Commit | Line | Data |
e1a479c5 |
1 | # mro.pm |
2 | # |
3 | # Copyright (c) 2007 Brandon L Black |
b2685f0c |
4 | # Copyright (c) 2008 Larry Wall and others |
e1a479c5 |
5 | # |
6 | # You may distribute under the terms of either the GNU General Public |
7 | # License or the Artistic License, as specified in the README file. |
8 | # |
9 | package mro; |
10 | use strict; |
11 | use warnings; |
12 | |
f58cd386 |
13 | # mro.pm versions < 1.00 reserved for MRO::Compat |
14 | # for partial back-compat to 5.[68].x |
b2685f0c |
15 | our $VERSION = '1.01'; |
e1a479c5 |
16 | |
17 | sub import { |
18 | mro::set_mro(scalar(caller), $_[1]) if $_[1]; |
19 | } |
20 | |
f58cd386 |
21 | package # hide me from PAUSE |
22 | next; |
23 | |
24 | sub can { mro::_nextcan($_[0], 0) } |
25 | |
26 | sub method { |
27 | my $method = mro::_nextcan($_[0], 1); |
28 | goto &$method; |
29 | } |
30 | |
31 | package # hide me from PAUSE |
32 | maybe::next; |
33 | |
34 | sub method { |
35 | my $method = mro::_nextcan($_[0], 0); |
36 | goto &$method if defined $method; |
37 | return; |
38 | } |
39 | |
b2685f0c |
40 | require XSLoader; |
41 | XSLoader::load('mro', $VERSION); |
42 | |
e1a479c5 |
43 | 1; |
44 | |
45 | __END__ |
46 | |
47 | =head1 NAME |
48 | |
49 | mro - Method Resolution Order |
50 | |
51 | =head1 SYNOPSIS |
52 | |
f58cd386 |
53 | use mro; # enables next::method and friends globally |
54 | |
2e4b0c4a |
55 | use mro 'dfs'; # enable DFS MRO for this class (Perl default) |
56 | use mro 'c3'; # enable C3 MRO for this class |
e1a479c5 |
57 | |
58 | =head1 DESCRIPTION |
59 | |
60 | The "mro" namespace provides several utilities for dealing |
61 | with method resolution order and method caching in general. |
62 | |
70cd14a1 |
63 | These interfaces are only available in Perl 5.9.5 and higher. |
64 | See L<MRO::Compat> on CPAN for a mostly forwards compatible |
65 | implementation for older Perls. |
66 | |
e1a479c5 |
67 | =head1 OVERVIEW |
68 | |
2e4b0c4a |
69 | It's possible to change the MRO of a given class either by using C<use |
70 | mro> as shown in the synopsis, or by using the L</mro::set_mro> function |
58d4c5df |
71 | below. |
f58cd386 |
72 | |
73 | The special methods C<next::method>, C<next::can>, and |
74 | C<maybe::next::method> are not available until this C<mro> module |
75 | has been loaded via C<use> or C<require>. |
e1a479c5 |
76 | |
77 | =head1 The C3 MRO |
78 | |
79 | In addition to the traditional Perl default MRO (depth first |
2e4b0c4a |
80 | search, called C<DFS> here), Perl now offers the C3 MRO as |
e1a479c5 |
81 | well. Perl's support for C3 is based on the work done in |
2e4b0c4a |
82 | Stevan Little's module L<Class::C3>, and most of the C3-related |
e1a479c5 |
83 | documentation here is ripped directly from there. |
84 | |
85 | =head2 What is C3? |
86 | |
2e4b0c4a |
87 | C3 is the name of an algorithm which aims to provide a sane method |
88 | resolution order under multiple inheritance. It was first introduced in |
89 | the language Dylan (see links in the L</"SEE ALSO"> section), and then |
90 | later adopted as the preferred MRO (Method Resolution Order) for the |
91 | new-style classes in Python 2.3. Most recently it has been adopted as the |
92 | "canonical" MRO for Perl 6 classes, and the default MRO for Parrot objects |
93 | as well. |
e1a479c5 |
94 | |
2e4b0c4a |
95 | =head2 How does C3 work |
e1a479c5 |
96 | |
2e4b0c4a |
97 | C3 works by always preserving local precendence ordering. This essentially |
98 | means that no class will appear before any of its subclasses. Take, for |
99 | instance, the classic diamond inheritance pattern: |
e1a479c5 |
100 | |
101 | <A> |
102 | / \ |
103 | <B> <C> |
104 | \ / |
105 | <D> |
106 | |
