1Role::Basic::PhilosophyU(s3e)r Contributed Perl DocumentaRtoiloen::Basic::Philosophy(3)
2
3
4
6 Role::Basic::Philosophy - Why Role::Basic exists.
7
9 Note: the words "trait" and "role" will be used interchangeably
10 throughout this documentation.
11
12 After years of using roles, your author has found that many people
13 would be happy to use roles but are not willing/comfortable with using
14 Moose. This module implements roles and nothing else. It does so in a
15 (relatively) simple bit of code. However, you should be aware that
16 there are some differences between Role::Basic and Moose::Role.
17
18 Moose is a fantastic technology and your author is quite happy with it.
19 He urges you to check it out and perhaps even consider Role::Basic a
20 "stepping-stone" to Moose. However, after an informal poll with many
21 respondents replying on blogs.perl.org, Twitter, Facebook and private
22 email unanimously saying they wanted this module for roles and not as a
23 stepping-stone to Moose, your author took the liberty of deciding to
24 implement traits in a rather faithful fashion, rather than strictly
25 adhere to the design of Moose::Role. For areas where we differ,
26 Role::Basic intends to be more restrictive when syntax is the same.
27 This allows an easier migration to Moose::Role when the time is right.
28 Otherwise, Role::Basic will offer a different syntax to avoid
29 confusion.
30
32 As most of you probably know, roles are the Perl implmentation of
33 traits as described in <http://scg.unibe.ch/research/traits/>. (The
34 name "role" was chosen because "trait" was already used in Perl 6.) In
35 particular, we direct you to two papers, both of which are easy to
36 read:
37
38 · <http://scg.unibe.ch/scgbib?_s=tgXJjGrs0380ejY6&_k=Swgdwx_C&query=nathanael+traits+composable+units+ecoop&display=abstract&_n&19>
39
40 The seminal "traits paper" which much of the documentation refers
41 to.
42
43 · <http://scg.unibe.ch/scgbib?_s=tgXJjGrs0380ejY6&_k=Swgdwx_C&query=traits+the+formal+model&display=abstract&_n&23>
44
45 "Traits: The Formal Model".
46
47 While less well-known, this relatively easy to read paper outlines
48 the mathematical underpinnings of traits and explains several
49 design decisions taken here.
50
51 It is important to refer back to those papers because Role::Basic
52 attempts to implements traits as described in the research, whereas
53 Moose::Role attempts to implement something very similar to traits, but
54 with more of a "Perlish" feel. This is not intended as a criticism of
55 Moose::Role, but merely an attempt to alert the reader to key
56 differences.
57
58 The Basics
59 Roles are simply bundles of behavior which classes may use. If you have
60 two completely unrelated classes, your code may still require each of
61 them to serialize themselves as JSON even though neither class
62 naturally has anything to do with JSON (for example, "Person" and
63 "Order" classes). There are a number of approaches to this problem but
64 if you're here I'll skip the explanation and assume that you already
65 understand roles and would like to know why we don't follow the
66 Moose::Role specification.
67
68 As you already probably know, roles allow you to state that your class
69 "DOES" some behaviour, and allows you to exclude or alias bits and
70 pieces of the roles you're including. The original specification of
71 traits made it clear that this was to be done in such a fashion that no
72 matter how you grouped the traits or in which order you used them, the
73 outcome behavior would be the same. That's why we have subtle but
74 forward-compatible differences with Moose::Role.
75
76 Commutative
77
78 The formal model
79 (<http://scg.unibe.ch/archive/papers/Scha02cTraitsModel.pdf>) states
80 that trait composition must be commutative (section 3.4, proposition
81 1). This means that:
82
83 (A + B) = (B + A)
84
85 In other words, it should not matter what order you compose the traits
86 in. It is well known that with both inheritance and mixins, this does
87 not hold (making refactoring a dicey proposition at times), but when
88 method modifiers are used with Moose::Role, the same issues arises
89 (from
90 <http://blogs.perl.org/users/ovid/2010/12/rolebasic---when-you-only-want-roles.html>):
91
92 {
93 package Some::Role;
94 use Moose::Role;
95 requires qw(some_method);
96
97 before some_method => sub {
98 my $self = shift;
99 $self->some_number( $self->some_number + 2 );
100 };
101 }
102 {
103 package Another::Role;
104 use Moose::Role;
105 requires qw(some_method);
106
107 before some_method => sub {
108 my $self = shift;
109 $self->some_number( $self->some_number / 2 );
110 };
111 }
112 {
113 package Some::Class;
114 use Moose;
115 my @roles =
116 int( rand(2) )
117 ? qw(Another::Role Some::Role)
118 : qw(Some::Role Another::Role);
119 with @roles;
120
121 has some_number => ( is => 'rw', isa => 'Num' );
122 sub some_method { print shift->some_number, $/ }
123 }
124 my $o = Some::Class->new( { some_number => 7 } );
125 $o->some_method;
126
127 If you run this code, it might print 4.5, but it might print 5.5. As
128 with mixins and multiple inheritance, you have no way of knowing the
129 exact behaviour which will be exhibited short of running the code. No
130 introspection will help. This is not an issue with Role::Basic because
131 we do not allow method modifiers. If you think you need them, please
132 consider Moose.
