Standards » TMCL

Consider the case where we have cities contained in provinces
contained in countries:

place isa tmcl:topic-type;
is-abstract();
plays-role(containee, contained-in, 0, 1);
plays-role(container, contained-in, 0, *).

city ako place.
province ako place.
country ako place.

contained-in isa tmcl:association-type;
has-role(containee, 1, 1);
has-role(container, 1, 1);
constrain-other-role(containee, city, container, province);
constrain-other-role(containee, province, container, country).

As given, this would allow the following:
contained-in(containee: some-country, container: some-city)
contained-in(containee: some-country, container: some-province)

This can be solved by specifying plays-role for the subtypes, but if
so we can recreate the problem by adding another level (continent). Of
course, it can still be solved by adding the necessary
constrain-other-role statements for the container roles.

However, another alternative would be to make this symmetrical (ie: in
a sense it constrains both role types at the same time) by turning it
into a declaration. The idea is: if the association type has at least
one role-combination-constraint then only the combinations listed are
allowed.

With this approach the association type could be rewritten as:

contained-in isa tmcl:association-type;
has-role(containee, 1, 1);
has-role(container, 1, 1);
role-combination(containee, city, container, province);
role-combination(containee, province, container, country).

This would now suffice to specify the association type.

Note that the issue we have with the regular-expression-constraint,
reifier-constraint etc would with this solution not force us to create
common supertypes but could instead be solved simply by enumerating
the cases.