2e4b0c4a |
107 | The standard Perl 5 MRO would be (D, B, A, C). The result being that B<A> |
108 | appears before B<C>, even though B<C> is the subclass of B<A>. The C3 MRO |
109 | algorithm however, produces the following order: (D, B, C, A), which does |
110 | not have this issue. |
e1a479c5 |
111 | |
2e4b0c4a |
112 | This example is fairly trivial; for more complex cases and a deeper |
113 | explanation, see the links in the L</"SEE ALSO"> section. |
e1a479c5 |
114 | |
115 | =head1 Functions |
116 | |
2e4b0c4a |
117 | =head2 mro::get_linear_isa($classname[, $type]) |
e1a479c5 |
118 | |
2e4b0c4a |
119 | Returns an arrayref which is the linearized MRO of the given class. |
e1a479c5 |
120 | Uses whichever MRO is currently in effect for that class by default, |
2e4b0c4a |
121 | or the given MRO (either C<c3> or C<dfs> if specified as C<$type>). |
e1a479c5 |
122 | |
70cd14a1 |
123 | The linearized MRO of a class is an ordered array of all of the |
124 | classes one would search when resolving a method on that class, |
125 | starting with the class itself. |
126 | |
127 | If the requested class doesn't yet exist, this function will still |
128 | succeed, and return C<[ $classname ]> |
129 | |
2e4b0c4a |
130 | Note that C<UNIVERSAL> (and any members of C<UNIVERSAL>'s MRO) are not |
131 | part of the MRO of a class, even though all classes implicitly inherit |
e1a479c5 |
132 | methods from C<UNIVERSAL> and its parents. |
133 | |
2e4b0c4a |
134 | =head2 mro::set_mro($classname, $type) |
e1a479c5 |
135 | |
2e4b0c4a |
136 | Sets the MRO of the given class to the C<$type> argument (either |
e1a479c5 |
137 | C<c3> or C<dfs>). |
138 | |
2e4b0c4a |
139 | =head2 mro::get_mro($classname) |
e1a479c5 |
140 | |
2e4b0c4a |
141 | Returns the MRO of the given class (either C<c3> or C<dfs>). |
e1a479c5 |
142 | |
2e4b0c4a |
143 | =head2 mro::get_isarev($classname) |
e1a479c5 |
144 | |
145 | Gets the C<mro_isarev> for this class, returned as an |
70cd14a1 |
146 | arrayref of class names. These are every class that "isa" |
2e4b0c4a |
147 | the given class name, even if the isa relationship is |
148 | indirect. This is used internally by the MRO code to |
149 | keep track of method/MRO cache invalidations. |
e1a479c5 |
150 | |
151 | Currently, this list only grows, it never shrinks. This |
152 | was a performance consideration (properly tracking and |
153 | deleting isarev entries when someone removes an entry |
154 | from an C<@ISA> is costly, and it doesn't happen often |
155 | anyways). The fact that a class which no longer truly |
156 | "isa" this class at runtime remains on the list should be |
157 | considered a quirky implementation detail which is subject |
158 | to future change. It shouldn't be an issue as long as |
159 | you're looking at this list for the same reasons the |
160 | core code does: as a performance optimization |
161 | over having to search every class in existence. |
162 | |
163 | As with C<mro::get_mro> above, C<UNIVERSAL> is special. |
164 | C<UNIVERSAL> (and parents') isarev lists do not include |
165 | every class in existence, even though all classes are |
166 | effectively descendants for method inheritance purposes. |
167 | |
2e4b0c4a |
168 | =head2 mro::is_universal($classname) |
e1a479c5 |
169 | |
170 | Returns a boolean status indicating whether or not |
171 | the given classname is either C<UNIVERSAL> itself, |
172 | or one of C<UNIVERSAL>'s parents by C<@ISA> inheritance. |
173 | |
174 | Any class for which this function returns true is |
175 | "universal" in the sense that all classes potentially |
176 | inherit methods from it. |
177 | |
178 | For similar reasons to C<isarev> above, this flag is |
179 | permanent. Once it is set, it does not go away, even |
