"Cannot infer a unique well-sorted term" error

[nolta]

Sorry if this is a naive question, but I'm wondering why the following schema throws an error:

schema S = literal : empty {
    entities
        A
        B
        C
        D
    foreign_keys
        x : A -> C
        x : B -> C
        y : A -> D
        y : B -> D
        y : C -> D
    path_equations
        A.x.y = A.y
        B.x.y = B.y
}

In equation A.x.y = A.y, Cannot infer a unique well-sorted term for y(x(v)) = y(v).
Candidates: FK v.x[class catdata.aql.exp.Fk].y[class catdata.aql.Term] = FK v[class catdata.aql.exp.Fk].y[class catdata.aql.Term]
FK v.x[class catdata.aql.exp.Fk].y[class catdata.aql.Term] = FK v[class catdata.aql.exp.Fk].y[class catdata.aql.Term]

I'm using the 2019-08-23 version of CQL.

[drever]

When you disambiguate the foreign keys x to e.g. x1 and x2 it compiles. I'm not sure what you are going for with the path equations, but I'm still learning CQL myself.

[nolta]

If the path_equations syntax matched observation_equations, then I could understand the error:

path_equations:
    forall e. e.x.y = e.y

But superficially A.x.y = A.y doesn't seem ambiguous.

[wisnesky]

Path equations were indeed first being translated into observation equations and then type inference was being performed, which in turn was not taking the given source entity into account. We've been moving away from type inference in general since it can be confusing when the same name can appear in multiple tables or types, so to fix I removed type inference from this step altogether. The example is not ambiguous and is working in the latest jar.

[nolta]

Thanks!

I've only been playing with CQL for a few days, but it's a really exciting project.