaclass,
just like all the other metaclasses in the system (see Figure 15.9).
300
15.7
The metaclass of Metaclass is an instance of Metaclass
Figure 15.9: All metaclasses are instances of the class Metaclass, even the
metaclass of Metaclass.
Figure 15.9 shows how all metaclasses are instances of Metaclass, including
the metaclass of Metaclass itself. If you compare Figures 15.8 and 15.9 you
will see how the metaclass hierarchy perfectly mirrors the class hierarchy, all
the way up to Object class.
The following examples show us how we can query the class hierarchy to
demonstrate that Figure 15.9 is correct. (Actually, you will see that we told a
white lie — Object class superclass --> ProtoObject class, not Class.
In Pharo, we must go one superclass higher to reach Class.)
Collection superclass
>>> Object
Collection class superclass
>>> Object class
Object class superclass superclass
>>> Class "NB: skip ProtoObject class"
Class superclass
>>> ClassDescription
ClassDescription superclass
>>> Behavior
Behavior superclass
>>> Object
Collection class class
>>> Metaclass
301
Classes and metaclasses
Object class class
>>> Metaclass
Behavior class class
>>> Metaclass
Metaclass class class
>>> Metaclass
Metaclass superclass
>>> ClassDescription
15.8
Chapter summary
This chapter gave an in-depth look into the uniform object model, and a more
thorough explanation of how classes are organized. If you get lost or confused,
you should always remember that message passing is the key: you look for the
method in the class of the receiver. This works on any receiver. If the method
is not found in the class of the receiver, it is looked up in its superclasses.
• Every class is an instance of a metaclass. Metaclasses are implicit. A
metaclass is created automatically when you create the class that is its
sole instance. A metaclass is simply a class whose unique instance is a
class.
• The metaclass hierarchy parallels the class hierarchy. Method lookup
for classes parallels method lookup for ordinary objects, and follows the
metaclass’s superclass chain.
• Every metaclass inherits from Class and Behavior. Every class is a
Class. Since metaclasses are classes too, they must also inherit from
Class. Behavior provides behaviour common to all entities that have
instances.
• Every metaclass is an instance of Metaclass. ClassDescription provides everything that is common to Class and Metaclass.
• The metaclass of Metaclass is an instance of Metaclass. The instance-of
relation forms a closed loop, so Metaclass class class is Metaclass.
302
CHAPTER
16
Reflection
Pharo is a reflective programming language. In a nutshell, this means that
programs are able to reflect on their own execution and structure. More technically, this means that the metaobjects of the runtime system can be reified
as ordinary objects, which can be queried and inspected. The metaobjects
in Pharo are classes, metaclasses, method dictionaries, compiled methods,
but also the run-time stack, processes, and so on. This form of reflection is
also called introspection, and is supported by many modern programming
languages.
Conversely, it is possible in Pharo to modify reified metaobjects and reflect
these changes back to the runtime system (see Figure 16.1). This is also called
Figure 16.1: Reification and reflection.
303
Reflection
intercession, and is supported mainly by dynamic programming languages,
and only to a very limited degree by static languages. So pay attention when
people say that Java is a reflective language, it is an introspective one not a
reflective one.
A program that manipulates other programs (or even itself) is a metaprogram.
For a programming language to be reflective, it should support both introspection and intercession. Introspection is the ability to examine the data
structures that define the language, such as objects, classes, methods and
the execution stack. Intercession is the ability to modify these structures, in
other words to change the language semantics and the behavior of a program
from within the program itself. Structural reflection is about examining and
modifying the structures of the run-time system, and behavioural reflection is
about modifying the interpretation of these structures.
In this chapter we will focus mainly on structural reflection. We will explore
many practical examples illustrating how Pharo supports introspection and
metaprogramming.
16.1
Introspection
Using the inspector, you can look at an object, change the values of its instance variables, and even send messages to it.
Evaluate the following code in a playground:
w := GTPlayground openLabel: 'My Playground'.
w inspect
This will open a second playground and an inspector. The inspector shows the
internal state of this new playground, listing its instance variables on the left
(borderColor, borderWidth, bounds...) and the value of the selected instance
variable on the right. The bounds instance variable represents the precise area
occupied by the playground.
Now choose the inspector and click the playground area of the inspector
which has a comment on top and type self bounds: (Rectangle origin:
10@10 corner: 300@300 ) in it of select asshown in Figure 16.2 and then Do It
like you do with a code of a Playground.
Immediately you will see the Playground we created change and resize itself.
Accessing instance variables
How does the inspector work? In Pharo, all instance variables are protected.
In theory, it is impossible to access them from another object if the class
doesn’t define any accessor. In practice, the inspector can access instance
variables without needing accessors, because it uses the reflective abilities
of Pharo. Classes define instance variables either by name or by numeric
304
16.1
Introspection
Figure 16.2: Inspecting a Workspace.
indices. The inspector uses methods defined by the Object class to access
them: instVarAt: index and instVarNamed: aString can be used to get
the value of the instance variable at position index or identified by aString,
respectively. Similarly, to assign new values to these instance variables, it uses
instVarAt:put: and instVarNamed:put:.
For instance, you can change the value of the w binding of the first workspace
by evaluating:
w instVarNamed:'bounds' put: (Rectangle origin: 10@10 corner: 500@500).
Important Caveat: Although these methods are useful for building development tools, using them to develop conventional applications is a bad idea: these
reflective methods break the encapsulation boundary of your objects and can
therefore make your code much harder to understand and maintain.
Both instVarAt: and instVarAt:put: are primitive methods, meaning that
they are implemented as primitive operations of the Pharo virtual machine. If
you consult the code of these methods, you will see the special pragma syntax
<primitive: N> where N is an integer.
Object >> instVarAt: index
"Primitive. Answer a fixed variable in an object. ..."
<primitive: 173 error: ec>
self primitiveFailed
305
Reflection
Figure 16.3: Displaying all instance variables of a GTPlayground.
Any Pharo code after the primitive declaration is executed only if the primitive fails. This also allows the debugger to be started on primitive methods.
In this specific case, there is no way to implement this method, so the whole
method just fails.
Other methods are implemented on the VM for faster execution. For example
some arithmetic operations on SmallInteger :
* aNumber
"Primitive. Multiply the receiver by the argument and answer with the
result if it is a SmallInteger. Fail if the argument or the result
is not a
SmallInteger. Essential. No Lookup. See Object documentation
whatIsAPrimitive."
<primitive: 9>
^ super * aNumber
If this primitive fails, for example if the VM does not handle the type of the
argument, the Pharo code is executed. Although it is possible to modify the
code of primitive methods, beware that this can be risky business for the
stability of your Pharo system.
Figure 16.3 shows how to display the values of the instance variables of an
arbitrary instance (w) of class GTPlayground. The method allInstVarNames
returns all the names of the instance variables of a given class.
306
16.1
Introspection
GTPlayground allInstVarNames
>>>
#(#registry #suspendAll #suspendedAnnouncemets #logger #pane #title
#titleIcon #transformation #actions #condition #implicitNotNil
#dynamicActionsBlock #color #customValidation #shouldValidate
#acceptsSelection #parentPrototype #registeredAnnouncers
#updateActions #selectionActions #selectionDynamicActionsBlock
#implicitAllNil #rawSelectionTransmissions #statusPane #sourceLink
#initializationBlock #cachedDisplayedValue #labelActionBlock
#portChangeActions #wantsSteps #stepTime #stepCondition
#presentations #arrangement)
w := GTPlayground someInstance.
w class allInstVarNames collect: [:each | each -> (w instVarNamed:
each)]
In the same spirit, it is possible to gather instances that have specific properties iterating over instances of a class using an iterator such as select:. For
instance, to get all objects who are directly included in the world morph (the
main root of the graphical displayed elements), try this expression:
Morph allSubInstances
select: [ :each |
| own |
own := (each instVarNamed: 'owner').
own isNotNil and: [ own isWorldMorph ]]
Querying classes and interfaces
The development tools in Pharo (system browser, debugger, inspector...) all
use the reflective features we have seen so far.
Here are a few other messages that might be useful to build development
tools:
isKindOf: aClass returns true if the receiver is instance of aClass or of one
of its superclasses. For instance:
1.5 class
>>> BoxedFloat64
1.5 isKindOf: Float
>>> true
1.5 isKindOf: Number
>>> true
1.5 isKindOf: Integer
>>> false
respondsTo: aSymbol returns true if the receiver has a method whose selector is aSymbol. For instance:
307
Reflection
1.5 respondsTo: #floor
>>> true "since Number implements floor"
1.5 floor
>>> 1
Exception respondsTo: #,
>>> true "exception classes can be grouped"
Although these features are especially useful for implementing development tools, they are normally not appropriate for typical
applications. Asking an object for its class, or querying it to discover which
messages it understands, are typical signs of design problems, since they
violate the principle of encapsulation. Development tools, however, are not
normal applications, since their domain is that of software itself. As such these
tools have a right to dig deep into the internal details of code.
Important Caveat:
Code metrics
Let’s see how we can use Pharo’s introspection features to quickly extract
some code metrics. Code metrics measure such aspects as the depth of the
inheritance hierarchy, the number of direct or indirect subclasses, the number
of methods or of instance variables in each class, or the number of locally
defined methods or instance variables. Here are a few metrics for the class
Morph, which is the superclass of all graphical objects in Pharo, revealing that
it is a huge class, and that it is at the root of a huge hierarchy. Maybe it needs
some refactoring!