180 | if the class in question really isn't universal anymore. |
181 | |
2e4b0c4a |
182 | =head2 mro::invalidate_all_method_caches() |
e1a479c5 |
183 | |
184 | Increments C<PL_sub_generation>, which invalidates method |
185 | caching in all packages. |
186 | |
2e4b0c4a |
187 | =head2 mro::method_changed_in($classname) |
e1a479c5 |
188 | |
2e4b0c4a |
189 | Invalidates the method cache of any classes dependent on the |
70cd14a1 |
190 | given class. This is not normally necessary. The only |
191 | known case where pure perl code can confuse the method |
192 | cache is when you manually install a new constant |
193 | subroutine by using a readonly scalar value, like the |
194 | internals of L<constant> do. If you find another case, |
195 | please report it so we can either fix it or document |
196 | the exception here. |
197 | |
198 | =head2 mro::get_pkg_gen($classname) |
199 | |
200 | Returns an integer which is incremented every time a |
201 | real local method in the package C<$classname> changes, |
202 | or the local C<@ISA> of C<$classname> is modified. |
203 | |
204 | This is intended for authors of modules which do lots |
205 | of class introspection, as it allows them to very quickly |
206 | check if anything important about the local properties |
207 | of a given class have changed since the last time they |
208 | looked. It does not increment on method/C<@ISA> |
209 | changes in superclasses. |
210 | |
211 | It's still up to you to seek out the actual changes, |
212 | and there might not actually be any. Perhaps all |
213 | of the changes since you last checked cancelled each |
214 | other out and left the package in the state it was in |
215 | before. |
216 | |
217 | This integer normally starts off at a value of C<1> |
218 | when a package stash is instantiated. Calling it |
219 | on packages whose stashes do not exist at all will |
220 | return C<0>. If a package stash is completely |
221 | deleted (not a normal occurence, but it can happen |
222 | if someone does something like C<undef %PkgName::>), |
223 | the number will be reset to either C<0> or C<1>, |
224 | depending on how completely package was wiped out. |
e1a479c5 |
225 | |
226 | =head2 next::method |
227 | |
228 | This is somewhat like C<SUPER>, but it uses the C3 method |
229 | resolution order to get better consistency in multiple |
230 | inheritance situations. Note that while inheritance in |
231 | general follows whichever MRO is in effect for the |
232 | given class, C<next::method> only uses the C3 MRO. |
233 | |
234 | One generally uses it like so: |
235 | |
236 | sub some_method { |
237 | my $self = shift; |
e1a479c5 |
238 | my $superclass_answer = $self->next::method(@_); |
239 | return $superclass_answer + 1; |
240 | } |
241 | |
242 | Note that you don't (re-)specify the method name. |
243 | It forces you to always use the same method name |
244 | as the method you started in. |
245 | |
246 | It can be called on an object or a class, of course. |
247 | |
248 | The way it resolves which actual method to call is: |
249 | |
2e4b0c4a |
250 | =over 4 |
251 | |
252 | =item 1 |
253 | |
254 | First, it determines the linearized C3 MRO of |
e1a479c5 |
255 | the object or class it is being called on. |
256 | |
2e4b0c4a |
257 | =item 2 |
258 | |
259 | Then, it determines the class and method name |
e1a479c5 |
260 | of the context it was invoked from. |
261 | |
2e4b0c4a |
262 | =item 3 |
263 | |
264 | Finally, it searches down the C3 MRO list until |
e1a479c5 |
265 | it reaches the contextually enclosing class, then |
266 | searches further down the MRO list for the next |
267 | method with the same name as the contextually |
268 | enclosing method. |
