1Algorithm::C3(3) User Contributed Perl Documentation Algorithm::C3(3)
2
3
4
6 Algorithm::C3 - A module for merging hierarchies using the C3 algorithm
7
9 use Algorithm::C3;
10
11 # merging a classic diamond
12 # inheritance graph like this:
13 #
14 # <A>
15 # / \
16 # <B> <C>
17 # \ /
18 # <D>
19
20 my @merged = Algorithm::C3::merge(
21 'D',
22 sub {
23 # extract the ISA array
24 # from the package
25 no strict 'refs';
26 @{$_[0] . '::ISA'};
27 }
28 );
29
30 print join ", " => @merged; # prints D, B, C, A
31
33 This module implements the C3 algorithm. I have broken this out into
34 it's own module because I found myself copying and pasting it way too
35 often for various needs. Most of the uses I have for C3 revolve around
36 class building and metamodels, but it could also be used for things
37 like dependency resolution as well since it tends to do such a nice job
38 of preserving local precedence orderings.
39
40 Below is a brief explanation of C3 taken from the Class::C3 module. For
41 more detailed information, see the "SEE ALSO" section and the links
42 there.
43
44 What is C3?
45 C3 is the name of an algorithm which aims to provide a sane method
46 resolution order under multiple inheritance. It was first introduced in
47 the language Dylan (see links in the "SEE ALSO" section), and then
48 later adopted as the preferred MRO (Method Resolution Order) for the
49 new-style classes in Python 2.3. Most recently it has been adopted as
50 the 'canonical' MRO for Perl 6 classes, and the default MRO for Parrot
51 objects as well.
52
53 How does C3 work.
54 C3 works by always preserving local precedence ordering. This
55 essentially means that no class will appear before any of it's
56 subclasses. Take the classic diamond inheritance pattern for instance:
57
58 <A>
59 / \
60 <B> <C>
61 \ /
62 <D>
63
64 The standard Perl 5 MRO would be (D, B, A, C). The result being that A
65 appears before C, even though C is the subclass of A. The C3 MRO
66 algorithm however, produces the following MRO (D, B, C, A), which does
67 not have this same issue.
68
69 This example is fairly trivial, for more complex examples and a deeper
70 explanation, see the links in the "SEE ALSO" section.
71
73 merge ($root, $func_to_fetch_parent, $cache)
74 This takes a $root node, which can be anything really it is up to
75 you. Then it takes a $func_to_fetch_parent which can be either a
76 CODE reference (see SYNOPSIS above for an example), or a string
77 containing a method name to be called on all the items being
78 linearized. An example of how this might look is below:
79
80 {
81 package A;
82
83 sub supers {
84 no strict 'refs';
85 @{$_[0] . '::ISA'};
86 }
87
88 package C;
89 our @ISA = ('A');
90 package B;
91 our @ISA = ('A');
92 package D;
93 our @ISA = ('B', 'C');
94 }
95
96 print join ", " => Algorithm::C3::merge('D', 'supers');
97
98 The purpose of $func_to_fetch_parent is to provide a way for
99 "merge" to extract the parents of $root. This is needed for C3 to
100 be able to do it's work.
101
102 The $cache parameter is an entirely optional performance measure,
103 and should not change behavior.
104
105 If supplied, it should be a hashref that merge can use as a private
106 cache between runs to speed things up. Generally speaking, if you
107 will be calling merge many times on related things, and the parent
108 fetching function will return constant results given the same
109 arguments during all of these calls, you can and should reuse the
110 same shared cache hash for all of the calls. Example:
111
112 sub do_some_merging {
113 my %merge_cache;
114 my @foo_mro = Algorithm::C3::Merge('Foo', \&get_supers, \%merge_cache);
115 my @bar_mro = Algorithm::C3::Merge('Bar', \&get_supers, \%merge_cache);
116 my @baz_mro = Algorithm::C3::Merge('Baz', \&get_supers, \%merge_cache);
117 my @quux_mro = Algorithm::C3::Merge('Quux', \&get_supers, \%merge_cache);
118 # ...
119 }
120
122 I use Devel::Cover to test the code coverage of my tests, below is the
123 Devel::Cover report on this module's test suite.
124
125 ------------------------ ------ ------ ------ ------ ------ ------ ------
126 File stmt bran cond sub pod time total
127 ------------------------ ------ ------ ------ ------ ------ ------ ------
128 Algorithm/C3.pm 100.0 100.0 100.0 100.0 100.0 100.0 100.0
129 ------------------------ ------ ------ ------ ------ ------ ------ ------
130 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0
131 ------------------------ ------ ------ ------ ------ ------ ------ ------
132
134 The original Dylan paper
135 <http://www.webcom.com/haahr/dylan/linearization-oopsla96.html>
136
137 The prototype Perl 6 Object Model uses C3
138 <http://svn.openfoundry.org/pugs/perl5/Perl6-MetaModel/>
139
140 Parrot now uses C3
141 <http://aspn.activestate.com/ASPN/Mail/Message/perl6-internals/2746631>
142 <http://use.perl.org/~autrijus/journal/25768>
143
144 Python 2.3 MRO related links
145 <http://www.python.org/2.3/mro.html>
146 <http://www.python.org/2.2.2/descrintro.html#mro>
147
148 C3 for TinyCLOS
149 <http://www.call-with-current-continuation.org/eggs/c3.html>
150
152 Stevan Little, <stevan@iinteractive.com>
153
154 Brandon L. Black, <blblack@gmail.com>
155
157 Copyright 2006 by Infinity Interactive, Inc.
158
159 <http://www.iinteractive.com>
160
161 This library is free software; you can redistribute it and/or modify it
162 under the same terms as Perl itself.
163
164
165
166perl v5.30.1 2020-01-29 Algorithm::C3(3)