Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

[svn:parrot-pdd] r17711 - trunk/docs/pdds/draft

5 views
Skip to first unread message

jona...@cvs.perl.org

unread,
Mar 24, 2007, 11:13:08 AM3/24/07
to perl6-i...@perl.org
Author: jonathan
Date: Sat Mar 24 08:13:06 2007
New Revision: 17711

Modified:
trunk/docs/pdds/draft/pdd15_objects.pod

Log:
[PDD15]: Incorporate some changes and updates. Also add a proposal for role conflict resolution, based upon the traits paper and discussion with allison++.

Modified: trunk/docs/pdds/draft/pdd15_objects.pod
==============================================================================
--- trunk/docs/pdds/draft/pdd15_objects.pod (original)
+++ trunk/docs/pdds/draft/pdd15_objects.pod Sat Mar 24 08:13:06 2007
@@ -145,6 +145,7 @@
a class. (Parrot may, at some point, allow imposition of runtime restrictions
on a class, but currently it doesn't)

+
=head2 Class PMC API

There are two PMC classes, C<Class> and C<Object>. Class PMCs hold all
@@ -271,13 +272,21 @@
=item attributes

An accessor for the attributes of the class. It returns the a Hash of
-all attributes, with a key of the attribute name, and a value of the
-Attribute object. The accessor is read-only.
+all attributes, with a key of the attribute name and a value of a Hash
+containing the attribute's metadata. The accessor is read-only.

=item add_attribute

-Adds a single attribute to the class. It takes a simple string name, and
-a simple string value for type.
+Adds a single attribute to the class. It takes a simple string name and,
+optionally, a simple string value for type. {{ Conjectural: Actually, the
+simple string value alone won't cut it. We need to take a key there too? }}
+
+=item methods
+
+An accessor for the methods of the class. It returns a Hash of all methods,
+with a key of the method name and a value of an invokable PMC. Note that the
+methods list includes any methods that were composed into the class from
+roles.

=item parents

@@ -286,7 +295,7 @@

=item add_parent

-Adds a single parent to the class. It takes a simple string name.
+Adds a single parent to the class. It takes an instance of the Class PMC.

=item roles

@@ -295,7 +304,9 @@

=item add_role

-Adds a single role to the class. It takes a simple string name.
+Adds a single role to the class. It takes an instance of the Role PMC as a
+required positional parameter, and the optional named parameters C<without>
+and C<alias>; see L<Role Conflict Resolution> for more details.

=back

@@ -327,6 +338,7 @@
and no Namespace object referencing it. When a class is garbage
collected, it should remove itself from the registry.

+
=head2 Object PMC API

C<Object> PMCs are the actual objects, and hold all the per-object
@@ -372,15 +384,17 @@
variable.

In a static language like C#.Net:
-B isa A
-A o1 = new B();
-B o2 = new B();

-o1.x; # retrieves A's attribute
-o2.x; # retrieves B's attribute
+ B isa A
+ A o1 = new B();
+ B o2 = new B();
+
+ o1.x; # retrieves A's attribute
+ o2.x; # retrieves B's attribute
+
+ o1.foo(); # call's B's method
+ o2.foo(); # call's B's method

-o1.foo(); # call's B's method
-o2.foo(); # call's B's method

=head2 Role PMC API

@@ -617,6 +631,45 @@
use of objects, though most class activity (including creation of subclasses
and modifications of existing classes) should be transparent as well.

+
+=head2 Role Conflict Resolution
+
+{{ Conjectural:
+
+When a role is added to a class, we try to compose it right away, and flag up
+any conflicts that are detected. A conflict occurs if two roles try to supply
+a method of the same name (but see the note on multi-methods below). High
+level languaes will provide varying facilities to deal with this, and Parrot
+provides the primitives to implement them.
+
+When adding a role to a class, you can optionally supply an array of method
+names from the role to exclude from the composition process. This is done
+using the named parameter C<without>. It is not an error to list a method name
+in this array that the role does not have. This makes it possible to implement
+languages that provide for explicit exclusions on a role-by-role basis as well
+as Perl 6, which can supply an array of all method names in the class (since
+writing a method in the class resolves any conflicts between methods of the
+same name from roles, and automatically overrides it).
+
+The second primitive is aliasing. When adding a role to a class, the optional
+C<alias> named parameter can be supplied, with a value of a hash of strings to
+strings. The key of this hash is a method name in the role, and the value is
+the name we should give it when we compose it in to the class. This will allow
+Parrot to support languages that allow resolution of conflicts in this form.
+
+It is fine to use both C<without> and C<alias> with a single role. However,
+placing a method name in the exclude list and also aliasing it has undefined
+behaviour.
+
+If a method in a role is a MultiSub PMC and there is either no method of that
+name yet OR what is in the method slot with that name is also a MultiSub PMC,
+there will be no error. Instead, the multi-methods from the role will be added
+to the multi-methods of the MultiSub PMC already in the class. Any attempt to
+combine a multi with a non-multi will result in an error.
+
+}}
+
+
=head1 EXAMPLES

The following examples all assume we're working with basic Object objects

Nicholas Clark

unread,
Mar 26, 2007, 6:13:07 PM3/26/07
to perl6-i...@perl.org
On Sat, Mar 24, 2007 at 08:13:08AM -0700, jona...@cvs.perl.org wrote:

> +It is fine to use both C<without> and C<alias> with a single role. However,
> +placing a method name in the exclude list and also aliasing it has undefined
> +behaviour.

Is it possible to detect this and merely make it a fatal error?

Nicholas Clark

Jonathan Worthington

unread,
Mar 26, 2007, 6:59:39 PM3/26/07
to perl6-i...@perl.org
Yes, it is easily possible. Now you mention it, that probably makes more
sense. I'll update the proposal to say that (and implementation - what's
in that section of the PDD is implemented now so it can be played with),
if I hear no other objections.

What should also be said is that this section is still my suggestion
rather than The Final Decision (though I believe it's along the right
lines), so it could all change.

Thanks for the feedback!

Jonathan

0 new messages