"inheritance depth"
Morph allSuperclasses size.
>>> 2
"number of methods"
Morph allSelectors size.
>>> 1304
"number of instance variables"
Morph allInstVarNames size.
>>> 6
"number of new methods"
Morph selectors size.
>>> 896
"number of new variables"
Morph instVarNames size.
>>> 6
"direct subclasses"
Morph subclasses size.
>>> 63
308
16.2 Browsing code
"total subclasses"
Morph allSubclasses size.
>>> 376
"total lines of code!"
Morph linesOfCode.
>>> 4964
One of the most interesting metrics in the domain of object-oriented languages is the number of methods that extend methods inherited from the
superclass. This informs us about the relation between the class and its superclasses. In the next sections we will see how to exploit our knowledge of the
runtime structure to answer such questions.
16.2
Browsing code
In Pharo, everything is an object. In particular, classes are objects that provide
useful features for navigating through their instances. Most of the messages
we will look at now are implemented in Behavior, so they are understood by
all classes.
For example, you can obtain a random instance of a given class by sending it
the message someInstance.
Point someInstance
>>> 0@0
You can also gather all the instances with allInstances, or the number of
active instances in memory with instanceCount.
ByteString allInstances
>>> #('collection' 'position' ...)
ByteString instanceCount
>>> 104565
String allSubInstances size
>>> 101675
These features can be very useful when debugging an application, because
you can ask a class to enumerate those of its methods exhibiting specific
properties. Here are some more interesting and useful methods for code
discovery through reflection.
whichSelectorsAccess: returns the list of all selectors of methods that read
or write the instance variable named by the argument
whichSelectorsStoreInto: returns the selectors of methods that modify the
value of an instance variable
whichSelectorsReferTo: returns the selectors of methods that send a given
message
309
Reflection
Point whichSelectorsAccess: 'x'
>>> #(#degrees #grid: #roundTo: #nearestPointAlongLineFrom:to: ...)
Point whichSelectorsStoreInto: 'x'
>>> #(#bitShiftPoint: #setR:degrees: #setX:setY: #fromSton:)
Point whichSelectorsReferTo: #+
>>> an OrderedCollection(#degrees #reflectedAbout: #grid: ...)
The following messages take inheritance into account:
whichClassIncludesSelector: returns the superclass that implements the
given message
unreferencedInstanceVariables returns the list of instance variables that
are neither used in the receiver class nor any of its subclasses
Rectangle whichClassIncludesSelector: #inspect
>>> Object
Rectangle unreferencedInstanceVariables
>>> #()
SystemNavigation is a facade that supports various useful methods for querying and browsing the source code of the system. SystemNavigation default
returns an instance you can use to navigate the system. For example:
SystemNavigation default allClassesImplementing: #yourself
>>> {Object}
The following messages should also be self-explanatory:
SystemNavigation default allSentMessages size
>>>370
(SystemNavigation default allUnsentMessagesIn: Object selectors) size
>>> 31
SystemNavigation default allUnimplementedCalls size
>>> 521
Note that messages implemented but not sent are not necessarily useless,
since they may be sent implicitly (e.g., using perform:). Messages sent but
not implemented, however, are more problematic, because the methods
sending these messages will fail at runtime. They may be a sign of unfinished
implementation, obsolete APIs, or missing libraries.
Point allCallsOn returns all messages sent explicitly to Point as a receiver.
All these features are integrated into the programming environment of Pharo,
in particular the code browsers. As we mentioned before, there are convenient keyboard shortcuts for browsing all implementors (CMD-b CMD-m) and
browsing senders (CMD-b CMD-n) of a given message. What is perhaps not so
well known is that there are many such pre-packaged queries implemented
310
16.3
Classes, method dictionaries and methods
Figure 16.4: Browse all implementations of ifTrue:.
as methods of the SystemNavigation class in the browsing protocol. For example, you can programmatically browse all implementors of the message
ifTrue: by evaluating:
SystemNavigation default browseAllImplementorsOf: #ifTrue:
Particularly useful are the methods browseAllSelect: and browseMethodsWithSourceString:matchCase:. Here are two different ways to browse all
methods in the system that perform super sends (the first way is rather brute
force, the second way is better and eliminates some false positives):
SystemNavigation default browseMethodsWithSourceString: 'super'
matchCase: true.
SystemNavigation default browseAllSelect: [:method | method
sendsToSuper ].
16.3
Classes, method dictionaries and methods
Since classes are objects, we can inspect or explore them just like any other
object.
Evaluate Point inspect.
In Figure 16.5, the inspector shows the structure of class Point. You can
see that the class stores its methods in a dictionary, indexing them by their
selector. The selector #* points to the decompiled bytecode of Point>>*.
Let us consider the relationship between classes and methods. In Figure 16.6
we see that classes and metaclasses have the common superclass Behavior.
This is where new is defined, amongst other key methods for classes. Every
class has a method dictionary, which maps method selectors to compiled
311
Reflection
Figure 16.5: Inspector on class Point and the bytecode of its #* method.
Figure 16.6: Classes, method dictionaries and compiled methods
methods. Each compiled method knows the class in which it is installed. In
Figure 16.5 we can even see that this is stored in an association in literal6.
We can exploit the relationships between classes and methods to pose queries
about the system. For example, to discover which methods are newly introduced in a given class, i.e., do not override superclass methods, we can
navigate from the class to the method dictionary as follows:
[:aClass| aClass methodDict keys select: [:aMethod |
(aClass superclass canUnderstand: aMethod) not ]] value: SmallInteger
>>> an IdentitySet(#threeDigitName #printStringBase:nDigits: ...)
312
16.4
Browsing environments
A compiled method does not simply store the bytecode of a method. It is also
an object that provides numerous useful methods for querying the system.
One such method is isAbstract (which tells if the method sends subclassResponsibility). We can use it to identify all the abstract methods of an
abstract class.
[:aClass| aClass methodDict keys select: [:aMethod |
(aClass>>aMethod) isAbstract ]] value: Number
>>> an IdentitySet(#storeOn:base: #printOn:base: #+ #- #* #/ ...)
Note that this code sends the >> message to a class to obtain the compiled
method for a given selector.
To browse the super-sends within a given hierarchy, for example within the
Collections hierarchy, we can pose a more sophisticated query:
class := Collection.
SystemNavigation default
browseMessageList: (class withAllSubclasses gather: [:each |
each methodDict associations
select: [:assoc | assoc value sendsToSuper]
thenCollect: [:assoc | RGMethodDefinition realClass: each
selector: assoc key]])
name: 'Supersends of ', class name, ' and its subclasses'
Note how we navigate from classes to method dictionaries to compiled methods to identify the methods we are interested in. A RGMethodDefinition is a
lightweight proxy for a compiled method that is used by many tools. There
is a convenience method CompiledMethod>>methodReference to return the
method reference for a compiled method.
(Object>>#=) methodReference selector
>>> #=
16.4
Browsing environments
Although SystemNavigation offers some useful ways to programmatically
query and browse system code, there are more ways. The Browser, which is
integrated into Pharo, allows us to restrict the environment in which a search
is to perform.
Suppose we are interested to discover which classes refer to the class Point
but only in its own package.
Open a browser on the class Point.
Action-click on the top level package Kernel in the package pane and select
Browse scoped. A new browser opens, showing only the package Kernel
and all classes within this package (and some classes which have extension
methods from this package). Now, in this browser, select again the class Point,
313
Reflection
Action-click on the class name and select Analyse > Class refs. This will
show all methods that have references to the class Point but only those from
the package Kernel. Compare this result with the search from a Browser
without restricted scope.
This scope is what we call a Browsing Environment (class RBBrowserEnvironment). All other searches, like senders of a method or implementors of a method
from within this browser are restricted to this environments too.
Browser environments can also be created programmatically. Here, for example, we create a new RBBrowserEnvironment for Collection and its subclasses, select the super-sending methods, and browse the resulting environment.
((RBBrowserEnvironment new forClasses: (Collection withAllSubclasses))
selectMethods: [:method | method sendsToSuper])
browse.
Note how this is considerably more compact than the earlier, equivalent
example using SystemNavigation.
Finally, we can find just those methods that send a different super message
programmatically as follows:
((RBBrowserEnvironment new forClasses: (Collection withAllSubclasses))
selectMethods: [:method |
method sendsToSuper
and: [(method parseTree superMessages includes: method selector)
not]])
browse
Here we ask each compiled method for its (Refactoring Browser) parse tree,
in order to find out whether the super messages differ from the method’s
selector. Have a look at the querying protocol of the class RBProgramNode to
see some the things we can ask of parse trees.
Instead of browsing the environment in a System Browser, we can spawn a
MessageBrower from the list of all methods in this environment.
MessageBrowser browse: ((RBBrowserEnvironment new forClasses:
(Collection withAllSubclasses))
selectMethods: [:method |
method sendsToSuper
and: [(method parseTree superMessages includes: method selector)
not]]) methods
title: 'Collection methods sending different super'
In Figure 16.7 we can see that 5 such methods have been found within the
Collection hierarchy, including Collection>>printNameOn:, which sends
super printOn:.
314
16.5
Accessing the run-time context
Figure 16.7: Finding methods
16.5
Accessing the run-time context
We have seen how Pharo’s reflective capabilities let us query and explore
objects, classes and methods. But what about the run-time environment?
