Method Ambiguity when an instance type defines the theories operations

[jpfairbanks]

@epatters, we previously discussed the case where an instance already defines the term constructor methods from the theory. I think we wanted the @instance macro to give a warning if you didn't define it in the instance. The current behavior I am seeing on Catlab v0.6 is that if you don't define the method (for example compose) you get an error when you evaluate the @instance macro that says the instance does not define the compose function. But if you do define it as

@instance Epidemiology(OpenPetriDom, OpenPetri) begin
  compose(f::OpenPetri, g::OpenPetri) = OpenPetris.compose(f,g)
  otimes(f::OpenPetri, g::OpenPetri) = OpenPetris.otimes(f,g)
end

you get a runtime error of method ambiguity, because Julia doesn't know if you want to call Main.compose(f,g) or OpenPetris.compose(f,g).

Is this a user error, or did we not resolve this?

[epatters]

IIRC, @mehalter resolved this problem by adding the @import keyword for methods that have already been defined elsewhere. So you could write

@instance Epidemiology(OpenPetriDom, OpenPetri) begin
  @import compose, otimes
end

[jpfairbanks]

I had forgotten about that and thought that the import statement brought the unicode aliases in from the theory definition. Definitely user error.