Changeset 2911 for ext/Perl-MetaClass/lib/Perl/MetaAssoc.pm

Timestamp:

05/10/05 02:16:17 (4 years ago)

Author:

mugwump

svk:copy_cache_prev:

4437

Message:

* explanatory essay in MetaAssoc?
* added missing clause on a Class.hs rule
* convert ∃ → ∀ in MetaClass?

Files:

• ext/Perl-MetaClass/lib/Perl/MetaAssoc.pm (modified) (3 diffs)

ext/Perl-MetaClass/lib/Perl/MetaAssoc.pm

@@ -7 +7 @@
 sub Perl::MetaAssoc::new(Str $class) returns Str is export {
     ($class.instance_isa('Perl::MetaClass'))
-            || die "Class must be a Perl::MetaClass instance";   
+            || die "Class must be a Perl::MetaClass instance";
     my $id = make_instance("Perl::MetaAssoc", { 
         'class'       => $class,   # Perl::MetaClass
@@ -13 +13 @@
         'range'       => [0, Inf], # Range
         'isComposite' => (0 == 1), # Bool
-        'ordered'     => (0 == 1), # Bool         
-        'keyed'       => (0 == 1), # Bool                 
-        'companion'   => undef,    # Str               
+        'ordered'     => (0 == 1), # Bool
+        'keyed'       => (0 == 1), # Bool
+        'companion'   => undef,    # Str
     });
     return $id;
 }
 
-# XXX - we need to follow these (?):
-# 
-#   ∃ MetaClass A, MetaAssoc C : A.clsAssoc ∋ C ↔ C.assocClass = A
-# (meta-class to meta-assoc is one-to-many, but meta-assoc to meta-class is one-to-one) <- am I right here?
-# 
-#   ∃ MetaAssoc C₁, C₂ : C₁.assocPair = C₂ ↔ C₂.assocPair = C₁
+
+#   ∀ MetaAssoc C₁, C₂ : C₁.assocPair = C₂ ↔ C₂.assocPair = C₁
 # (this is just bi-directional pairing of the two associations)
-# 
+#
 # -- can't be composite both ways
-# 
-# ∃ MetaAssoc C₁, C₂ : C₁.assocPair = C₂ ∧ C₁.assocIsComposite
-#      → ¬(C₂.catIsComposite)
-# 
+#
+#
 # -- this seems the simplest way to specify complementary categories
-# 
-# ∃ MetaAssoc C₁, C₂, MetaClass M₁, M₂
-#    : C₁.assocPair = C₂ ∧ C₁.assocClass = M₁ ∧ C₂.assocClass = M₂
-#    → (   ∃ M₁.clsAssoc{C₂.assocCompanion}
-#        ∧ ∃ M₂.clsAssoc{C₁.assocCompanion}
-#        ∧ M₁.clsAssoc{C₂.assocCompanion}[1] = C₁
-#        ∧ M₂.clsAssoc{C₁.assocCompanion}[1] = C₂
-#        ∧ M₁.clsAssoc{C₂.assocCompanion}[0] = M₂.clsAssoc{C₁.assocCompanion}[0]
-#        )
+#
+#
+# This is the rule that 
 
 
@@ -96 +84 @@
   use Perl::MetaAssoc;
 
+  # represent a one to many relationship between the
+  # "Property" and "Class" MetaClasses.  The association is
+  # called .properties on the Class end, and .class on the
+  # property end.
+
+  my $Property_mc = Perl::MetaClass->new("Property");
+  my $Class_mc    = Perl::MetaClass->new("Class");
+
+  $Class_mc->clsAssocs
+       (properties => Perl::MetaAssoc->new
+           (
+             assocOrdered => false,
+             assocRange => [0, inf],
+             assocCompanion => "class",
+             assocIsComposite => true,
+             assocPair => Perl::MetaAssoc->new
+                             ( assocRange => [1, 1],
+                               assocClass => $Property_mc )
+           )
+       );
+
 =head1 DESCRIPTION
+
+A Perl::MetaAssoc represents a fairly abstract concept.  Firstly, it
+is a class of the B<meta-model>, which is not C<Class.meta()> objects
+at all.  That distinction is explained at L<Perl::MetaClass>.
+
+This class is used to describe I<relationships> between two classes in
+the Class model.  For instance, a Class can have any number of defined
+methods, and each method can only belong to one class.  These two
+comments are frequently viewed as being seperate relationships, but in
+fact they are simply different ways of looking at the same
+relationship; that a method belongs to exactly one class (or package,
+or whatever).
+
+Each side of the relationship is identical to the other; so, in this
+metamodel, relationships are represented with I<two> objects - one for
+each side.
+
+There is a lot of implicit 'magic' assumed to be happening in the
+example.  This magic stems from rules such as the following;
+
+   ∀ MetaClass A, MetaAssoc C : A.clsAssoc ∋ C ↔ C.assocClass = A
+
+This is read as "For every MetaClass A and MetaAssoc C, such that
+A.clsAssoc contains C, it is automatically implied that C.assocClass
+is A".
+
+We could have written this explicitly:
+
+  my $properties_class_ma = Perl::MetaAssoc->new();
+  $Class_mc->clsAssocs("properties" => $properties_class_ma);
+  $properties_class_ma->assocClass($Class_mc);
+
+However, this would be needless duplication of expressing the
+relationship; we want this all to happen automatically.
+
+This rule is almost identical:
+
+  ∀ MetaAssoc C₁, C₂ : C₁.assocPair = C₂ ↔ C₂.assocPair = C₁
+
+Just by stating that the C<assocPair> for one of the C<MetaAssoc>
+objects is the other, we don't need to specify the reverse relation.
+
+Similarly, as we have stated;
+
+  ∀ MetaAssoc C₁, C₂, MetaClass M₁, M₂
+     :   C₁.catPair = C₂  ∧ C₁.assocCompanion
+       ∧ C₁.catClass = M₁ ∧ C₂.catClass = M₂
+     → (   ∃ M₁.clsAssocs{C₂.catCompanion}
+         ∧ ∃ M₂.clsAssocs{C₁.catCompanion}
+         ∧ M₁.clsAssocs{C₂.catCompanion}[1] = C₁
+         ∧ M₂.clsAssocs{C₁.catCompanion}[1] = C₂
+         ∧ M₁.clsAssocs{C₂.catCompanion}[0] = M₂.clsAssocs{C₁.catCompanion}[0]
+         )
+
+Then simply by specifying that the C<properties> MetaAssoc object's
+companion is called "class", this will mean that its pair MetaAssoc
+object's name within the "Property" clsAssocs collection is "class".
+It also implies that the pair MetaAssoc will have a C<companion>
+property of "properties".
+
+Finally, it is implied that the other side of the association cannot
+be composite;
+
+  ∀ MetaAssoc C₁, C₂ : C₁.assocPair = C₂ ∧ C₁.assocIsComposite
+     → ¬(C₂.catIsComposite)
 
 =head1 AUTHORS