Method contexts
In fact, the run-time context of an executing method is in the virtual machine
— it is not in the image at all! On the other hand, the debugger obviously has
access to this information, and we can happily explore the run-time context,
just like any other object. How is this possible?
Actually, there is nothing magical about the debugger. The secret is the
pseudo-variable thisContext, which we have encountered only in passing
before. Whenever thisContext is referred to in a running method, the entire
run-time context of that method is reified and made available to the image as
a series of chained Context objects.
We can easily experiment with this mechanism ourselves.
Change the definition of Integer>>factorial by inserting the expression
thisContext inspect. self halt. as shown below:
Integer>>factorial
"Answer the factorial of the receiver."
self = 0 ifTrue: [thisContext inspect. self halt. ^ 1].
self > 0 ifTrue: [^ self * (self - 1) factorial].
self error: 'Not valid for negative integers'
315
Reflection
Figure 16.8: Inspecting thisContext.
Now evaluate 3 factorial in a workspace. You should obtain both a debugger window and an inspector, as shown in Figure 16.8.
Inspecting thisContext gives you full access to the current execution context,
the stack, the local tempories and arguments, the senders chain and the
receiver. Welcome to the poor man’s debugger! If you now browse the class
of the explored object (i.e., by evaluating self browse in the bottom pane of
the inspector) you will discover that it is an instance of the class Context, as is
each sender in the chain.
thisContext is not intended to be used for day-to-day programming, but it is
essential for implementing tools like debuggers, and for accessing information
about the call stack. You can evaluate the following expression to discover
which methods make use of thisContext:
SystemNavigation default browseMethodsWithSourceString: 'thisContext'
matchCase: true
As it turns out, one of the most common applications is to discover the sender
of a message. Here is a typical application:
subclassResponsibility
"This message sets up a framework for the behavior of the class'
316
16.5
Accessing the run-time context
subclasses.
Announce that the subclass should have implemented this message."
SubclassResponsibility signalFor: thisContext sender selector
By convention, methods that send self subclassResponsibility are considered to be abstract. But how does Object>>subclassResponsibility provide
a useful error message indicating which abstract method has been invoked?
Very simply, by asking thisContext for the sender.
Intelligent breakpoints
The Pharo way to set a breakpoint is to evaluate self halt at an interesting
point in a method. This will cause thisContext to be reified, and a debugger
window will open at the breakpoint. Unfortunately this poses problems for
methods that are intensively used in the system.
Suppose, for instance, that we want to explore the execution of Morph>>openInWorld. Setting a breakpoint in this method is problematic.
Pay attention the following experiment will break everything! Take a fresh
image and set the following breakpoint:
Morph >> openInWorld
"Add this morph to the world."
self halt.
self openInWorld: self currentWorld
Notice how your image immediately freezes as soon as you try to open any
new Morph (Menu/Window/...)! We do not even get a debugger window. The
problem is clear once we understand that 1) Morph>>openInWorld is used by
many parts of the system, so the breakpoint is triggered very soon after we
interact with the user interface, but 2) the debugger itself sends openInWorld as
soon as it opens a window, preventing the debugger from opening! What we
need is a way to conditionally halt only if we are in a context of interest. This is
exactly what Object>>haltIf: offers.
Suppose now that we only want to halt if openInWorld is sent from, say, the
context of MorphTest>>testOpenInWorld.
Fire up a fresh image again, and set the following breakpoint:
Morph>>openInWorld
"Add this morph to the world."
self haltIf: #testOpenInWorld.
self openInWorld: self currentWorld
This time the image does not freeze. Try running the MorphTest.
MorphTest run:#testOpenInWorld.
317
Reflection
How does this work? Let’s have a look at Object>>haltIf:. It first calls if:
with the condition to the Exception class Halt. This method itself will check
if the condition is a symbol, which is true in this case and finally calls
Object >> haltIfCallChainContains: aSelector
| cntxt |
cntxt := thisContext.
[cntxt sender isNil] whileFalse: [
cntxt := cntxt sender.
(cntxt selector = aSelector) ifTrue: [self signal]].
Starting from thisContext, haltIfCallChainContains: goes up through the
execution stack, checking if the name of the calling method is the same as the
one passed as parameter. If this is the case, then it signals itself, the exception
which, by default, summons the debugger.
It is also possible to supply a boolean or a boolean block as an argument to
haltIf:, but these cases are straightforward and do not make use of thisContext.
16.6
Intercepting messages not understood
So far we have used Pharo’s reflective features mainly to query and explore
objects, classes, methods and the run-time stack. Now we will look at how to
use our knowledge of its system structure to intercept messages and modify
behaviour at run time.
When an object receives a message, it first looks in the method dictionary of
its class for a corresponding method to respond to the message. If no such
method exists, it will continue looking up the class hierarchy, until it reaches
Object. If still no method is found for that message, the object will send itself
the message doesNotUnderstand: with the message selector as its argument.
The process then starts all over again, until Object>>doesNotUnderstand: is
found, and the debugger is launched.
But what if doesNotUnderstand: is overridden by one of the subclasses of Object in the lookup path? As it turns out, this is a convenient way of realizing
certain kinds of very dynamic behaviour. An object that does not understand a
message can, by overriding doesNotUnderstand:, fall back to an alternative
strategy for responding to that message.
Two very common applications of this technique are 1) to implement lightweight
proxies for objects, and 2) to dynamically compile or load missing code.
Lightweight proxies
In the first case, we introduce a minimal object to act as a proxy for an existing
object. Since the proxy will implement virtually no methods of its own, any
318
16.6
Intercepting messages not understood
message sent to it will be trapped by doesNotUnderstand:. By implementing
this message, the proxy can then take special action before delegating the
message to the real subject it is the proxy for.
Let us have a look at how this may be implemented.
We define a LoggingProxy as follows:
ProtoObject subclass: #LoggingProxy
instanceVariableNames: 'subject invocationCount'
classVariableNames: ''
package: 'PBE-Reflection'
Note that we subclass ProtoObject rather than Object because we do not
want our proxy to inherit around 400 methods (!) from Object.
Object methodDict size
>>> 397
Our proxy has two instance variables: the subject it is a proxy for, and a
count of the number of messages it has intercepted. We initialize the two
instance variables and we provide an accessor for the message count. Initially
the subject variable points to the proxy object itself.
LoggingProxy >> initialize
invocationCount := 0.
subject := self.
LoggingProxy >> invocationCount
^ invocationCount
We simply intercept all messages not understood, print them to the Transcript, update the message count, and forward the message to the real subject.
LoggingProxy >> doesNotUnderstand: aMessage
Transcript show: 'performing ', aMessage printString; cr.
invocationCount := invocationCount + 1.
^ aMessage sendTo: subject
Here comes a bit of magic. We create a new Point object and a new LoggingProxy object, and then we tell the proxy to become: the point object:
point := 1@2.
LoggingProxy new become: point.
This has the effect of swapping all references in the image to the point to now
refer to the proxy, and vice versa. Most importantly, the proxy’s subject
instance variable will now refer to the point!
point invocationCount
>>> 0
point + (3@4)
>>> 4@6
319
Reflection
point invocationCount
>>> 1
This works nicely in most cases, but there are some shortcomings:
point class
>>> LoggingProxy
Actually the method class is implemented in ProtoObject, but even if it
were implemented in Object, which LoggingProxy does not inherit from, it
isn’t actually send to the LoggingProxy or its subject. The message is directly
answered by the virtual machine. yourself is also never truly sent.
Other messages that may be directly interpreted by the VM, depending on the
receiver, include:
+- < > <= >= = ~= * / \ == @ bitShift: // bitAnd: bitOr: at: at:put:
size next nextPut: atEnd blockCopy: value value: do: new new: x y.
Selectors that are never sent, because they are inlined by the compiler and
transformed to comparison and jump bytecodes:
ifTrue: ifFalse: ifTrue:ifFalse: ifFalse:ifTrue: and: or: whileFalse: whileTrue: whileFalse whileTrue to:do: to:by:do: caseOf:
caseOf:otherwise: ifNil: ifNotNil: ifNil:ifNotNil: ifNotNil:ifNil:
Attempts to send these messages to non-boolean normally results in an exception from the VM as it can not use the inlined dispatching for non-boolean
receivers. You can intercept this and define the proper behavior by overriding mustBeBoolean in the receiver or by catching the NonBooleanReceiver
exception.
Even if we can ignore such special message sends, there is another fundamental problem which cannot be overcome by this approach: self-sends cannot
be intercepted:
point := 1@2.
LoggingProxy new become: point.
point invocationCount
>>> 0
point rectangle: (3@4)
>>> 1@2 corner: 3@4
point invocationCount
>>> 1
Our proxy has been cheated out of two self-sends in the rect: method:
Point >> rect: aPoint
^ Rectangle origin: (self min: aPoint) corner: (self max: aPoint)
Although messages can be intercepted by proxies using this technique, one
should be aware of the inherent limitations of using a proxy. In Section 16.7
we will see another, more general approach for intercepting messages.
320
16.6
Intercepting messages not understood
Generating missing methods
The other most common application of intercepting not understood messages
is to dynamically load or generate the missing methods. Consider a very large
library of classes with many methods. Instead of loading the entire library, we
could load a stub for each class in the library. The stubs know where to find
the source code of all their methods. The stubs simply trap all messages not
understood, and dynamically load the missing methods on demand. At some
point, this behaviour can be deactivated, and the loaded code can be saved as
the minimal necessary subset for the client application.