269 | |
2e4b0c4a |
270 | =back |
271 | |
e1a479c5 |
272 | Failure to find a next method will result in an |
273 | exception being thrown (see below for alternatives). |
274 | |
275 | This is substantially different than the behavior |
2e4b0c4a |
276 | of C<SUPER> under complex multiple inheritance. |
277 | (This becomes obvious when one realizes that the |
e1a479c5 |
278 | common superclasses in the C3 linearizations of |
279 | a given class and one of its parents will not |
2e4b0c4a |
280 | always be ordered the same for both.) |
e1a479c5 |
281 | |
2e4b0c4a |
282 | B<Caveat>: Calling C<next::method> from methods defined outside the class: |
e1a479c5 |
283 | |
2e4b0c4a |
284 | There is an edge case when using C<next::method> from within a subroutine |
285 | which was created in a different module than the one it is called from. It |
286 | sounds complicated, but it really isn't. Here is an example which will not |
287 | work correctly: |
e1a479c5 |
288 | |
289 | *Foo::foo = sub { (shift)->next::method(@_) }; |
290 | |
2e4b0c4a |
291 | The problem exists because the anonymous subroutine being assigned to the |
292 | C<*Foo::foo> glob will show up in the call stack as being called |
293 | C<__ANON__> and not C<foo> as you might expect. Since C<next::method> uses |
294 | C<caller> to find the name of the method it was called in, it will fail in |
295 | this case. |
296 | |
297 | But fear not, there's a simple solution. The module C<Sub::Name> will |
298 | reach into the perl internals and assign a name to an anonymous subroutine |
299 | for you. Simply do this: |
e1a479c5 |
300 | |
e1a479c5 |
301 | use Sub::Name 'subname'; |
302 | *Foo::foo = subname 'Foo::foo' => sub { (shift)->next::method(@_) }; |
303 | |
304 | and things will Just Work. |
305 | |
306 | =head2 next::can |
307 | |
2e4b0c4a |
308 | This is similar to C<next::method>, but just returns either a code |
309 | reference or C<undef> to indicate that no further methods of this name |
310 | exist. |
e1a479c5 |
311 | |
312 | =head2 maybe::next::method |
313 | |
2e4b0c4a |
314 | In simple cases, it is equivalent to: |
e1a479c5 |
315 | |
76051f89 |
316 | $self->next::method(@_) if $self->next::can; |
e1a479c5 |
317 | |
318 | But there are some cases where only this solution |
2e4b0c4a |
319 | works (like C<goto &maybe::next::method>); |
e1a479c5 |
320 | |
2e4b0c4a |
321 | =head1 SEE ALSO |
e1a479c5 |
322 | |
323 | =head2 The original Dylan paper |
324 | |
325 | =over 4 |
326 | |
327 | =item L<http://www.webcom.com/haahr/dylan/linearization-oopsla96.html> |
328 | |
329 | =back |
330 | |
331 | =head2 The prototype Perl 6 Object Model uses C3 |
332 | |
333 | =over 4 |
334 | |
335 | =item L<http://svn.openfoundry.org/pugs/perl5/Perl6-MetaModel/> |
336 | |
337 | =back |
338 | |
339 | =head2 Parrot now uses C3 |
340 | |
341 | =over 4 |
342 | |
343 | =item L<http://aspn.activestate.com/ASPN/Mail/Message/perl6-internals/2746631> |
344 | |
345 | =item L<http://use.perl.org/~autrijus/journal/25768> |
346 | |
347 | =back |
348 | |
349 | =head2 Python 2.3 MRO related links |
350 | |
351 | =over 4 |
352 | |
353 | =item L<http://www.python.org/2.3/mro.html> |
354 | |
355 | =item L<http://www.python.org/2.2.2/descrintro.html#mro> |
356 | |
357 | =back |
358 | |
359 | =head2 C3 for TinyCLOS |
360 | |
361 | =over 4 |
362 | |
363 | =item L<http://www.call-with-current-continuation.org/eggs/c3.html> |
364 | |
365 | =back |
366 | |
367 | =head2 Class::C3 |
368 | |
369 | =over 4 |
370 | |
371 | =item L<Class::C3> |
372 | |
373 | =back |
374 | |
375 | =head1 AUTHOR |
376 | |
377 | Brandon L. Black, E<lt>blblack@gmail.comE<gt> |
378 | |
379 | Based on Stevan Little's L<Class::C3> |
380 | |
381 | =cut |