133
134 Associative
135
136 The formal model
137 (<http://scg.unibe.ch/archive/papers/Scha02cTraitsModel.pdf>) states
138 that trait composition must be associative (section 3.4, proposition
139 1). This means that:
140
141 (A + B) + C = A + (B + C)
142
143 Moose is associative if and only if you do not have multiple methods
144 with the same name. In Moose, if a role providing method M consumes
145 one other role which also provides method M, we have a conflict:
146
147 package Some::Role;
148 use Moose::Role;
149 sub bar { __PACKAGE__ }
150
151 package Some::Other::Role;
152 use Moose::Role;
153 with 'Some::Role';
154 sub bar { __PACKAGE__ }
155
156 package Some::Class;
157 use Moose;
158 with 'Some::Other::Role';
159
160 package main;
161 my $o = Some::Class->new;
162 print $o->bar;
163
164 However, if the role consumes two or more other roles which provide the
165 same method, we don't have a conflict:
166
167 package Some::Role;
168 use Moose::Role;
169 sub bar { __PACKAGE__ }
170
171 package Some::Other::Role;
172 use Moose::Role;
173 sub bar { __PACKAGE__ }
174
175 package Another::Role;
176 use Moose::Role;
177 with qw(Some::Role Some::Other::Role);
178 sub bar { __PACKAGE__ }
179
180 package Some::Class;
181 use Moose;
182 with 'Another::Role';
183
184 package main;
185 my $o = Some::Class->new;
186 print $o->bar;
187
188 This is because, in Moose, when you have two or more roles consumed,
189 any conflicting methods are excluded and considered to be requirements.
190
191 See "Moose::Role composition edge cases" for more explanation:
192 <http://search.cpan.org/~drolsky/Moose-1.21/lib/Moose/Spec/Role.pod#Composition_Edge_Cases>.
193
194 This makes roles easy to use at times, but it means that the following
195 three groups of roles are not guaranteed to provide the same behavior:
196
197 RoleA does RoleB, RoleC
198 RoleB does RoleA, RoleC
199 RoleC does RoleA, RoleB
200
201 Further, you as a developer have no way of knowing that we have had
202 methods silently excluded without reading all of the code.
203
204 For Role::Basic there are no edge cases. If "RoleA", "RoleB", and
205 "RoleC" all provide method M, you are guaranteed to get a conflict at
206 composition time and must specifically address the problem. This
207 addresses the associative issue because strictly speaking, a trait is
208 merely a bundle of services provided, not its name. Thus, a trait with
209 its "foo" method excluded is not the same as itself without the "foo"
210 method excluded.
211
212 Benefits of associative and commutative behaviour
213
214 While we recognize that not everyone will be happy with the decisions
215 we have made, we have several benefits here:
216
217 · We adhere to the formal definition of traits
218
219 · Ordering and grouping of traits does not alter their behavior
220
221 · We're forward-compatible with Moose::Role
222
224 The primary goal of Role::Basic is to provide traits in a simple and
225 safe manner. We are huge fans of Moose and Moose::Role and suggest that
226 everyone check them out. The decision of Moose::Role to deviate from
227 the "associative" and "commutative" deviations from the original traits
228 model is, in our experience, less likely to occur with roles than with
229 mixins and inhertance, so please do not take this as an indictment, but
230 rather in the spirit of TIMTOWTDI.
231
232
233
234perl v5.30.0 2019-07-26 Role::Basic::Philosophy(3)