Let us look at a simple variant of this technique where we have a class that
automatically adds accessors for its instance variables on demand:
DynamicAcccessors >> doesNotUnderstand: aMessage
| messageName |
messageName := aMessage selector asString.
(self class instVarNames includes: messageName)
ifTrue: [
self class compile: messageName, String cr, ' ^ ', messageName.
^ aMessage sendTo: self ].
^ super doesNotUnderstand: aMessage
Any message not understood is trapped here. If an instance variable with the
same name as the message sent exists, then we ask our class to compile an
accessor for that instance variables and we re-send the message.
Suppose the class DynamicAccessors has an (uninitialized) instance variable
x but no pre-defined accessor. Then the following will generate the accessor
dynamically and retrieve the value:
myDA := DynamicAccessors new.
myDA x
>>> nil
Let us step through what happens the first time the message x is sent to our
object (see Figure 16.9).
(1) We send x to myDA, (2) the message is looked up in the class, and (3) not
found in the class hierarchy. (4) This causes self doesNotUnderstand: #x
to be sent back to the object, (5) triggering a new lookup. This time doesNotUnderstand: is found immediately in DynamicAccessors, (6) which asks its
class to compile the string 'x ^ x'. The compile method is looked up (7), and
(8) finally found in Behavior, which (9-10) adds the new compiled method to
the method dictionary of DynamicAccessors. Finally, (11-13) the message is
resent, and this time it is found.
The same technique can be used to generate setters for instance variables, or
other kinds of boilerplate code, such as visiting methods for a Visitor.
Note the use of Object>>perform: in step (12) which can be used to send
messages that are composed at run-time:
321
Reflection
Figure 16.9: Dynamically creating accessors.
5 perform: #factorial
>>> 120
6 perform: ('fac', 'torial') asSymbol
>>> 720
4 perform: #max: withArguments: (Array with: 6)
>>> 6
16.7
Objects as method wrappers
We have already seen that compiled methods are ordinary objects in Pharo,
and they support a number of methods that allow the programmer to query
the runtime system. What is perhaps a bit more surprising, is that any object
can play the role of a compiled method. All it has to do is respond to the
method run:with:in: and a few other important messages.
Define an empty class Demo. Evaluate Demo new answer42 and notice how the
usual Message Not Understood error is raised.
Now we will install a plain object in the method dictionary of our Demo class.
Evaluate Demo methodDict at: #answer42 put: ObjectsAsMethodsExample new.
Now try again to print the result of Demo new answer42. This time we get the
answer 42.
If we take look at the class ObjectsAsMethodsExample we will find the following methods:
322
16.7
Objects as method wrappers
answer42
^42
run: oldSelector with: arguments in: aReceiver
^self perform: oldSelector withArguments: arguments
When our Demo instance receives the message answer42, method lookup proceeds as usual, however the virtual machine will detect that in place of a
compiled method, an ordinary Pharo object is trying to play this role. The
VM will then send this object a new message run:with:in: with the original
method selector, arguments and receiver as arguments. Since ObjectsAsMethodsExample implements this method, it intercepts the message and delegates
it to itself.
We can now remove the fake method as follows:
Demo methodDict removeKey: #answer42 ifAbsent: []
If we take a closer look at ObjectsAsMethodsExample, we will see that its
superclass also implements some methods like flushcache, methodClass:
and selector:, but they are all empty. These messages may be sent to a
compiled method, so they need to be implemented by an object pretending
to be a compiled method. (flushcache is the most important method to
be implemented; others may be required by some tools and depending on
whether the method is installed using Behavior>>addSelector:withMethod:
or directly using MethodDictionary>>at:put:.)
Using method wrappers to perform test coverage
Method wrappers are a well-known technique for intercepting messages. In
the original implementation(http://www.squeaksource.com/MethodWrappers.html), a method wrapper is an instance of a subclass of CompiledMethod.
When installed, a method wrapper can perform special actions before or after invoking the original method. When uninstalled, the original method is
returned to its rightful position in the method dictionary.
In Pharo, method wrappers can be implemented more easily by implementing
run:with:in: instead of by subclassing CompiledMethod. In fact, there exists
a lightweight implementation of objects as method wrappers(http://www.squeaksource.com/ObjectsAsMethodsWrap.html), but it is not part of standard Pharo
at the time of this writing.
Nevertheless, the Pharo Test Runner uses precisely this technique to evaluate
test coverage. Let’s have a quick look at how it works.
The entry point for test coverage is the method TestRunner>>runCoverage:
TestRunner >> runCoverage
| packages methods |
... "identify methods to check for coverage"
self collectCoverageFor: methods
323
Reflection
The method TestRunner>>collectCoverageFor: clearly illustrates the coverage checking algorithm:
TestRunner >> collectCoverageFor: methods
| wrappers suite |
wrappers := methods collect: [ :each | TestCoverage on: each ].
suite := self
resetResult;
suiteForAllSelected.
[ wrappers do: [ :each | each install ].
[ self runSuite: suite ] ensure: [ wrappers do: [ :each | each
uninstall ] ] ] valueUnpreemptively.
wrappers := wrappers reject: [:each | each hasRun].
wrappers := wrappers collect: [:each | each reference].
wrappers isEmpty
ifTrue: [ UIManager default inform: 'Congratulations. Your tests
cover all code under analysis.' ]
ifFalse: ...
A wrapper is created for each method to be checked, and each wrapper is
installed. The tests are run, and all wrappers are uninstalled. Finally the user
obtains feedback concerning the methods that have not been covered.
How does the wrapper itself work? The TestCoverage wrapper has three
instance variables, hasRun, reference and method. They are initialized as
follows:
TestCoverage class >> on: aMethodReference
^ self new initializeOn: aMethodReference
TestCoverage
hasRun :=
reference
method :=
>> initializeOn: aMethodReference
false.
:= aMethodReference.
reference compiledMethod
The install and uninstall methods simply update the method dictionary in the
obvious way:
TestCoverage >> install
reference actualClass methodDict
at: reference selector
put: self
TestCoverage >> uninstall
reference actualClass methodDict
at: reference selector
put: method
The run:with:in: method simply updates the hasRun variable, uninstalls the
wrapper (since coverage has been verified), and resends the message to the
324
16.8
Pragmas
original method.
run: aSelector with: anArray in: aReceiver
self mark; uninstall.
^ aReceiver withArgs: anArray executeMethod: method
mark
hasRun := true
Take a look at ProtoObject>>withArgs:executeMethod: to see how a method
displaced from its method dictionary can be invoked.
That’s all there is to it!
Method wrappers can be used to perform any kind of suitable behaviour
before or after the normal operation of a method. Typical applications are
instrumentation (collecting statistics about the calling patterns of methods),
checking optional pre- and post-conditions, and memoization (optionally
cacheing computed values of methods).
16.8
Pragmas
A pragma is an annotation that specifies data about a program, but is not
involved in the execution of the program. Pragmas have no direct effect on
the operation of the method they annotate. Pragmas have a number of uses,
among them:
Information for the compiler: pragmas can be used by the compiler to make a
method call a primitive function. This function has to be defined by the virtual
machine or by an external plug-in.
Runtime processing: Some pragmas are available to be examined at runtime.
Pragmas can be applied to a program’s method declarations only. A method
may declare one or more pragmas, and the pragmas have to be declared prior
any Smalltalk statement. Each pragma is in effect a static message send with
literal arguments.
We briefly saw pragmas when we introduced primitives earlier in this chapter.
A primitive is nothing more than a pragma declaration. Consider <primitive: 173 error:ec> as contained in instVarAt:. The pragma’s selector
is primitive:error: and its arguments is an immediate literal value, 173.
The variable ec is an error code, filled by the VM in case the execution of the
implementation on the VM side failed.
The compiler is probably the bigger user of pragmas. SUnit is another tool
that makes use of annotations. SUnit is able to estimate the coverage of an
application from a test unit. One may want to exclude some methods from the
coverage. This is the case of the documentation method in SplitJointTest
class:
325
Reflection
SplitJointTest class >> documentation
<ignoreForCoverage>
"self showDocumentation"
^ 'This package provides function.... "
By simply annotating a method with the pragma <ignoreForCoverage> one
can control the scope of the coverage.
As instances of the class Pragma, pragmas are first class objects. A compiled
method answers to the message pragmas. This method returns an array of
pragmas.
(SplitJoinTest class >> #showDocumentation) pragmas.
>>> an Array(<ignoreForCoverage>)
(Float>>#+) pragmas
>>> an Array(<primitive: 41>)
Methods defining a particular query may be retrieved from a class. The class
side of SplitJoinTest contains some methods annotated with <ignoreForCoverage>:
Pragma allNamed: #ignoreForCoverage in: SplitJoinTest class
>>> an Array(<ignoreForCoverage> <ignoreForCoverage>
<ignoreForCoverage>)
A variant of allNamed:in: may be found on the class side of Pragma.
A pragma knows in which method it is defined (using method), the name of
the method (selector), the class that contains the method (methodClass),
its number of arguments (numArgs), about the literals the pragma has for
arguments (hasLiteral: and hasLiteralSuchThat:).
16.9
Chapter summary
Reflection refers to the ability to query, examine and even modify the metaobjects of the runtime system as ordinary objects.
• The Inspector uses instVarAt: and related methods to view private
instance variables of objects.
• Send Behavior>>allInstances to query instances of a class.
