QVT — 7.13 Implicit Variable Declarations

The RelationsToCore example declares only those variables that are necessary. It is
possible to deduce the types of variables not declared explicitly from the property
with which the variable is associated in a property template, or from the relation
whose invocation the variable participates in.

Are QVTrelation tools required to deduce implicit declarations?

If so, what are the detailed semantics for resolution of conflicting inference between
explicit and implicit declarations and between implicit and implicit declarations.

Do these semantics depend on collection kind, enforcement direction and relation
invocation binding?

Note that invocation of a relation with ridiculous types is probably but not
necessarily an error. In some circumstances the user may not be aware that a match
is impossible. When using third party models, the user perhaps should not be aware
that a match is impossible. The ridiculous match may be a precautionary cover for
a hoped for impossible case.

Add the following sub section in section 7.13 Concrete Syntax:

Implicit declaration of variables:

There is no need to explicitly declare variables that are bound to object or collection templates. All other variables must be explictly declared. Consider the following template declarations:
c:Class

Here there is no need to have explcit declaration for variables c and cs as their types are obvious from the template declarations.

If there exists an explcit declaration for a variable that is bound to an object or a collection template, then the explicitly declared type must either be the same or more generic than the type specified by the template. If an implictly declared variable has bindings with more than one object template (resp collection template), then the types of all such templates must be the same.