• The messages class, isKindOf:, respondsTo: etc. are useful for gathering metrics or building development tools, but they should be avoided in
regular applications: they violate the encapsulation of objects and make
your code harder to understand and maintain.
• SystemNavigation is a utility class holding many useful queries for
navigation and browsing the class hierarchy. For example, use SystemNavigation default browseMethodsWithSourceString: 'pharo'
326
16.9
Chapter summary
matchCase:true. to find and browse all methods with a given source
string. (Slow, but thorough!)
• Every Pharo class points to an instance of MethodDictionary which
maps selectors to instances of CompiledMethod. A compiled method
knows its class, closing the loop.
• RGMethodDefinition is a leightweight proxy for a compiled method,
providing additional convenience methods, and used by many Pharo
tools.
• RBBrowserEnvironment, part of the Refactoring Browser infrastructure,
offers a more refined interface than SystemNavigation for querying the
system, since the result of a query can be used as a the scope of a new
query. Both GUI and programmatic interfaces are available.
• thisContext is a pseudo-variable that reifies the runtime stack of the
virtual machine. It is mainly used by the debugger to dynamically
construct an interactive view of the stack. It is also especially useful for
dynamically determining the sender of a message.
• Intelligent breakpoints can be set using haltIf:, taking a method selector as its argument. haltIf: halts only if the named method occurs as a
sender in the run-time stack.
• A common way to intercept messages sent to a given target is to use
a minimal object as a proxy for that target. The proxy implements as
few methods as possible, and traps all message sends by implementing
doesNotunderstand:. It can then perform some additional action and
then forward the message to the original target.
• Send become: to swap the references of two objects, such as a proxy and
its target.
• Beware, some messages, like class and yourself are never really sent,
but are interpreted by the VM. Others, like +, - and ifTrue: may be
directly interpreted or inlined by the VM depending on the receiver.
• Another typical use for overriding doesNotUnderstand: is to lazily load
or compile missing methods.
• doesNotUnderstand: cannot trap self-sends.
• A more rigorous way to intercept messages is to use an object as a
method wrapper. Such an object is installed in a method dictionary
in place of a compiled method. It should implement run:with:in:
which is sent by the VM when it detects an ordinary object instead of a
compiled method in the method dictionary. This technique is used by
the SUnit Test Runner to collect coverage data.
327
CHAPTER
17
Regular expressions in Pharo
Regular expressions are widely used in many scripting languages such as Perl,
Python and Ruby. They are useful to identify strings that match a certain
pattern, to check that input conforms to an expected format, and to rewrite
strings to new formats. Pharo also supports regular expressions due to the
Regex package contributed by Vassili Bykov. Regex is installed by default
in Pharo. If you are using an older image that does not include Regex the
Regex package, you can install it yourself from SqueakSource (http://www.
squeaksource.com/Regex.html).
A regular expression1 is a pattern that matches a set of strings. For example, the regular expression 'h.*o' will match the strings 'ho', 'hiho' and
'hello', but it will not match 'hi' or 'yo'. We can see this in Pharo as follows:
'ho' matchesRegex: 'h.*o'
>>> true
'hiho' matchesRegex: 'h.*o'
>>> true
'hello' matchesRegex: 'h.*o'
>>> true
'hi' matchesRegex: 'h.*o'
>>> false
'yo' matchesRegex: 'h.*o'
>>> false
1 http://en.wikipedia.org/wiki/Regular_expression
329
Regular expressions in Pharo
In this chapter we will start with a small tutorial example in which we will
develop a couple of classes to generate a very simple site map for a web site.
We will use regular expressions
1. to identify HTML files,
2. to strip the full path name of a file down to just the file name,
3. to extract the title of each web page for the site map, and
4. to generate a relative path from the root directory of the web site to the
HTML files it contains.
After we complete the tutorial example, we will provide a more complete
description of the Regex package, based largely on Vassili Bykov’s documentation provided in the package. (The original documentation can be found on
the class side of RxParser.)
17.1
Tutorial example — generating a site map
Our job is to write a simple application that will generate a site map for a
web site that we have stored locally on our hard drive. The site map will
contain links to each of the HTML files in the web site, using the title of the
document as the text of the link. Furthermore, links will be indented to reflect
the directory structure of the web site.
Accessing the web directory
To do If you do not have a web site on your machine, copy a few HTML
files to a local directory to serve as a test bed.
We will develop two classes, WebDir and WebPage, to represent directories and
web pages. The idea is to create an instance of WebDir which will point to the
root directory containing our web site. When we send it the message makeToc,
it will walk through the files and directories inside it to build up the site map.
It will then create a new file, called toc.html, containing links to all the pages
in the web site.
One thing we will have to watch out for: each WebDir and WebPage must remember the path to the root of the web site, so it can properly generate links
relative to the root.
To do Define the class WebDir with instance variables webDir and homePath, and define the appropriate initialization method.
Also define class-side methods to prompt the user for the location of the web
site on your computer, as follows:
330
17.1
Tutorial example — generating a site map
Figure 17.1: A WebDir instance.
WebDir >> setDir: dir home: path
webDir := dir.
homePath := path
WebDir class >> onDir: dir
^ self new setDir: dir home: dir pathString
WebDir class >> selectHome
^ self onDir: UIManager default chooseDirectory
The last method opens a browser to select the directory to open. Now, if
you inspect the result of WebDir selectHome, you will be prompted for the
directory containing your web pages, and you will be able to verify that
webDir and homePath are properly initialized to the directory holding your
web site and the full path name of this directory.
It would be nice to be able to programmatically instantiate a WebDir, so let’s
add another creation method.
To do Add the following methods and try it out by inspecting the result
of WebDir onPath: 'path to your web site'.
WebDir class>>onPath: homePath
^ self onPath: homePath home: homePath
WebDir class>>onPath: path home: homePath
^ self new setDir: path asFileReference home: homePath
Pattern matching HTML files
So far so good. Now we would like to use regexes to find out which HTML files
this web site contains.
331
Regular expressions in Pharo
If we browse the FileDirectory class, we find that the method fileNames
will list all the files in a directory. We want to select just those with the file
extension .html. The regex that we need is '.*\.html'. The first dot will
match any character except a newline:
'x' matchesRegex: '.'
>>> true
' ' matchesRegex: '.'
>>> true
Character cr asString matchesRegex: '.'
>>> true
The * (known as the Kleene star, after Stephen Kleene, who invented it) is a
regex operator that will match the preceding regex any number of times
(including zero).
'' matchesRegex: 'x*'
>>> true
'x' matchesRegex: 'x*'
>>> true
'xx' matchesRegex: 'x*'
>>> true
'y' matchesRegex: 'x*'
>>> false
Since the dot is a special character in regexes, if we want to literally match a
dot, then we must escape it.
'.' matchesRegex: '.'
>>> true
'x' matchesRegex: '.'
>>> true
'.' matchesRegex: '\.'
>>> true
'x' matchesRegex: '\.'
>>> false
Now let’s check our regex to find HTML files works as expected.
'index.html' matchesRegex: '.*\.html'
>>> true
'foo.html' matchesRegex: '.*\.html'
>>> true
'style.css' matchesRegex: '.*\.html'
>>> false
332
17.1
Tutorial example — generating a site map
'index.htm' matchesRegex: '.*\.html'
>>> false
Looks good. Now let’s try it out in our application.
To do Add the following method to WebDir and try it out on your test
web site.
WebDir >> htmlFiles
^ webDir fileNames select: [ :each | each matchesRegex: '.*\.html' ]
If you send htmlFiles to a WebDir instance and print it, you should see
something like this:
(WebDir onPath: '...') htmlFiles
>>> #('index.html' ...)
Caching the regex
Now, if you browse matchesRegex:, you will discover that it is an extension
method of String that creates a fresh instance of RxParser every time it is
sent. That is fine for ad hoc queries, but if we are applying the same regex to
every file in a web site, it is smarter to create just one instance of RxParser
and reuse it. Let’s do that.
To do Add a new instance variable htmlRegex to WebDir and initialize it
by sending asRegex to our regex string. Modify WebDir>>htmlFiles to use
the same regex each time as follows:
WebDir >> initialize
htmlRegex := '.*\.html' asRegex
WebDir >> htmlFiles
^ webDir fileNames select: [ :each | htmlRegex matches: each ]
Now listing the HTML files should work just as it did before, except that we
reuse the same regex object many times.
Accessing web pages
Accessing the details of individual web pages should be the responsibility of a
separate class, so let’s define it, and let the WebDir class create the instances.
To do Define a class WebPage with instance variables path, to identify
the HTML file, and homePath, to identify the root directory of the web site.
(We will need this to correctly generate links from the root of the web site
to the files it contains.) Define an initialization method on the instance side
and a creation method on the class side.
333
Regular expressions in Pharo
WebPage >> initializePath: filePath homePath: dirPath
path := filePath.
homePath := dirPath
WebPage class >> on: filePath forHome: homePath
^ self new initializePath: filePath homePath: homePath
A WebDir instance should be able to return a list of all the web pages it contains.
To do Add the following method to WebDir, and inspect the return value
to verify that it works correctly.
WebDir >> webPages
^ self htmlFiles collect:
[ :each | WebPage
on: webDir pathString, '/', each
forHome: homePath ]
You should see something like this:
(WebDir onPath: '...') webPages
>>> an Array(a WebPage a WebPage ...)
String substitutions
That’s not very informative, so let’s use a regex to get the actual file name for
each web page. To do this, we want to strip away all the characters from the
path name up to the last directory. On a Unix file system directories end with
a slash (/), so we need to delete everything up to the last slash in the file path.
The String extension method copyWithRegex:matchesReplacedWith: does
what we want:
'hello' copyWithRegex: '[elo]+' matchesReplacedWith: 'i'
>>> 'hi'
In this example the regex [elo] matches any of the characters e, l or o. The
operator + is like the Kleene star, but it matches exactly one or more instances
of the regex preceding it. Here it will match the entire substring 'ello' and
replay it in a fresh string with the letter i.
To do
Add the following method and verify that it works as expected.
WebPage >> fileName
^ path copyWithRegex: '.*/' matchesReplacedWith: ''
Now you should see something like this on your test web site:
(WebDir onPath: '...') webPages collect: [:each | each fileName ]
>>> #('index.html' ...)
334
17.1
Tutorial example — generating a site map
Extracting regex matches
Our next task is to extract the title of each HTML page. First we need a way to
get at the contents of each page. This is straightforward.
To do
Add the following method and try it out.
WebPage >> contents
^ (FileStream oldFileOrNoneNamed: path) contents
Actually, you might have problems if your web pages contain non-ascii characters, in which case you might be better off with the following code:
WebPage >> contents
^ (FileStream oldFileOrNoneNamed: path)
converter: Latin1TextConverter new;
contents
You should now be able to see something like this:
(WebDir onPath: '...') webPages first contents
>>> '<head>
<title>Home Page</title>
...
'
Now let’s extract the title. In this case we are looking for the text that occurs
between the HTML tags <title> and </title>.
What we need is a way to extract part of the match of a regular expression.
Subexpressions of regexes are delimited by parentheses. Consider the regex
([CARETaeiou]+)([aeiou]+); it consists of two subexpressions, the first of
which will match a sequence of one or more non-vowels, and the second of
which will match one or more vowels: the operator CARET at the start of a
bracketed set of characters negates the set. (NB: In Pharo the caret is also
the return keyword, which we write as ^. To avoid confusion, we will write
CARET when we are using the caret within regular expressions to negate sets of
characters, but you should not forget, they are actually the same thing.) Now
we will try to match a prefix of the string 'pharo' and extract the submatches:
| re |
re := '([CARETaeiou]+)([aeiou]+)' asRegex.
re matchesPrefix: 'pharo'
>>> true
re subexpression: 1
>>> 'pha'
re subexpression: 2
>>> 'ph'
re subexpression: 3
>>> 'a'
335
Regular expressions in Pharo
After successfully matching a regex against a string, you can always send it
the message subexpression: 1 to extract the entire match. You can also send
subexpression: n where n-1 is the number of subexpressions in the regex.
The regex above has two subexpressions, numbered 2 and 3.
We will use the same trick to extract the title from an HTML file.
To do
Define the following method:
WebPage >> title
| re |
re := '[\w\W]*<title>(.*)</title>' asRegexIgnoringCase.
^ (re matchesPrefix: self contents)
ifTrue: [ re subexpression: 2 ]
ifFalse: [ '(', self fileName, ' -- untitled)' ]
There are a couple of subtle points to notice here. First, HTML does not care
whether tags are upper or lower case, so we must make our regex case insensitive by instantiating it with asRegexIgnoringCase.
Second, since dot matches any character except a newline, the regex .*<title>(.*)</title> will not work as expected if multiple lines appear before
the title. The regex \w matches any alphanumeric, and \W will match any
non-alphanumeric, so [\w\W] will match any character including newlines. (If
we expect titles to possible contain newlines, we should play the same trick
with the subexpression.)
Now we can test our title extractor, and we should see something like this:
(WebDir onPath: '...') webPages first title
>>> 'Home page'
More string substitutions
In order to generate our site map, we need to generate links to the individual
web pages. We can use the document title as the name of the link. We just
need to generate the right path to the web page from the root of the web site.
Luckily this is trivial — it is simple the full path to the web page minus the full
path to the root directory of the web site.
We must only watch out for one thing. Since the homePath variable does
not end in a /, we must append one, so that relative path does not include a
leading /. Notice the difference between the following two results:
'/home/testweb/index.html' copyWithRegex: '/home/testweb'
matchesReplacedWith: ''
>>> '/index.html'
'/home/testweb/index.html' copyWithRegex: '/home/testweb/'
matchesReplacedWith: ''
>>> 'index.html'
336
17.1
Tutorial example — generating a site map
The first result would give us an absolute path, which is probably not what we
want.
To do
Define the following methods:
WebPage >> relativePath
^ path
copyWithRegex: homePath, '/'
matchesReplacedWith: ''
WebPage >> link
^ '<a href="', self relativePath, '">', self title, '</a>'
You should now be able to see something like this:
(WebDir onPath: '...') webPages first link
>>> '<a href="index.html">Home Page</a>'
Generating the site map
Actually, we are now done with the regular expressions we need to generate
the site map. We just need a few more methods to complete the application.
To do If you want to see the site map generation, just add the following
methods.
If our web site has subdirectories, we need a way to access them:
WebDir >> webDirs
^ webDir directoryNames
collect: [ :each | WebDir onPath: webDir pathString, '/', each
home: homePath ]
We need to generate HTML bullet lists containing links for each web page of a
web directory. Subdirectories should be indented in their own bullet list.
WebDir >> printTocOn: aStream
self htmlFiles
ifNotEmpty: [
aStream nextPutAll: '<ul>'; cr.
self webPages
do: [:each | aStream nextPutAll: '<li>';
nextPutAll: each link;
nextPutAll: '</li>'; cr].
self webDirs
do: [:each | each printTocOn: aStream].
aStream nextPutAll: '</ul>'; cr]
We create a file called toc.html in the root web directory and dump the site
map there.
337
Regular expressions in Pharo
Figure 17.2: A small site map.
WebDir >> tocFileName
^ 'toc.html'
WebDir >> makeToc
| tocStream |
tocStream := (webDir / self tocFileName) writeStream.
self printTocOn: tocStream.
tocStream close.
Now we can generate a table of contents for an arbitrary web directory!
WebDir selectHome makeToc
17.2
Regex syntax
We will now have a closer look at the syntax of regular expressions as supported by the Regex package.
The simplest regular expression is a single character. It matches exactly that
character. A sequence of characters matches a string with exactly the same
sequence of characters:
'a' matchesRegex: 'a'
>>> true
'foobar' matchesRegex: 'foobar'
>>> true
'blorple' matchesRegex: 'foobar'
>>> false
Operators are applied to regular expressions to produce more complex regular expressions. Sequencing (placing expressions one after another) as an
operator is, in a certain sense, invisible — yet it is arguably the most common.
338
17.2 Regex syntax
We have already seen the Kleene star (*) and the + operator. A regular expression followed by an asterisk matches any number (including 0) of matches of
the original expression. For example:
'ab' matchesRegex: 'a*b'
>>> true
'aaaaab' matchesRegex: 'a*b'
>>> true
'b' matchesRegex: 'a*b'
>>> true
'aac' matchesRegex: 'a*b'
>>> false "b does not match"
The Kleene star has higher precedence than sequencing. A star applies to the
shortest possible subexpression that precedes it. For example, ab* means a
followed by zero or more occurrences of b, not zero or more occurrences of ab:
'abbb' matchesRegex: 'ab*'
>>> true
'abab' matchesRegex: 'ab*'
>>> false
To obtain a regex that matches zero or more occurrences of ab, we must enclose
ab in parentheses:
'abab' matchesRegex: '(ab)*'
>>> true
'abcab' matchesRegex: '(ab)*'
>>> false "c spoils the fun"
Two other useful operators similar to * are + and ?. + matches one or more
instances of the regex it modifies, and ? will match zero or one instance.
'ac' matchesRegex: 'ab*c'
>>> true
'ac' matchesRegex: 'ab+c'
>>> false "need at least one b"
'abbc' matchesRegex: 'ab+c'
>>> true
'abbc' matchesRegex: 'ab?c'
>>> false "too many b's"
As we have seen, the characters *, +, ?, (, and ) have special meaning within
regular expressions. If we need to match any of them literally, it should be
escaped by preceding it with a backslash \ . Thus, backslash is also special
character, and needs to be escaped for a literal match. The same holds for all
further special characters we will see.
339
Regular expressions in Pharo
'ab*' matchesRegex: 'ab*'
>>> false "star in the right string is special"
'ab*' matchesRegex: 'ab\*'
>>> true
'a\c' matchesRegex: 'a\\c'
>>> true
The last operator is |, which expresses choice between two subexpressions. It
matches a string if either of the two subexpressions matches the string. It has
the lowest precedence — even lower than sequencing. For example, ab*|ba*
means a followed by any number of b’s, or b followed by any number of a’s:
'abb' matchesRegex: 'ab*|ba*'
>>> true
'baa' matchesRegex: 'ab*|ba*'
>>> true
'baab' matchesRegex: 'ab*|ba*'
>>> false
A bit more complex example is the expression c(a|d)+r, which matches the
name of any of the Lisp-style car, cdr, caar, cadr, ... functions:
'car' matchesRegex: 'c(a|d)+r'
>>> true
'cdr' matchesRegex: 'c(a|d)+r'
>>> true
'cadr' matchesRegex: 'c(a|d)+r'
>>> true
It is possible to write an expression that matches an empty string, for example
the expression a| matches an empty string. However, it is an error to apply *,
+, or ? to such an expression: (a|)* is invalid.
So far, we have used only characters as the smallest components of regular
expressions. There are other, more interesting, components. A character set
is a string of characters enclosed in square brackets. It matches any single
character if it appears between the brackets. For example, [01] matches
either 0 or 1:
'0' matchesRegex: '[01]'
>>> true
'3' matchesRegex: '[01]'
>>> false
'11' matchesRegex: '[01]'
>>> false "a set matches only one character"
Using plus operator, we can build the following binary number recognizer:
340
17.2 Regex syntax
'10010100' matchesRegex: '[01]+'
>>> true
'10001210' matchesRegex: '[01]+'
>>> false
If the first character after the opening bracket is CARET, the set is inverted: it
matches any single character not appearing between the brackets:
'0' matchesRegex: '[CARET01]'
>>> false
'3' matchesRegex: '[CARET01]'
>>> true
For convenience, a set may include ranges: pairs of characters separated by a
hyphen (-). This is equivalent to listing all characters in between: '[0-9]' is
the same as '[0123456789]'. Special characters within a set are CARET, -, and
], which closes the set. Below are examples how to literally match them in a
set:
'CARET' matchesRegex: '[01CARET]'
>>> true "put the caret anywhere except the start"
'-' matchesRegex: '[01-]'
>>> true "put the hyphen at the end"
']' matchesRegex: '[]01]'
>>> true "put the closing bracket at the start"
Thus, empty and universal sets cannot be specified.
Character classes
Regular expressions can also include the following backquote escapes to refer
to popular classes of characters: \w to match alphanumeric characters, \d to
match digits, and \s to match whitespace. Their upper-case variants, \W, \D
and \S, match the complementary characters (non-alphanumerics, non-digits
and non-whitespace). Here is a summary of the syntax seen so far:
341
Regular expressions in Pharo
Syntax
What it represents
a
.
(...)
\x
*
+
?
|
[abcd]
[^abcd]
[0-9]
\w
\W
\d
\D
\s
\S
literal match of character a
match any char (except newline)
group subexpression
escape the following special character where ’x’ can be ’w’,’s’,’d’,’W’,’S’,’D’
Kleene star — match previous regex zero or more times
match previous regex one or more times
match previous regex zero times or once
match choice of left and right regex
match choice of characters abcd
match negated choice of characters
match range of characters 0 to 9
match alphanumeric
match non-alphanumeric
match digit
match non-digit
match space
match non-space
As mentioned in the introduction, regular expressions are especially useful for
validating user input, and character classes turn out to be especially useful for
defining such regexes. For example, non-negative numbers can be matched
with the regex \d+:
'42' matchesRegex: '\d+'
>>> true
'-1' matchesRegex: '\d+'
>>> false
Better yet, we might want to specify that non-zero numbers should not start
with the digit 0:
'0' matchesRegex: '0|([1-9]\d*)'
>>> true
'1' matchesRegex: '0|([1-9]\d*)'
>>> true
'42' matchesRegex: '0|([1-9]\d*)'
>>> true
'099' matchesRegex: '0|([1-9]\d*)'
>>> false "leading 0"
We can check for negative and positive numbers as well:
'0' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> true
342
17.2 Regex syntax
'-1' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> true
'42' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> true
'+99' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> true
'-0' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> false "negative zero"
'01' matchesRegex: '(0|((\+|-)?[1-9]\d*))'
>>> false "leading zero"
Floating point numbers should require at least one digit after the dot:
'0' matchesRegex: '(0|((\+|-)?[1-9]\d*))(\.\d+)?'
>>> true
'0.9' matchesRegex: '(0|((\+|-)?[1-9]\d*))(\.\d+)?'
>>> true
'3.14' matchesRegex: '(0|((\+|-)?[1-9]\d*))(\.\d+)?'
>>> true
'-42' matchesRegex: '(0|((\+|-)?[1-9]\d*))(\.\d+)?'
>>> true
'2.' matchesRegex: '(0|((\+|-)?[1-9]\d*))(\.\d+)?'
>>> false "need digits after ."
For dessert, here is a recognizer for a general number format: anything like
999, or 999.999, or -999.999e+21.
'-999.999e+21' matchesRegex: '(\+|-)?\d+(\.\d*)?((e|E)(\+|-)?\d+)?'
>>> true
Character classes can also include the following grep(1)-compatible elements:
Syntax
What it represents
[:alnum:]
[:alpha:]
[:cntrl:]
[:digit:]
[:graph:]
[:lower:]
[:print:]
[:punct:]
[:space:]
[:upper:]
[:xdigit:]
any alphanumeric
any alphabetic character
any control character (ascii code below 32)
any decimal digit
any graphical character (ascii code above 32)
any lowercase character
any printable character (here, the same as [:graph:])
any punctuation character
any whitespace character
any uppercase character
any hexadecimal character
343
Regular expressions in Pharo
Note that these elements are components of the character classes, i.e., they
have to be enclosed in an extra set of square brackets to form a valid regular
expression. For example, a non-empty string of digits would be represented as
[[:digit:]]+. The above primitive expressions and operators are common to
many implementations of regular expressions.
'42' matchesRegex: '[[:digit:]]+'
>>> true
Special character classes
The next primitive expression is unique to this Smalltalk implementation. A
sequence of characters between colons is treated as a unary selector which
is supposed to be understood by characters. A character matches such an
expression if it answers true to a message with that selector. This allows a
more readable and efficient way of specifying character classes. For example,
[0-9] is equivalent to :isDigit:, but the latter is more efficient. Analogously
to character sets, character classes can be negated: :CARETisDigit: matches
a character that answers false to isDigit, and is therefore equivalent to
[CARET0-9].
So far we have seen the following equivalent ways to write a regular expression that matches a non-empty string of digits: [0-9]+, \d+, [\d]+,
[[:digit:]]+, :isDigit:+.
'42' matchesRegex: '[0-9]+'
>>> true
'42' matchesRegex: '\d+'
>>> true
'42' matchesRegex: '[\d]+'
>>> true
'42' matchesRegex: '[[:digit:]]+'
>>> true
'42' matchesRegex: ':isDigit:+'
>>> true
Matching boundaries
The last group of special primitive expressions shown next is used to match
boundaries of strings.
344
17.3
Regex API
Syntax
What it represents
CARET
\$
\b
\B
\<
\>
match an empty string at the beginning of a line
match an empty string at the end of a line
match an empty string at a word boundary
match an empty string not at a word boundary
match an empty string at the beginning of a word
match an empty string at the end of a word
'hello world' matchesRegex: '.*\bw.*'
>>> true "word boundary before w"
'hello world' matchesRegex: '.*\bo.*'
>>> false "no boundary before o"
17.3
Regex API
Up to now we have focussed mainly on the syntax of regexes. Now we will
have a closer look at the different messages understood by strings and
regexes.
Matching prefixes and ignoring case
So far most of our examples have used the String extension method matchesRegex:.
Strings also understand the following messages: prefixMatchesRegex:,
matchesRegexIgnoringCase: and prefixMatchesRegexIgnoringCase:.
The message prefixMatchesRegex: is just like matchesRegex, except that the
whole receiver is not expected to match the regular expression passed as the
argument; matching just a prefix of it is enough.
'abacus' matchesRegex: '(a|b)+'
>>> false
'abacus' prefixMatchesRegex: '(a|b)+'
>>> true
'ABBA' matchesRegexIgnoringCase: '(a|b)+'
>>> true
'Abacus' matchesRegexIgnoringCase: '(a|b)+'
>>> false
'Abacus' prefixMatchesRegexIgnoringCase: '(a|b)+'
>>> true
345
Regular expressions in Pharo
Enumeration interface
Some applications need to access all matches of a certain regular expression
within a string. The matches are accessible using a protocol modeled after the
familiar Collection-like enumeration protocol.
regex:matchesDo: evaluates a one-argument aBlock for every match of the
regular expression within the receiver string.
| list |
list := OrderedCollection new.
'Jack meet Jill' regex: '\w+' matchesDo: [:word | list add: word].
list
>>> an OrderedCollection('Jack' 'meet' 'Jill')
regex:matchesCollect: evaluates a one-argument aBlock for every match of
the regular expression within the receiver string. It then collects the results
and answers them as a SequenceableCollection.
'Jack meet Jill' regex: '\w+' matchesCollect: [:word | word size]
>>> an OrderedCollection(4 4 4)
allRegexMatches: returns a collection of all matches (substrings of the
receiver string) of the regular expression.
'Jack and Jill went up the hill' allRegexMatches: '\w+'
>>> an OrderedCollection('Jack' 'and' 'Jill' 'went' 'up' 'the' 'hill')
Replacement and translation
It is possible to replace all matches of a regular expression with a certain
string using the message copyWithRegex:matchesReplacedWith:.
'Krazy hates Ignatz' copyWithRegex: '\<[[:lower:]]+\>'
matchesReplacedWith: 'loves'
>>> 'Krazy loves Ignatz'
A more general substitution is match translation. This message evaluates a
block passing it each match of the regular expression in the receiver string
and answers a copy of the receiver with the block results spliced into it in
place of the respective matches.
'Krazy loves Ignatz' copyWithRegex: '\b[a-z]+\b'
matchesTranslatedUsing: [:each | each asUppercase]
>>> 'Krazy LOVES Ignatz'
All messages of enumeration and replacement protocols perform a casesensitive match. Case-insensitive versions are not provided as part of a String
protocol. Instead, they are accessible using the lower-level matching interface
presented in the following section.
346
17.3
Regex API
Lower-level interface
When you send the message matchesRegex: to a string, the following happens:
• A fresh instance of RxParser is created, and the regular expression
string is passed to it, yielding the expression’s syntax tree.
• The syntax tree is passed as an initialization parameter to an instance of
RxMatcher. The instance sets up some data structure that will work as a
recognizer for the regular expression described by the tree.
• The original string is passed to the matcher, and the matcher checks for
a match.
The Matcher
If you repeatedly match a number of strings against the same regular expression using one of the messages defined in String, the regular expression
string is parsed and a new matcher is created for every match. You can avoid
this overhead by building a matcher for the regular expression, and then
reusing the matcher over and over again. You can, for example, create a
matcher at a class or instance initialization stage, and store it in a variable for
future use. You can create a matcher using one of the following methods:
• You can send asRegex or asRegexIgnoringCase to the string.
• You can directly invoke the RxMatcher constructor methods forString:
or forString:ignoreCase: (which is what the convenience methods
above will do).
Here we send matchesIn: to collect all the matches found in a string:
| octal |
octal := '8r[0-9A-F]+' asRegex.
octal matchesIn: '8r52 = 16r2A'
>>> an OrderedCollection('8r52')
| hex |
hex := '16r[0-9A-F]+' asRegexIgnoringCase.
hex matchesIn: '8r52 = 16r2A'
>>> an OrderedCollection('16r2A')
| hex |
hex := RxMatcher forString: '16r[0-9A-Fa-f]+' ignoreCase: true.
hex matchesIn: '8r52 = 16r2A'
>>> an OrderedCollection('16r2A')
Matching
A matcher understands these messages (all of them return true to indicate
successful match or search, and false otherwise):
347
Regular expressions in Pharo
matches: aString — true if the whole argument string (aString) matches.
'\w+' asRegex matches: 'Krazy'
>>> true
matchesPrefix: aString — true if some prefix of the argument string (not
necessarily the whole string) matches.
'\w+' asRegex matchesPrefix: 'Ignatz hates Krazy'
>>> true
search: aString — Search the string for the first occurrence of a matching
substring. Note that the first two methods only try matching from the very
beginning of the string. Using the above example with a matcher for a+, this
method would answer success given a string 'baaa', while the previous two
would fail.
'\b[a-z]+\b' asRegex search: 'Ignatz hates Krazy'
>>> true "finds 'hates'"
The matcher also stores the outcome of the last match attempt and can report
it: lastResult answers a Boolean: the outcome of the most recent match
attempt. If no matches were attempted, the answer is unspecified.
| number |
number := '\d+' asRegex.
number search: 'Ignatz throws 5 bricks'.
number lastResult
>>> true
matchesStream:, matchesStreamPrefix: and searchStream: are analogous
to the above three messages, but takes streams as their argument.
| ignatz names |
ignatz := ReadStream on: 'Ignatz throws bricks at Krazy'.
names := '\<[A-Z][a-z]+\>' asRegex.
names matchesStreamPrefix: ignatz
>>> true
Subexpression matches
After a successful match attempt, you can query which part of the original
string has matched which part of the regex. A subexpression is a parenthesized part of a regular expression, or the whole expression. When a regular
expression is compiled, its subexpressions are assigned indices starting from 1,
depth-first, left-to-right.
For example, the regex ((\d+)\s*(\w+)) has four subexpressions, including
itself.
1: ((\d+)\s*(\w+)) "the complete expression"
2: (\d+)\s*(\w+) "top parenthesized subexpression"
348
17.3
Regex API
3: \d+ "first leaf subexpression"
4: \w+ "second leaf subexpression"
The highest valid index is equal to 1 plus the number of matching parentheses.
(So, 1 is always a valid index, even if there are no parenthesized subexpressions.)
After a successful match, the matcher can report what part of the original
string matched what subexpression. It understands these messages:
subexpressionCount answers the total number of subexpressions: the highest
value that can be used as a subexpression index with this matcher. This value
is available immediately after initialization and never changes.
subexpression: takes a valid index as its argument, and may be sent only
after a successful match attempt. The method answers a substring of the
original string the corresponding subexpression has matched to.
subBeginning: and subEnd: answer the positions within the argument string
or stream where the given subexpression match has started and ended, respectively.
| items |
items := '((\d+)\s*(\w+))' asRegex.
items search: 'Ignatz throws 1 brick at Krazy'.
items subexpressionCount
>>> 4
items subexpression: 1
>>> '1 brick' "complete expression"
items subexpression: 2
>>> '1 brick' "top subexpression"
items subexpression: 3
>>> '1' "first leaf subexpression"
items subexpression: 4
>>> 'brick' "second leaf subexpression"
items subBeginning: 3
>>> 14
items subEnd: 3
>>> 15
items subBeginning: 4
>>> 16
items subEnd: 4
>>> 21
As a more elaborate example, the following piece of code uses a MMM DD, YYYY
date format recognizer to convert a date to a three-element array with year,
month, and day strings:
| date result |
date :=
'(Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\s+(\d\d?)\s*,\s*19(\d\d)'
asRegex.
349
Regular expressions in Pharo
result := (date matches: 'Aug 6, 1996')
ifTrue: [{ (date subexpression: 4) .
(date subexpression: 2) .
(date subexpression: 3) } ]
ifFalse: ['no match'].
result
>>> #('96' 'Aug' '6')
Enumeration and Replacement
The String enumeration and replacement protocols that we saw earlier in
this section are actually implemented by the matcher. RxMatcher implements
the following methods for iterating over matches within strings: matchesIn:,
matchesIn:do:, matchesIn:collect:, copy:replacingMatchesWith: and
copy:translatingMatchesUsing:.
| seuss aWords |
seuss := 'The cat in the hat is back'.
aWords := '\<([^aeiou]|[a])+\>' asRegex. "match words with 'a' in them"
aWords matchesIn: seuss
>>> an OrderedCollection('cat' 'hat' 'back')
aWords matchesIn: seuss collect: [:each | each asUppercase ]
>>> an OrderedCollection('CAT' 'HAT' 'BACK')
aWords copy: seuss replacingMatchesWith: 'grinch'
>>> 'The grinch in the grinch is grinch'
aWords copy: seuss translatingMatchesUsing: [ :each | each asUppercase ]
>>> 'The CAT in the HAT is BACK'
There are also the following methods for iterating over matches within
streams: matchesOnStream:, matchesOnStream:do:, matchesOnStream:collect:, copyStream:to:replacingMatchesWith: and copyStream:to:translatingMatchesUsing:.
Error Handling
Several exceptions may be raised by RxParser when building regexes. The
exceptions have the common parent RegexError. You may use the usual
Smalltalk exception handling mechanism to catch and handle them.
• RegexSyntaxError is raised if a syntax error is detected while parsing a
• regex RegexCompilationError is raised if an error is detected while
• building a matcher RegexMatchingError is raised if an error occurs
while
• matching (for example, if a bad selector was specified using ':<selector>:'
• syntax, or because of the matcher’s internal error).
350
17.4
Implementation Notes by Vassili Bykov
['+' asRegex] on: RegexError do: [:ex | ^ ex printString ]
>>> 'RegexSyntaxError: nullable closure'
17.4
Implementation Notes by Vassili Bykov
In 90% of the cases, the method String>>matchesRegex: is all you need to access the package.
What to look at first.
RxParser accepts a string or a stream of characters with a regular expression,
and produces a syntax tree corresponding to the expression. The tree is made
of instances of Rxs* classes.
RxMatcher accepts a syntax tree of a regular expression built by the parser
and compiles it into a matcher: a structure made of instances of Rxm* classes.
The RxMatcher instance can test whether a string or a positionable stream of
characters matches the original regular expression, or it can search a string
or a stream for substrings matching the expression. After a match is found,
the matcher can report a specific string that matched the whole expression,
or any parenthesized subexpression of it. All other classes support the same
functionality and are used by RxParser, RxMatcher, or both.
The matcher is similar in spirit, but not in design to Henry Spencer’s
original regular expression implementation in C. The focus is on simplicity,
not on efficiency. I didn’t optimize or profile anything. The matcher passes H.
Spencer’s test suite (see test suite protocol), with quite a few extra tests added,
so chances are good there are not too many bugs. But watch out anyway.
Caveats
Since the first release of the matcher, thanks to the input
from several fellow Smalltalkers, I became convinced a native Smalltalk
regular expression matcher was worth the effort to keep it alive. For the
advice and encouragement that made this release possible, I want to thank:
Felix Hack, Eliot Miranda, Robb Shecter, David N. Smith, Francis Wolinski and
anyone whom I haven’t yet met or heard from, but who agrees this has not
been a complete waste of time.
Acknowledgments
17.5
Chapter Summary
Regular expressions are an essential tool for manipulating strings in a trivial
way. This chapter presented the Regex package for Pharo. The essential points
of this chapter are:
• For simple matching, just send matchesRegex: to a string
• When performance matters, send asRegex to the string representing the
regex, and reuse the resulting matcher for multiple matches
351
Regular expressions in Pharo
• Subexpression of a matching regex may be easily retrieved to an arbitrary depth
• A matching regex can also replace or translate subexpressions in a new
copy of the string matched
• An enumeration interface is provided to access all matches of a certain
regular expression
• Regexes work with streams as well as with strings.
352